From 01d105ed7da3e4a033fad3923f01def537c5a362 Mon Sep 17 00:00:00 2001
From: IndrajeetPatil <IndrajeetPatil@users.noreply.github.com>
Date: Sun, 7 Jan 2024 00:48:09 +0000
Subject: [PATCH] =?UTF-8?q?Deploying=20to=20gh-pages=20from=20@=20Indrajee?=
 =?UTF-8?q?tPatil/Advanced-R-exercises@d118af9badc3bf2c77c0d2ae3e0bb78cf35?=
 =?UTF-8?q?ffa27=20=F0=9F=9A=80?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 404.html                                      |   2 +-
 Conditions.md                                 |   4 +-
 Control-flow.md                               |   2 +-
 Environments.md                               |  16 ++---
 Evaluation.md                                 |  12 ++--
 Expressions.md                                |   4 +-
 Function-factories.md                         |  50 ++++++-------
 .../figure-html/Function-factories-28-1.png   | Bin 51560 -> 53440 bytes
 Function-operators.md                         |  16 ++---
 Functionals.md                                |  14 ++--
 Functions.md                                  |  16 ++---
 Names-values.md                               |  64 ++++++++---------
 .../figure-html/Names-values-31-1.png         | Bin 62192 -> 62664 bytes
 Perf-improve.md                               |  48 ++++++-------
 .../figure-html/Perf-improve-28-1.png         | Bin 50778 -> 52143 bytes
 Perf-measure.md                               |  16 ++---
 Quotation.md                                  |   6 +-
 R6.md                                         |   6 +-
 Rcpp.md                                       |  62 ++++++++--------
 S3.md                                         |  12 ++--
 S4.md                                         |   2 +-
 Subsetting.md                                 |   2 +-
 Translation.md                                |   2 +-
 Vectors.md                                    |   6 +-
 base-types.html                               |   2 +-
 big-picture.html                              |   2 +-
 conditions.html                               |   6 +-
 control-flow.html                             |   4 +-
 debugging.html                                |   2 +-
 environments.html                             |  18 ++---
 evaluation.html                               |  14 ++--
 expressions.html                              |   6 +-
 function-factories.html                       |  52 +++++++-------
 function-operators.html                       |  18 ++---
 functionals.html                              |  16 ++---
 functions.html                                |  18 ++---
 improving-performance.html                    |  50 ++++++-------
 index.html                                    |   2 +-
 index.md                                      |   2 +-
 introduction.html                             |   2 +-
 measuring-performance.html                    |  18 ++---
 names-and-values.html                         |  66 +++++++++---------
 quasiquotation.html                           |   8 +--
 r6.html                                       |   8 +--
 rewriting-r-code-in-c.html                    |  64 ++++++++---------
 s3.html                                       |  14 ++--
 s4.html                                       |   4 +-
 search.json                                   |   2 +-
 subsetting.html                               |   4 +-
 trade-offs.html                               |   2 +-
 translation.html                              |   4 +-
 vectors.html                                  |   8 +--
 52 files changed, 389 insertions(+), 389 deletions(-)

diff --git a/404.html b/404.html
index 51f124d9..a30e101f 100644
--- a/404.html
+++ b/404.html
@@ -101,7 +101,7 @@ <h1>Page not found<a class="anchor" aria-label="anchor" href="#page-not-found"><
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/Conditions.md b/Conditions.md
index bb4b9f98..d0a3cb0c 100644
--- a/Conditions.md
+++ b/Conditions.md
@@ -292,7 +292,7 @@ rlang::catch_cnd
 #>         return(NULL)
 #>     })))
 #> }
-#> <bytecode: 0x564fa21357a8>
+#> <bytecode: 0x55f2c19fa068>
 #> <environment: namespace:rlang>
 ```
 
@@ -691,7 +691,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Control-flow.md b/Control-flow.md
index 2d402f2a..6aa7b7e6 100644
--- a/Control-flow.md
+++ b/Control-flow.md
@@ -246,7 +246,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Environments.md b/Environments.md
index 4b0d2e00..f19f5aae 100644
--- a/Environments.md
+++ b/Environments.md
@@ -40,7 +40,7 @@ library(rlang)
 e <- env()
 e$loop <- e
 env_print(e)
-#> <environment: 0x558a72b692e8>
+#> <environment: 0x5592b255c2e8>
 #> Parent: <environment: global>
 #> Bindings:
 #> • loop: <env>
@@ -51,8 +51,8 @@ The binding `loop` should have the same memory address as the environment `e`:
 
 ```r
 lobstr::ref(e$loop)
-#> █ [1:0x558a72b692e8] <env> 
-#> └─loop = [1:0x558a72b692e8]
+#> █ [1:0x5592b255c2e8] <env> 
+#> └─loop = [1:0x5592b255c2e8]
 ```
 
 **Q3.** Create a pair of environments as illustrated by this picture.
@@ -71,9 +71,9 @@ e2$deloop <- e1
 
 # following should be the same
 lobstr::obj_addrs(list(e1, e2$deloop))
-#> [1] "0x558a751c8be8" "0x558a751c8be8"
+#> [1] "0x5592b4bbb1e8" "0x5592b4bbb1e8"
 lobstr::obj_addrs(list(e2, e1$loop))
-#> [1] "0x558a75216348" "0x558a75216348"
+#> [1] "0x5592b4c08948" "0x5592b4c08948"
 ```
 
 **Q4.** Explain why `e[[1]]` and `e[c("a", "b")]` don't make sense when `e` is an environment.
@@ -300,7 +300,7 @@ fget("mean", inherits = FALSE)
 fget("mean", inherits = TRUE)
 #> function (x, ...) 
 #> UseMethod("mean")
-#> <bytecode: 0x558a7059e5a8>
+#> <bytecode: 0x5592aff915a8>
 #> <environment: namespace:base>
 
 mean <- 5
@@ -506,7 +506,7 @@ rlang::caller_env
 #> {
 #>     parent.frame(n + 1)
 #> }
-#> <bytecode: 0x558a710bd8d0>
+#> <bytecode: 0x5592b0ab08d0>
 #> <environment: namespace:rlang>
 ```
 
@@ -559,7 +559,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Evaluation.md b/Evaluation.md
index 93469a5a..83d167ca 100644
--- a/Evaluation.md
+++ b/Evaluation.md
@@ -49,7 +49,7 @@ withr::with_tempdir(
     foo()
   }
 )
-#> <environment: 0x56269fbe7e18>
+#> <environment: 0x564fa1bed6a8>
 #> Parent: <environment: global>
 ```
 
@@ -250,17 +250,17 @@ q1 <- new_quosure(expr(x), env(x = 1))
 q1
 #> <quosure>
 #> expr: ^x
-#> env:  0x56269fa39110
+#> env:  0x564fa1a3e9a0
 q2 <- new_quosure(expr(x + !!q1), env(x = 10))
 q2
 #> <quosure>
 #> expr: ^x + (^x)
-#> env:  0x5626a008bd48
+#> env:  0x564fa20915d8
 q3 <- new_quosure(expr(x + !!q2), env(x = 100))
 q3
 #> <quosure>
 #> expr: ^x + (^x + (^x))
-#> env:  0x5626a034a820
+#> env:  0x564fa23500b0
 ```
 
 **A1.** Correctly predicted 😉
@@ -298,7 +298,7 @@ enenv(x)
 
 foo <- function(x) enenv(x)
 foo()
-#> <environment: 0x5626a12b70c8>
+#> <environment: 0x564fa32bc708>
 ```
 
 ---
@@ -674,7 +674,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Expressions.md b/Expressions.md
index d5291c6c..e9a43887 100644
--- a/Expressions.md
+++ b/Expressions.md
@@ -374,7 +374,7 @@ call_standardise(quote(mean(x = 1:10, , TRUE)))
 mean
 #> function (x, ...) 
 #> UseMethod("mean")
-#> <bytecode: 0x55d34a3275a8>
+#> <bytecode: 0x562fc13895a8>
 #> <environment: namespace:base>
 ```
 
@@ -993,7 +993,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Function-factories.md b/Function-factories.md
index 9596903a..4e304a5c 100644
--- a/Function-factories.md
+++ b/Function-factories.md
@@ -21,7 +21,7 @@ library(ggplot2, warn.conflicts = FALSE)
 force
 #> function (x) 
 #> x
-#> <bytecode: 0x556cfd712ab0>
+#> <bytecode: 0x56479c5e3ab0>
 #> <environment: namespace:base>
 ```
 
@@ -47,8 +47,8 @@ f <- approxfun(x, y)
 f
 #> function (v) 
 #> .approxfun(x, y, v, method, yleft, yright, f, na.rm)
-#> <bytecode: 0x556d01983cb0>
-#> <environment: 0x556d01982ba0>
+#> <bytecode: 0x5647a0854cb0>
+#> <environment: 0x5647a0853ba0>
 f(x)
 #>  [1] -0.7786629 -0.3894764 -2.0337983 -0.9823731  0.2478901
 #>  [6] -2.1038646 -0.3814180  2.0749198  1.0271384  0.4730142
@@ -212,15 +212,15 @@ new_counter3()
 #>     i <- i + 1
 #>     i
 #>   }
-#> <environment: 0x556d01b28e68>
+#> <environment: 0x5647a09f9e68>
 
 new_counter3()
 #> function() {
 #>     i <- i + 1
 #>     i
 #>   }
-#> <bytecode: 0x556d01ced600>
-#> <environment: 0x556d01ba0c80>
+#> <bytecode: 0x5647a0bbe600>
+#> <environment: 0x5647a0a71c80>
 ```
 
 ---
@@ -258,7 +258,7 @@ ggplot2::label_bquote
 #>     }
 #>     structure(fun, class = "labeller")
 #> }
-#> <bytecode: 0x556d01e7f0b0>
+#> <bytecode: 0x5647a0d500b0>
 #> <environment: namespace:ggplot2>
 ```
 
@@ -281,7 +281,7 @@ scales::number_format
 #>             scale_cut = scale_cut, trim = trim, ...)
 #>     }
 #> }
-#> <bytecode: 0x556d020cfbc8>
+#> <bytecode: 0x5647a0fa0bc8>
 #> <environment: namespace:scales>
 ```
 
@@ -377,7 +377,7 @@ Let's have a look at one example with each:
 ```r
 boxcox2(1)
 #> function(x) (x^lambda - 1) / lambda
-#> <environment: 0x556cfdac76a0>
+#> <environment: 0x56479c998358>
 
 boxcox3(mtcars$wt)
 #> function(lambda) {
@@ -387,7 +387,7 @@ boxcox3(mtcars$wt)
 #>       (x^lambda - 1) / lambda
 #>     }
 #>   }
-#> <environment: 0x556cfefae4d0>
+#> <environment: 0x56479de7f188>
 ```
 
 As can be seen:
@@ -420,7 +420,7 @@ boot_permute(mtcars, "mpg")
 #>     col <- df[[var]]
 #>     col[sample(n, replace = TRUE)]
 #>   }
-#> <environment: 0x556d00bef850>
+#> <environment: 0x56479fac0b28>
 ```
 
 This is why we don't need to worry about a copy being made because the `df` in the function environment points to the memory address of the data frame. We can confirm this by comparing their memory addresses:
@@ -429,7 +429,7 @@ This is why we don't need to worry about a copy being made because the `df` in t
 ```r
 boot_permute_env <- rlang::fn_env(boot_permute(mtcars, "mpg"))
 rlang::env_print(boot_permute_env)
-#> <environment: 0x556d0181e450>
+#> <environment: 0x5647a06ef728>
 #> Parent: <environment: global>
 #> Bindings:
 #> • n: <int>
@@ -488,8 +488,8 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression      min   median `itr/sec` mem_alloc `gc/sec`
 #>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
-#> 1 LL1          27.5µs   29.3µs    33174.    12.9KB     43.2
-#> 2 LL2          14.6µs   15.3µs    62143.        0B     49.8
+#> 1 LL1          28.3µs   30.2µs    31979.    12.9KB     41.6
+#> 2 LL2          14.6µs   15.7µs    60147.        0B     48.2
 ```
 
 As can be seen, the second version is much faster than the first version.
@@ -515,16 +515,16 @@ generate_ll_benches <- function(n) {
 #> # A tibble: 10 × 5
 #>    length expression      min   median `itr/sec`
 #>     <dbl> <bch:expr> <bch:tm> <bch:tm>     <dbl>
-#>  1     10 LL1          38.9µs   40.9µs    23209.
-#>  2     10 LL2          17.2µs     18µs    54710.
-#>  3     20 LL1          42.1µs     44µs    22427.
-#>  4     20 LL2          16.9µs   17.7µs    55356.
-#>  5     50 LL1          46.2µs   47.9µs    20558.
-#>  6     50 LL2          16.2µs   17.2µs    57361.
-#>  7    100 LL1          61.9µs   63.6µs    15410.
-#>  8    100 LL2          17.6µs   18.4µs    53603.
-#>  9   1000 LL1         854.3µs  872.3µs     1092.
-#> 10   1000 LL2          54.6µs   55.9µs    17653.
+#>  1     10 LL1          39.9µs   42.1µs    22543.
+#>  2     10 LL2          17.3µs   18.3µs    53968.
+#>  3     20 LL1            43µs   44.6µs    22159.
+#>  4     20 LL2          17.1µs     18µs    54815.
+#>  5     50 LL1          46.8µs   48.3µs    20374.
+#>  6     50 LL2          16.3µs   17.4µs    56753.
+#>  7    100 LL1          62.4µs   64.1µs    15369.
+#>  8    100 LL2          17.6µs   18.4µs    53652.
+#>  9   1000 LL1         857.4µs    965µs     1081.
+#> 10   1000 LL2          54.6µs   56.2µs    17530.
 
 ggplot(
   df_bench,
@@ -648,7 +648,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Function-factories_files/figure-html/Function-factories-28-1.png b/Function-factories_files/figure-html/Function-factories-28-1.png
index 1e1412cd5f124986238fad69ee4b571fa3ab3aa4..1ba1f6ed6e8967ff12d611a630630a88147072d5 100644
GIT binary patch
literal 53440
zcmcG$bzD?i+de*^SfGeVsFb6Kk_spt>L5x>N(&+(CEX0@K?x}-k%j|^lz?<9C?x{Y
zjdXX{eD6JT&ikC_eb1l2AD?p+XU|@H#l7yhu4`={1v$x+C(fThp-?9u+{Y-QP$c3g
z)S;?l#PEsBVN3x0b^Q5#bsH3l{x$MHQQ-LUQxu8`^#F5Q*)e)<z$x_Xdimz^V)n;~
znb>oNdWQ`t58u{8qovNo4!<b%u6t8G!W0;|Y{9VjHhNaJY_LB%Dr7vmzosSx{`uzK
z?QlH`O#kiblgF}{(9+$Z65)EUdP>qGJ~|&<t!9ot#52HCxe&WKo1K;T%$?CWrF5oE
zMLSuznG!mNLIw4a4G`gftBQO$jQ_3f&{;?PCo3ot2ILnHVR6r+$S+Lf3jg1Jg=l~%
zDELWjfKzOjotPLcS$KH3ueW($W@e@~pMoCSwQD!XlG4&1I0U&;`QJ-w{owui^XJQF
zjt^L^s7Oe7nfK*9l7VX~8XD)xj!68OknqJ$8?L=Ic|Sfr-qh4oZ=mQ9?%Gl6V3`~`
z%J1**Uy?V*)|l(-=U3`X{?UVk*fTemo8rOcGaMWo8r)Ztoi`UUpBjso3SV6+FmH|#
za@gNlhi0l@oxC98Y-?wiZ}MJUt1#(qhEDm;axEivE~j>9eYVT|C9>8i=UgqWrS<i7
zv8^xU8*@FE(8<ZkngwP(xlbEYWTPr}2NuIaLk*q|$dN7gnr8PD3^e`EeYjJ5Nr9pz
zM$NH)ToqgOEOF61Mn6Xb#YBx;;sZr@rW`7FwF^4uTTbUZnuDwTxlg4XP9Y|T5|^Xd
zGVd!eA1o^_D%$$}nlURQ<81K3{w{XWWr#<1)t_&W-+9e=Vm5aFbav&z&Rud5UYCuz
zr;Xuueep;Gs4eS~s%*_dIn4I}dY2XTakt$K`kT)}C!J=yvYTB8@)-3O2g{F=GmyQm
zE_Ys6l(K97&wY!%`O=v0Z_e2?xMBlv+2^ucHtxEvMh!*p8tHD_T+blK_tAqZm2~gR
z%cBxCV)c6l2hoR)o-(MgxN>)_4&RbTEK_RsouSUoE9g|l1x`**>fnt<L)Y+GLvD=2
zsS?Yf3IUsmXi=OcZo~3Y@cq|!?7j6|4=CiX)2@vsP!fB3DEhK|Y9sb+9!g_xT(PY2
zms61Px|2Z3^~|>2r_Ys|ippwt>gRH$hoQRa)w921n02?C6ZB}!lmwfw95hr^)2H5M
z%b7_Md!n$f#Wx$LiQ}Y_F=J~l&imh!xEy>R-&NbmcvxBK2u=~VgG1!PkJVM78v!Qo
z`;$cP4y;D8ntgd3L>%WM&h&hw?X;~1KV2vN#md$}p43qNsJ9nmtA3dD3ydi#uNerq
zUHvY0k3O(gK5~&U$pYs-ZB+c<Q$hYsSy`D*%)Kl>e?4^jAbMqG<wsYOf%gKP?O0u~
zuHy+e!5C5ZJxAz3oli*88K0WzIR0UFiKK_s*X3~jqwMP8<BY66-}Djdk9O@ySIaiw
zHt8!^VA}pzRMahoO(-tzA?y=9T3FF%rlzaSn7b^Y;o+>wT3zNA7M(CfFoDLrz3uHa
z5Bq=KkZlv|4VAefv}4|8_E0J;Jp9?Y^DUgIq^|1fk>0$#ywTj1IXMq8uzg~-OCrfy
zV_w2$eVBAFMikeb^JLn#<nBeGv6_>s!p*b4JWoVTIthD5@;?lxtNm^A<B!=AVb2(Q
zeO<8_Ii+EtTCu)x8taMf6oaq0%bdtZJbx}PE;TMpk*ysTa67R-{N<&D!RUAY)E`5i
zO+viM=mmLK?Rd_iC6ACG)(l_L9?mYAVWn`|9B`1Y$kMd2S)FK#x|`6_p{9%PUht*A
z_T7tU&ZP(~>-xIjOBXI&V04+gQEP>P72g|U5qprVNO?tufBS%uO)X0+NB`T;7$KpO
zg&y+YZ&f|?XOTy>P=27p9W)!oxox!ECBJ#|rbJRJSUKz#p-7pH#UWg(*#30AnrU`c
zmfV$)Jfr5Uj~{2IN{ouT>tW8LqN1>{Q`TUY*?0=wn4%H1pL-N>aOCBec&z{FzJlT-
znyF!-p|r1B+uQ#Xx?kaYFX=#*T2n(q(Ah>ZTH)bCy}p7h(o<9(zkVs=IA5JSdD8W0
zg^f%bI~SLwg~i0P(3sA@{N90`P0R7{r&KgF4(klcM5Kp_55}X<_+X@^{ldwl(gJkS
zPF$rGa67Dn`6;s5uY0f(!8hpm<no!GVjClM^`6jcO6~`%adEPZB8~r;=(A;7V#I+P
zR)<Boo*x#yOXt~qBB091`_fY6=rPv`e&RR*!aqH9iro+6!wBlwU*V+e|F3ZJcia*e
zcR&^hwOwWBFTW|sNKaE6oT%TC8b`zNE9Ur=`qd@m>QAOmHZG1GjrVMncA9N`WqItT
z$9+t*HeXzfsrQKegP0t?6EE_`e^Dd5@^b7ZjSW3Z(X#ev%a^S+m#HQ3&0!&YGZmq&
z-H$Wf$J;j!j3yFZCWuw^l>2zZ`r+>%mL?H%l%z{7C%ILT`YX+%@z9z9p_jT1d$#hs
zq#tY>_#NY2zM@3v@Uv4D!QI>YQ$I4?;y$s*UrjhAEJa2f*Bl5sObEzT5!O3vBGqT4
z1SZ#J!`a2<CGp{TDLI;yOSGDJGG;L>)K;E0{&iB-AsEA!enY|CMRQ{#A}+du?DCwx
z6_kk9Uo%vE|J!Krx90P&Px*RZy(_#KN>ZkioYR-y-(bW0$06s-CxoE^52dt%&ieye
zc?aI!j$ag9XKcAnAG6%Q+ODve+@_PQ=rZgy?N>QMa`mkE=I`|*aK|v$hlv#90lK7|
z+ZbWCGwJVtZ!lg(r+vBW+t7gQus=15?C4qs2<vbV(P}H4yL2vTy2#tHHPKNv)8%I*
zCx$#;laOt={^VC>%#2FX(8ScRu?BRsmxQ_5z9wE0z;_NsL9tqwaDRY~D$vhcO@lf;
zay&e2_7ur5mXOYPJ54<{Sk7~IabR<<eP-yyT#npJgMYNRywSq$<y_kesVfpSnZhip
zJdJFuo%AQF(g>FWsk`&qG*nZ6nNV2m#9k<!weZ2%!QCW;%gt2X`PCW~_XPsH`41iW
zmToA;pCNSCgX<oCIh;QXcLrwAMEK15dh;0HFMC!(LpBZllE(MTBlc?1wA*Zc*oYJQ
zttgJzr9nIPfe_02Jl8|S>}rp@F9qb!JgqEGy{BU^WS{yjWkCV;lVA~R><K4Lt4&q6
z1C?mBpAA{RrhPSdANBJH-hifV-!UB`%hntflk;aV)y4W<P3Ak}k%uAl=}0qa<?^$s
z@{<S3D<7AA(phF?569mn42OKUq%)oEWt7-b>f6!pELU0vJ%#V$Z#xk)drwD)qC58i
z29s9P(Z!Oz_kKYJCHMw!pl5k&eab%ZetMGb?XBoDYata-Eah+o6?2<#Z@5%y#^emK
zi-V6h?JKvoKiip;U=CXz;I~2i^)zMHTwBlWX$A_4w$CiGvi`KmFo!=0rb}|a$KH0@
zb+k&9Vp*oGqMX(3i&x2AqB?>nxysAhNv+P?#c%xxQzm}h-N>KE;2lg8w*=wIzRyhk
zIp3SoiJ;Ls-M(r~&ySG(qCX){jW;WguL<u1Bxfhi*IvNX<e_suMq1#AkhV|Z%}Rmk
z&G#p*CB+|$A7ILiAC?)0yT~yPB%lmV;%E8l;qN&+VBj6S(02B(N3xlkgPIOkQ9Xot
zUWs+u%BxE(W@VtEC{#0X?G>?5FNU7w5@u|SZ?&qT?(H_Ko}aho!*gJ=VWCr}NSJmB
z3&Fs%>2^k4P?tFbJ)*kd)6woM@eO)5bp}6WOu8iYi=V^0(rh&_(eC?CI(V4RbilCG
z67aj->Dnio()B(4aenj&8691m;N7o!&~Mv|_+bTMUug1aa&gjhQBfEz1#PuY>s<d6
z<~xfpWh*kz+gv(LxqK&m?mKrM+YNZuY*5Q^)wS#=R0bZdXIydSrt56(F7n!rIc~FO
zZ^$AoA0@QRcjrRqtFDX{##F8yU$;hs8)W6dDBBB!i;Q=$t~`64=W$lg)*kKdXkxA*
z7g-1wSB1IuzNaT_-r7>Cw;H9Syn-0zE@7MWPk>Q!9p<xKOEq2;4<vNc`W@J#yXdF!
zyA%}3XS(`>=yp{GB2iNWi?a?J#ppab)14!WSgRc$ORZL#5GBf-Fu9zH)5-?wF~vp^
z-rjoMd;_s1W85%`DZ;{82S_nrlBRhdw?IL0L)PV2d%O2TnEIB#){XjUZIblFg`=B(
z-bU{?#%9)*f|09%_<b;Tn#$C#z3FBE12qK~?Hl>X2{Hj)un7TA!sQbko#!*7OxqbL
zD7ehNr5g$EADw|kS0k8XbUfZ1X=^V?fH^MDME)#31fD90usJ>4?^Wj@_V$Y%t@HV+
zsCl?jk0i8R?1gXJ*sz_^zq9F~c_w;foz=hNhjk+xa<dJ7%pP(Rznjs16CWM5F#W%B
zw?4b{|3bz*hQ`l-yuZ_9U4wezu>0QsWsQ;q2P*Qz&E=9b^#^4|Otk3!$jyA6%OAmF
z`Uy+Hdu;9?B;PF9`WJd+ZI&@T_S$C$)O;e|6bCqp_Me!J+l&xQG5h_sgdo@i*7O8N
zrRo_7rpN)N7`)c<dn5EbSCxf;2hV5>QcX>X{vCK((9?RC=7c_qgt%FwjJ)_r5&szV
z$Dj_4PF{oMVj3w|fn^LjB@OyiMlgVqyT;YMOYc<<M%&fCck>Nik(%~D%v4F3XU++y
z-(qD9&XXKiR=e%6jhZNvMU<c~ej06=n@O|Xnb!{b4wS9?w`%_Iai%1yQzCe(#-gnC
z2GTEqw|123m(^G*-17%9s1+K*Sk4DL$#ms};r^FFj@=*_xLbo&29-21w)YA;G1P7J
z&vrG=F4(c|1ShAS@++;*#%^Lz8;yQNQ8~S_#vB2Gf@zVk1I!eU3*{PAI6aQ@_BQDa
zNlB4>gLtuv@%TAbPnvVItrs30CPF-9tX2$zs3O<}$qW2|3|{-@OUz8rN(Op+(~x~%
zj9!T5Q-;f4FA0}D!9#X7-IxjV^)^z^N#p4!dI@Hupg`CvDQ_lH>b@Gi8TChxNPtO}
z4!G}w>sBy^;;@7#e)ZS%XT^`R?iGp+*U2cBg`4dLvMzl1wB3AB{&8x|mXNJ>L}7*;
zk8B8y^%b(}r6EiD+SVB6z{r$(QUS^olO6Icp*db`)Z2BI79r<uw-u{wNB};()tSKU
zUYK^Vuzy`gQT)kf%1MW*O7H%f!rt^`p#gVKGRffeH_!Rm(E+1>DyAV1^ty8~S(!Vs
zo&K^e2{92?_hO7(UURQW_|xS(KQa7t8$*^#X=nHIY(f=NOjh8d%V#JknpeA1Y~P!_
zPf)m8DjzTC{77V9)5M&D;u?pI;m~5}W;pRaeN4B2t+v|e&_xmk5o+wtHU~qX!?e2r
zl$}FKZPT`znwniFmAlERuamum`ZKZB_&V*`U<^9={$<<RdbQur`6Fe%>)#l-z0%=B
z{#HMSmWA6ED#q}qJq5DcbG(hYp^90vxw6yn+nodjcIw7XhERwaQ;9~*43rdwpV>v{
z8TjaezwVR3sOE)UOI^cNv&SDLY^@l*1spFVE|<@sM>ewFbo^NHl}1*sj<9mCBjXnQ
z*|p;pAK(+0%Jk0BPuD#-35ki|l~1Ox^IdcP+OpXdQ5&b?FA#Resbdy>`sGD3bds3p
z+mqi?e=gm*2%ci?7=8~7n~{hfv^Y;YS!q)|Ag8yGPg^cR3wkVk?K=sZ^114Z3q$!j
z*ayP$m|xYBy&>r$ZS7#G<ahAl_$T{kJ#)t<RtN#@%o~|XHP_ef5x2Avp3Af4x8dGH
zyZfpS9~suf3D0YE(%EY2y0e(InW`9}Ya(A3xp7LI6u+sKn1#8@cE($F7ylVGSO{B6
zx5h40P4%!?@0dJut!~KH&f#(&)Y@}Gqr3r6%0KZI5z-Y?OBtOTa*(cw5WLbBDeTor
znxpm>G!!?;cE9&2M(AV-okbD48S(+L6|;n5he8M*HI^kJZuL(g0C*7c5&qlEb9Ez5
z6JPex=#+DDyf;PkL&4wum8deI6XzTamW5aI4xvX#UTnnNoW58MYvzF?cv8%_UNq_C
zyLCxS4e60lrBtDpuOU33u=wcV5i56-Phxh0*8e1WL|rxGSxO%^9@>7+ik}Q;-lXDd
zxx*&^NTKC)VcmvF@%?VGId^iswM5`!vIM3qaim*2e?b6Aq5n!%#mS#C2%JreN*}pF
zu`Jc7e>2}g-3#e>GTxe|e16<Ln&db6K==yULShl_hMXT7j~Btu=NkXSfjnAzLYn&;
z+G4TcQ4@Q15e9q={~<!xsBJdoOndHx$@zJ=ZT4@$-Io(#LZi>&J*S9w>H4MFVy}h=
ze%?Hf70RByqA5UnmWMYQad|zn?d$oj!PayXR!2_y$j2;r-GK?O>cMY=c*$e#TNT?g
zQp>^=58}l7E0VsGe&Qixid%{WFu>z0&E7D*X-8vR;U+0pz>uJXj1u%~iWc9m6Iy=-
zi6>9~laa(z63|c&&a)c3uPNsT=xtP<@37zKK6)f5f*<jj9>n8Lzxi}Y%S#92r2LDe
zEXrR#S^PT;c-5_DbYveM!uG{U1zd=GcfB>S{7{_{ewxnGb@`X&qL*5)U|F}ZSr1&z
zUqJi_A`~$v{6-&fv7emlW<@6|kMv#$QG3%Ca~ze6&-WoyrK@#zw|YEjPeD=r$uc^$
z#P%~RPY^w!;bUP^65$-Yn+wzwH{NVDg|yLNC19L&=LzHV%$=Y5wf)`G4R4gnVGLFY
z$j!3&o5w_`0wt-%Oz8x#qIH;P;}l29=u=>I3<#^UEt{WzZ)V~>-Y8#<D5VMzjIsw`
zaf&l2HOYcm-Fx!V+fnth>doqTF$H9X35n_$Thg?n!Ec^>ZB`V^i4x22H8!IwVDbLS
zRGGNW)y^c@*(jU(c{@sc%XzO+m>UQ^Yw^NcU7grO@ZI1N0VIXa{>%=g83!8#O;iar
z-fo<}3)^1L-BPbCFQqDfax<^&tq#(31b+WB#l71V^<j_6ehdDOt=R{0kG>GOJQncC
zgZSXFxgHaabfxLf-u!OO{sVS}m_+1mK70zzBzD+-{!=)YV!$QzPmkNnA0yA4f`Sq?
z@EUIb!l(Opg4aj5!3QEHsU{160kRu3@!>jX_C?fSp^oOYSjbyexjv=#k4Mhc5>Ht<
zjzTRiE=T%rbzM9WeoVz|e|yN8MCp0Wh11n77mlNB37wMk>Z~}YQ}`VleO{!q;(+?0
zl;veqi#NVY{i9%$Iw#8uzr;!D+!VE1$AG&N7s9W<^_y9Z>?|^igDQeein|;|5ia=t
zYf%?ONtpg~0hCU9CXZ`X@fpoPD4k6hlzc&KgsaC{V{&8DN#_Qb(fYI-vLT8QZ=h~Q
zSrZ)L&c7CQs4;5%gL#z-=(Poh_&>>Ggxfw_W=rWl$6F*JS#5a$e$a2?yyPiAZ<=G@
z7c;D0k+mUBQ{d;Q_d`MFRATY>)=TISuS~}dGNURcn0+QfvKs&TnIHq?*mRbsDVlw2
zJ{p$nABDmSRZKu*PX2pM$CQd+x|V@APw-wcAG{=>AYhPXg~(Q!CAdM=e`L`*$;&32
zMNh{Ot6{)rV)480j&4YV|4=)wiRt%kfQnG<mo%~s$?+S_meOX3B{+Fi1@V$k<B-eB
z02ZLePU7dJ<nCkel1?#FsS-8QU1F*tO6Z5ku;K{*dz@IPP_vm!6|u;wFUx5>mjPft
zfm$KM&xtl$kzL;Uh&A3K$q^?>au?~?0p4=rNlEz1jmkF?9b>)!eC_{f@xz9U{cWOn
z9>>kYV3BIaD{uw*J5+!)aF%sf*!kj-hM;W1n9ux;G_=vUeXDqyi$@xQx(G&4(ZV=H
zR+<0bCRyizAFn~!q}Pz+&9%WJSXKcGSN*SuNycBt4<|?@%d~vw&hyoKSl^9PD0O`4
zB_3X6F*ZJVNK^DI#Y0xBrjzwKx*gyHPaMUMv1gb+m?r$Ny%{BPb)c#KhT6<ez{BwM
zkB3Z}Eo*MZ&;C4VOwIMVC<}RQ>quBR1!F=-z5lhS6DJ}Bk=0Oh$%r33rxs`^ntA8z
zJEnhMLVU$nLbq7b9%=q*k!7{^39zrnt_{3Kb=%0~ej(h<qpWHA*&^#97K17KM3o&o
zy~g+fmh~(*epv-&Zqf%rf*rsI3AV&Y2TzO*-X2K;@cTC4#`ypU>L0=l@ea}?FS<b#
zY5O<ci^U&$vy;>+VWWaL$Z{!ga{VhO5KkkCH@IQ5$k$?i?C;Y@tSGFc8%Np};7=rU
z#^Q5-k34;bZW?uNd1govRh7RZ&=)h(I3XRYij3UdxEIc@koG6|z@2ZV&&uy3mBC-&
zDF4f<1u^wT>T4_<Xt~ORoSDcwE=N5q2yWlyMW0xKswm4})(HOb1k@`i5yB4DWnEe}
zoJ}4dDM7zw5bTdt93|=YJd6@q8Q$+Im@l@O%+jr_C@yyD{-l>A8JLxobxW4VZ`RN>
zr?<CvVq(I6Wp8g%j8W9J$eUfOXsO32ZvX4kgW1Xh(<%=lqt#0AqEm|hr#9pv-(Vn$
zR~X;XeD3D9Pwu|-Qp9D$`@V_EbT|i|gEMJMS#;klWi~3AnO4t=8u->rv$e|7`T7+7
zOjRBvuPyU;EVD_o)yE4r+r4P|3gv?HgkBJ%rqIPY!;Efgt<ap7>+-jugN;HqHa2nw
z;jL=MgC!vCb_Sr`K-^ZKx;meqpU83zrS7{ipzaP1HY>5?VQQz&ao5C+(rm@;&l$Zu
za#HXq6bAh{lgEHf?fW0Zq=Nf<#quq@OZP|+5_c&8O3!I^Vq|3GHKTh`ZmuF`XLAu(
zzUbD>+XUeJ=u~S`WHt;UgG>MY1f{(2LoKbnszc<_-jg(2w_hHKsVLK*g(vW?lr2SN
z?_0W5?n(6up~k#{@R;q)qzuk^QXA_U;5<~}Ch32E+-VS4OdegnQa`tZl9H1B-OV8^
zZl8*Z>L3~lu9g-BRaHav&K_X|5Fe;n{hF?OynHoWER2qC5CYNP4VT)_p9#M4<a;#t
zP~0fa^wq4PLr$C9N<E7o4HpBC;g9?-9T)wH-_@^0nt}7s71{<!sU&JvgDd-3Ij#U&
z!`czN1d63`da;U8*iiwqr_?Dw(Jb(MSbPHC-2Mb#rIoe=u&XT1Yt3<7Y;2mj2K7;v
z6;4}AzZ)7F67IfMRZ)q-9&FF=r9y8@d`BO##gSUz9wAWBO}@*u?kbS$SLxPG6B7tW
z-O=)*qDL70oJaU4D4F#aSpi=(wDq0u)hT-C71zqDY~4zCAc8%`<JGctb~hJ?DzUf{
zyV-^E&B6K*mbinYQw+jI2K8^-CWqWM2U?BSNF|$3RB!Fk`HM3tC@Z(y9il1_XNnIU
zIX?H9ve5F_>uMqk(SCKQw8z)$<_Z-u!4mLVfFZrqm?9Hn-t+6i@|N#}$92b;=}sMe
zd3!*8B@9P*{#EdPm$hIfb+Z!IVsUU@<;x8<>aoRoEQiDNf|A~o$<r-tsSaLBX`$Xw
zgDOqzE*AS>(5rTNrQO$JohHszUpM^m&Z+^%LqkJ@+aJR112uQZQ(RNC?*$2IIG66h
zSctriZL7pV`G~wdOlNAG`_9_Mh2qKBHYGbRaVBxA28GSNloOi3p#r%{^76K?(nOK#
zgI7g8QWK$y>VqkNzD+ZVv&Mav1t(10!PCT+)iBZAEMgYQWAr{HD>YdDVQa{^Ei=&i
z0>EDiGl)H{z}A%RmuAO}m-J}Q8Gu;~fq&Rq9zA-B&apvE8@s=*re*aFC_nk=5M5<u
zTbSURw{B&r=cQz2EdbeVF<A+8L(olo;q`WGsf&Z3_dCPq-_1sUwU+P@p(-8$IqvK0
zJ6ao9SWplPcB!PJ6T6Ln7SQa8-ZIf!BoXfC1Z}6Tv2^TOluW1GH0_)lEZ-$n+pd>)
zx88BBwbIbgkc*4s0KCHJ+@Ffq`U*-d;=dZIhwIVKW2X<zAM?6%a8anQV82bT<@mR6
z-|7vxBV}fQQBG|AfSoWmHDx=B6|5Co7GelYZ8JfZ68oddP8-+g7O7+}d>ywrQ{_$(
zEY2ii=WJbfeY!!LbpFMc_^0aHCA-Sq)&`TG_Afb!69Fl$tEzeh4J=>k4@)z%v%!Pq
zE=Cm=_B2AjfxV2nWzNdQCAR-{eSWI1Fk8%b(kZq7WJN~C+|F#am|GYqEb?(GFUVr}
z`#P<nIq3nUVR2tQY~Sm%$dIadE%v({8!PLGSM>45+&!^x2HevnSquD5^}J1qiHXXr
z-;O?rQPtAQ2wf58V#|Jt2RR@97ov1xU;FFOH>oQju4lRgJjJWHma=ANdQ+}0n0hm?
zr&?a(b6k2%(NknKdTo`@y!SF%*YR{|60)HwRWmif{QxuTHX_K#Qlp<k#ip8}RlKs+
zbkEJLT#R~A3cx;Tpr59`8p%|aYpTv=X-t-FY-FHN#Y~_oP*KxnMlV+EF4QwF-sI!6
z94r$!dqUf7y#wza<+er{aaYm677K|yx_l-rJ$-i(cd&@Kw@|F}s6QXZ04VkOy-IKu
z+y=aS5|biN+Zm}SxX$<tsmz|D1!NpG`~&n6y4Ll~mm~ws%*<Rt^Fdcj0XWZcNZj7#
zFe%x0*ME;)P=iiefAggha#)y~0t5TuFxQ;}N{Yx7r!OQl(FbvgQUmsV=I9ka-Mz13
z9$e@GbFtmIyzGjdXAd6KPpf5uzNU^kDM3_l@!F_k5>gN?Pv^zp@itU+SoFcEGH=&e
z%^~Nhgx3d?*n>%MJ{N=?mlg&}L+TB12m6cgK?se@Oj_#O`mwp5JoyTuK75=TYo6>N
zzi?}=(Zb2DOWWbw=<;N&%fisX{E(EC6gs%BHAzyW8MfqhXJN%Sb_keSrr9?;mk{bB
zQw5&~o8<N9B6$bOsBsU1TDH(#7q~U6c;fvdcKfxd)@IS&jIy%5R~Ia{HVP}XwX|GU
z^zH2I0!H63i0mZ#aRTSvqb-9$0<pv5prz~Rg$P~uZJL8rvg)F}mB!cPjDx>^{c34B
z9a!5ehTF?_n>WR+aN&L((Co@P?8MCHHuB2i0&1si-<zJY18Ie-l-2SlIU%df6|mGe
z;+JyZ@Ze(OmP4<PU-!3Bt1Pz;fmeFq1u^I^q{V>DeZW$b{?y*@6W&vQIG{ykN`Tkg
z3N$+hMchs(GM$E9tX+Oor3zq5hf@Gk=G%M-UE%uk=$g21c~F@>tR?6ENucI|Yp2LD
z-(4%6`py2o$$9&d<Ycer+k%;iiC+uMEsVNNK>u@a^a1QyEw6tt^OV2kW(P_O6&tMR
zaH9P5`UD_P>N2KgUK^o;s5WaZSgD7rXleC#xG0dYpFXyg>UZVLaczfDC8>O@5No{0
z_|CO8Fxk4G?did~UnHzyXDLz_i<e6wheMDT@z?lz=7jp=u2ZJl`s07%YYH7VzIl4o
z%2_0xj(<#2Xr7~^bp=V`;#WtMj2v~XgSAoO3;4|QY)VxvqXw4<R~8$3I-z?0+D#n~
zH!)P*lQJJjzs|-Fe|O02(mm5m84Lu-#@MrFUPzcx26)O`!59*$lQ_LF9RJOkGrOJF
z^l6<h5uv#8spHkF_<G`?e~|0KlcDEP2sephBGi;#Usu0cR!}=Tl#7Lx>`A$dGy#)v
zl@_TcJYEbt{qEp;J&3LS3s`w*iAIUr<A=Et2L*1%G562)aINyn^7?TA?NPb-3_EBN
zkP|(V=YLhPgYU}pqoIoRn8Ocq?8Z2M=c?Z!p+Y}|njAGxljcGoh8AXg*|05@tQh>1
zw5hyQ0z-Dmy={J4niHu+TU79uogVxv+ULQEuN&&b8UiG#n|Z{9m;hKY9#Ris;Fdww
zrVC#;jGY5F${zis6Cs0uD~eelV*pkK72il0laE8$4R8f`{=CswrO9@e9H3NG3q9e{
z)uv|8<F?i(=S4KmW~bRTPp2dIM&YZgm6r}{&v$%uU1Vj^BUDti3m@Tn5^{Z)aNWac
zOu>0sI2+mCzj6`kQZ)g&OaL&fV}LNXT^8m75L%@tHUHVGu5_d@5O*NV8LthFC75}Y
zIr1z;M{4#m?x52~3vmwT@r_%nJ#Oaw3++39{i@^8P;odj=<aNYP`uY|5fNb(4e4C3
zp2((?i^CDJ6N5MpR#u=hQ1K@v5ow|I!{>d}v)`WLDI-BHQM-FJzDNPz;)4`SK{hWK
zD=f=P6cL1ZlDu&7!!q$)0g<?TZ8)^dRCrj-u~uU}cW6-Q5?8KesS+OV^Ki$T5Z#&H
zPq)ep@n)&G+bL8toG5`pQ6OcS_`<Qw{kx6>jC-|djai>8i&R2?0q_zegr97Is{h#K
zd7Ch|0a>uis(Spj5~tYwov`6u%iaGlm#Dhi_zE~vi*KdDw@NZ<FKrB0d}TQwZ<p4T
z$o>;&tj*DRbiO-B1<6bE`L1NsrUFIz7mBf_5b#8#k{@_nJVhxdr7fs|tiS@^bd2K_
z)9P})CI$!abdi(w&PX;gMRg|laW15o36<DZSmEiS?<^|Gl3GYX%G8B7tdcO_+Dm?M
zH&4-@rQkaKMn28z)hWcLgO1|IGa&izcpm}8@UTdHAC+il^XB^zisg*^8P8r3K<K}C
zx2aS2+LjT%C}kA*B0pdX0cA!U!x$xgmLqrqI)HHh&`C|P?CbjkQ0AuyqmFWVg{*B$
z{*Ozq(RMZ}fHLctsAm|4YpNo^I};9XSS3LO9S5FKPjrCG1b9VU2R|vcRR4fi#szZT
ztLr?>0R+Iq6%VdYo$(YA%6w|a8bghE%AC@LaD25Z&VV;+16{nQMB_Pg+Ot4Z{{DpA
zyoV=D>T>^8$bqVJ!B4o^2@B^TXY*tPUTLnZinWd~%B_XvAil){MyXCoz@ULM_b;of
zn3uzI=6~^XLG*vaEB_R7c-+Bb-*qoqdiknnJGTOrl4^UqhR_I$iMo0n|Ii<fI$Qo~
zpy~mwhMK>;mI&S<-k$kL94D80h)9$k9_{rLND#FneAr<iLHy-q;>!=AN-&)j5$hl!
z6<(>k1O?f=z-083_lCe+QFPfcI*kQum)buy1$mnX-AGArgqrK%V3?Ga%^`>!Eq1v|
z!`Qu1#=!G8XI_i#{f4qjT~~LpBui{)wG4{k*ECwszg!9wejDZqPq+3O#i+6$H|x2K
zK-wiyZ`Isoq4W~J*m%bi*Ba=)wp2}CJ8DHyJp6+6ihQi=N+qacw{9LV9{gtHW(1O#
ziRJ@<I<~j*5G26f6(Rx?tcbmkaJCUwn)ltpP3=JL3pH6e<hSFE5pi2;ZM=|Sxb4mt
z`qgE-uct#5&R_i-I`dg6P1&ISMHND^rkKdTe_vvNAm|=nBTPK8YpqBI#KGaVJsy!=
zzR}x8^C2TctIToPi$-UvWTDx8BaiGg#Jv`25Tmqet+t7vM<NaPHa@o3`2pSn<X9y2
zmfdWpPn|AqU6=eE_gG{2aj!erbu}zAn$dkr?^QJs3T6DKzIm{h{Dq?%J_xRvDko2X
zJw0O4Z9?VqJ$aIttfCR$4Vnu#^#AaNjN=l%=eEG<0|w-ugwR5o%OB4HA`K`G$~j0T
zu<G@lixei$o;}0n$cybOV?q_rodOpd?*$-e%;s)$Un!N3%8)DozP5QXq<bVZT1;qA
zgIZv7K(y?4UnsQ)Mm5HswEU>`MXLWOFiWs&+l)#HnCX-iH^-8SiV8aS@s}5qaoB^#
z@NK|6u44~eDT>F#HM!|d!Ly<e5T7*mVG{%!P#!exRsO_NOQ<yr%*s&)$H&7}o@hlC
zmGx%aUNh8n{M))B&6*Xi+wEyeN@?v?o`)DAsl`w56rxi|T#ZxA<N#Ka)nB%>mSZs!
z2~DpXK39QKR|L1W3d-(j+Z{37su<E36c)I+ii7p+IG2TDr4$(xV`BxWeaKR|xVW16
zuY;%MyFoq&kv0iI$=Fz*@GROjc1IGSj6+O3X@VQMDqTT5O)k0>kG)+A`TF+ikn?P2
z^=op!Iw&pjEe1+_C|L|esiC3G>xe36FGH#uuga0FtYw=9sS?0(jcd28D@5IPgg3gM
z;MU}EYrmwq_NV2sy->^o-T@esp!HZCTUL?&%2>S?7X1HMaMeOg=ZYYn6wAhDIH%wV
z2$XZl+*!w=mV~>wg~bC-Ke5v=!@DH*7KO3883-lo`Epdr=suz{1JA;N$Pc$U{}ua1
zF}=+s9_sA{oT9VbWwRu!@8E4)ifm(pH{7F6)4vwG-s^`*&lqlftw`$S5$GjC)lsQ;
zU)i5gtK3;|sO;<Qb-qQa1Tk6M;>5op#HhV&u&IVHHb&Q_FBurH@D}F3>5HAj?CwGt
z>Pw`2vqUH+Kd>WE!pck3Tz^x!9aR|#Qw7U@vcMiTdG;R%TZUViH$kg0m9w#4c#N-w
z2sIW9O#=po8T)dSaqk<Kd$2#ELma4sNL+T_+43cVJ&E4XO9TcD<r*nN>Z+s^;;pQv
zW^{=RS#oRb*$$c)QG7c}5URxk8rN($0TuJPFU-(<j~0B}0^-}^aAoCeVMQDF0Kj6Z
zeX22>Di_0w_M_%#z<<Z$H-9jd-`{XF0L}zB+JmR=bCuZMQuW<eWZUzW*xhv9{f5=D
z3<D6YObN2a_=*-XeR>jw58U!{0mX<I1ND==ZLw-|ro(i7rUSqT?9!K`ajq-%vstAJ
zWvf3;C-1D03onVkx&&?5zW-OO8pw5+Mf6G^0FYoF=oOe4nD?XeTrvo`UUIcBfHTsF
zuAd7X3mTF1wrWX)JFe=1eIvHst|WG_9GG_oIMMjo5B~~hqgseyyXxY0y3l9Hj#O@z
zFRG}ik<(s}k6E3BrTwe`UZ|w_-aZAz@>kU|=ae=)`;4-cNg;9H-IzzO#4Nf_21_#3
z-vjIRU{{i@1!fN~98nK;T$h>bN2GPA7QJ&+TaVZ~dwW^Yh5gpj;VZDJ1|hA_*){ki
z^Lc^r;qaA@8du(Eg_OkOXWfS?z)X@sZ30I(dQJda5yWwCzZK}8kAG7%>K0B2Q;o7!
zyf>t_*pc1^)~iAh$!DoUX!|vysCY57J3qj|;U^o1_`<P#87}_Kt<gZh!H^nUY^ODl
zuC`w43Ur(-5;z=c(f@H43wC%&9mRAhQm4iO;L@S(pAH8Ic!s$8IH!ro<CIs{=KBg`
z#<5E;84dW+V1B#*Dwsl{B4IJ)-}6q_jKU4OG|85DX2Uk^Uyydx*+0Ojt*oq6t&tf*
zghl^XM};zwOd+XAPG)8kfG=^M-Ez!Z|Aml1Ak8oY^%)8GKnH9;hfgZt=DnCeyF9xd
zQL~D<EMoSDn`FLGZrmZ%)vIRR(B(nW@^Vf{jv;fV(J}j*DM#hXIK`$49TMb<lwnOS
zS(s!?hSL__MOlsJ#6nHHvdPMVF`b~gIrmEMeTp}Kz$gM6ZL`vnAu1cD)1dfio92!B
z`E$AOn)dQxXH!*|?5sW4&>USA2W4qB3%Z*pj=V&VI5G5v%B>b&OH-DDvpiv>La$>d
zy^0D*oIeK8d*T3HpA>s1U}JdsjDXWm^%B5(1V0P+PCDsCEJ&VzWu?#LtCR|oBFc9@
zlyQp1S&ePoCI@{apg?kpoy1)$$&&V5q{AKmu`-k}S)b1Kbe>T(u54y@&|FElIphwG
zepzGTGbNV+UZvyD-FH^p51uNv>?>kB6@;za3}85Tn#xidKrhspEN^n1mYsuxilVOo
zv_A+8sH&)_PkRr151g%Fh2tVg4=GGcOyE3G$>><RukVSzf+soekaHKGKevl*0g_%N
zYC8EYMO{Z{u)jZA=1Sfwr3i>v!dm`QKp0N4OnwA#k1d(5*2K)L9gbrB7Zwm27Upvw
z9o%Af(==HbjvC~Xoiwg>K-@F~c)c*17mVg>Am7cFbKQRij2mj|89f_vDCyuO^an>I
z`4qnCU_%Gi^ktsW&--dv(Zs8<?relh6OWjE<SDjq5C*8<oEPkR{zFh7`!aiqnFHz8
zP+_Gw!xn)M;o=TcKP2Y92Qp%Rd(qg-B0HWd`Cm0`;ESgpdEx-IzlKJVF{Jjf<q7+A
zNqQm*Nrc{|h?qQcuDOb&byK2RHIPw^5#cqK=Z=zl|5Ia$YRN<@6s9gb9T#A=hDgyL
z@BN(q>2w#VG2%;+`bJz%=|7H&o$Z$0E1}1T<Pc?yC;p2YjV*lHUBB}Wg4*Hone2+b
z`7T7EiCRM#H4nO{yzNk+d_?FtK#gza%Ml8cmJ}Y6gUrIgT=sOE(WY3DQX4j6ln2AN
z4`OFBqtbgdBx+_=&PUy9<VbG;zlDksR2l~Cd>R1s5}Q$^TUF-}s*clH6$Cc_r0P)O
zJaP%(zKg&0<4g>uT_5}h1d2dPIuG}lYfGDhE{iojlz9x`zN_|ZEe)oL!=!xwOG|?i
zNZ=uH2j;?I$jzZM3D$#w^Z#nRP*umH??9zxedmAG*so~t{GIqIJYT#|%!oBEUn?uh
z0=oEAVQRZ0(FUAq+#)qQPlWO?{0H^+`x3CFu_j2Q5&wz5H|AflpP(pY>BOU2UgiJ3
z3FJkDe~x95EWXC}i0;TC#??^Dd)EQ%UwQ??_tXuL15cPHICSkQr8Q=MW*L+LuN_Yh
zb_^$abZN%H*@<{hdZ|^?IpgU6!7iif-0<ToiTF1_C<rh4kG~)gkgxyXe*;3OF}#R*
zigVib`8090<>D*7AxF)bUyXJ~#3&|9qyiOpP)Jjr*~FRenIL8Pf5cy?AbDhX@r`MJ
z(CcsyiGZx3hIjODohBTf#B1oarF)h^8}{AkDCF^n=Y>pv-F=0SrUa7SgA-pJ*!<YJ
zghnSQl%(8pTM$P!MV&NK?zN~W+bDgu2x#qQLdQP!TG%bcsC!)wK~R;Th+NfSD0<3m
zCgqw}dv)hhqAVIZA<JWmIM*Z_Od@|lwpOCXkD57U;K@k_bW;=lrtDmW;jT<gHE?5a
z<}gov>&;0<0QOJ-KqHa(v$waJGF@r9>iT>tl4@>qK*a@LgoK3?4#Y8o+V3g~FV6vm
zJ0)J8qcj_?@~%V)I<yFh3T~MZ_q%X2O(}H{n8e8#hgdz@o?}4Oth^<lOFp%j8jeN!
z^t5&7FueU-gtW@C`_pYqicBO>dT-6fzyUe^=*K-UsCUJ2+x=5nhB|?{M>4ot2DPeB
z-TvlKTry;bAdBCa%MIsH{!7#Y0KNnoNjenxxmlOY&CTIB9EO#REiV){+*oo|7JcW+
zxBsO&L&e7%thuc#TvgW6Vn;)QbsNbdJoYI23z&0VTPNq;TAV-z3xZ*kfi`45y`N3c
zD~-Abx%v4wrGvr5MUR6XZT+jF6&<2S6V`Dn{A+6wHLUqABPTh5xh=F95Oi7@%hW71
zWQIeybu?ed2Xl0^ui@g#g#!IP_Tx|oA~uFc4JaH89uisAljOBak)D(mKTAT>AK}tA
z%azy8J3EMOl;LTcNj9-4x`>WXNiNo@BWaA05P$eESVG+>ZgqFmu~-Xp_-B>gb{X!y
zz+~jAtI;DvQJdIRVW=aHG;_YMIxWX)>#GHu=|fOUVYgDM9EZkd<JLrp+CWAyCl}xp
zM1+Mk3N4zh%XrOKfBkxR;~0E-{f|Af=#lotJ@c1p4)cB0$J6T}Yi&QDKs4CgoV`|1
zZy@Nlvo_t9%4QOXLSP;Kx=}gS5&4zElOkpLcL!69T;1Sf6!k^ks@0HgSY&a^uU5X4
zl4Hk>wp?Bfy5(DLp#i8swcko0s`&J&1#Ji47O$?dYIsuFa3EYe)K<wWzWF||hzjoI
zr6^KBq^BQ$n5ohS>{~ixpPv3DgyMY)oWI`3d&!DE+GZ-_6B8>pE&Gce&S{)~N6n#e
zoop2j?{?|Xykmj74)yW7fg*$V+Z2!m-*~gI&bL0u<)aPj(%wH%>cGRpldPClSXkJd
zFCO-Yy(&s31&#%O|Mm^AefTlS*KlxfASVw5M#E{O#z!&ZDs;Ui=*E>0?puoQ-^;*Z
zC*z?t9^1uh_HeFe{Tm#n({WDAUF1~{&&<>R2=%2MRMy2wNl)zU3-Hglq9!RR$)!_P
z)GxTSxiBCEmIB578c<{$x)rX#%TiAx3iuh?C)YnDI^-4p^6|cWrl|$>q>HA86L3_9
z<>lpbaC)>`<<YB`FRNcd9Q~l4XQ<yrJg&}Eg#~N`s3Ha|)OWoD0|Ol$Dy(6HWllWI
z%*pZbOw7#8*RMZ^JlE6H^J^&##iOLgIDh6t%-bJA7z}&*GAXySEp;hG%nux7)t{er
zNX@av>)$!#s;KyDZ+q3J?#q|opow@7`C5<HIPdu;5Eg*xSlig7%Et{BS`I-`cVdnL
z=|349=SCMwGxT3V%dxf{J`&Y0=&OAeCEoV*IxX3emr}|4=X*zDP8O~>OH8q^nRaH}
zIPY&=EPU&hJ#=?G+)FJ@8R7llr1QA7lZC|soU%S_4_g~$_v#v_;(YJdm1*rOU5EW|
zJ0C@zMzgW`#SOcGUCS@}`}q9cG&C>>wfLNqqenY|94M8L0D*!roS_a23sX#+>&>72
zY|;jcqoOAn!KE7u!>J9RS7B9Y4CesUp~!w-qg+R~cR0pj5E@u*7Gp__j;5;^g0r@i
z4<T!3&TAPecX9X`Euf>L)7ja1>((v!#b)Big9i_sN}e{c(j53wJal(=fApw3{`Sl0
z=;(}$j1uQB(t_Z}@LB--`SWMbo~5DL%+P_;>Tqs*79<EaEeHF*e@~GNWLPX)m9w^9
z$?Ik0fBqPRG~i2AS;H(WEU>$ML#(bO&tA6hELGwTgeYWV1nodL)BLOrWDKw^aR9qX
z2GCJ-BM4#p$rL7C`W)GntAXD9$saEpXB20KbV~qM{}8BbASY+O`O$2bI+q{7I)Y{c
z3}bhBw03uWR;^$eCOn*7U7tI0t5hTNmXOdc06L~!S^Mi<y1)TG)4C@k^8-FwU0IPT
zybOlL%nZF;n9Z(qhcHf}YC!=Lb#QPH$RNTx@O6`xiVB8T7RII)Zi-zVk-%bcr4EbX
z7R9dHD}e1muqw2j{w3i@9cFIEdmbskbB`7SxB(q21hI_UuDzdFMCx&oL(O8%lNAe~
zVc{}|WCFDweHoYI+!f9;t!<usj)!8J9wm-}BEQ6OIWZ!li(G8)`NWS%*d;4Af6UsJ
zScgoS!-0KpgI7Y*pm|bi_HH=23oif|be`ovOUcP?tgqiZ?~hpniH|j1p+Wr_w6(46
z7!*H6Cnz1$y1Tops;bJ$gp0LAVZCL6nFn))Cs-HvKfeAWMu<;PP~-kM^)gsp@Fhu6
z=e4?qh8f_barO)S=O`&x$0KwLOuGWSE4C-1;GBG@d0b4){`U)(OSf-czH9*Q0%j#v
z3|o<z$x&+s${J2S-#swH-)*Jk<@*`xdGNl0m2Wh<WVF}ClUm;}idDqNSB*4(pi3|s
z76?dp&SSlWoX@0qU~SV*(f2O#4tgXbL;A(trT~s_i1*MDa^EXbO?T@Y%+fBc0Cx6H
zKwNAr@*0Bf?niJ)+P$&QvXUN5lZwg?WHM`OYXEMY4pvPc=P8Ei1hgwACT9CPKuwwe
zpn_M7jE*h?cotAJ)k0KA{paVu?;sQQ#^0Yq<fV|wuRr0+k;8|_#>YWqB?<O0PW>ZY
ze>{8zpgCdJtsw8jq$GIxzz<8ZW5<r2I<>K~@@sbe;X`EeBqLFYLBttI_)C{ANq5PT
zo;no^0uI1OLPA22(jH_^E7jC6k`36Sl)<T~lb(l;3fj%o6zUc||J`3~)0LtA<O|CR
z`<>>h)ggmD2n%6+qZd9~4(U8bP7o`kcfTfY7vd9puvlQvJetA;fl0XQyZ?n$u|8nl
zNf_du_AFN7DtHgAg1F|((0y6iaZ4~w^1b>o=tyXa!yudvH?*{r$AJGxzfJV}_iu0k
zn6+l?L0<pXEYYDOt}tqhEy#-?taN5rOaHE&X&kp^WQ@!06SAH91kfvJj3V|chC4Gg
zLArG*5mNlZ{*RNOn}bboh{A;TJ=7XLLtEL{{YJA}l&wIsOSK8XOD*1lH*%Q%^)0tK
z`Ocj~dqEfPdK<&Kgc)psL9eVbhPsnL<VLNZ-W`r*)xVH*?IJqw?&?TRNR$1BJ`EHb
z?Vxi?UDccJm`lDCv;Cm91l%5H#hGAb)_s_$2o6ovJcaMgC?g{y5OsV`-3;fvx3Pw?
zQpxC7gtS2x<6Q0zZ_XIE_ww?B>pqMLRl8u_TR!^}KVx>+5<tdw5`rh|>Yt!qz*4rt
zj?i_UPPX<x_v)IqX1VI|lS=@eCzO|$%cYVYJNBs5-n`iAl1Apn?qcQU=4RKlEqA0D
zxL9I*oYMF@Sg$ZY*kH_yF|ap#$ZdGWs7+=r5qQa0OH0oedO&ExytKu@z#wlKdcPsH
z_yV8B>4!iaunOP5f4}al7wO%*cTE*g*pFp4lRsI*W~R%*Z@YHNN=vg9ru&9zHbCN*
znUUdb40{>0LnFVErAJ3bYz4Kw{I?_0ArqC8qiSCLid}!cKBIzWWoLI?f+JZXv(LPN
zWobIoJS^~s&LPBqLb!LT@um<EfR|eQ*ajF1k&MjDNMX*MJLlu$6UZnwq3I4zl<EMA
z3Y56*uc)+!sbk-bFQb*xZ@q6#(2`(mh0o9K`HVN5^S|eQGrzaUI+A?JwIw|$J$+Ue
zG*h-7*;3=wV+#um^z{3il{gW2y~#!7$J6`liEruI6%bo^E&7WLn+|pdcyWJn^YF0g
zzK@NwtF--$%j4|ja%j54kyFLT2w6aDYin$5EM^U=Gk7HlL`HD&+StA2qj7^GA|jjf
zeQMjGub`t$)Fv9f)<X*{)8*POvUQk09Z(+5$t!=<x6Q8ak*%KaN?d3Rg7<2yt;x8n
zr1j?>A!l%%@688nS}s;N61O9s+D0$pyaujpd1XZq<Y6X_5EDPrwV2(y;Q#%X+PZr4
z{WRr|9nZdui@k>6RtBzScY7OLKZmHO=*^p9IO{~nVIkVhfQsyo<CI&BEYPiXzTBGh
z2HnJZvnyyN7d6=yTDdDUkiU4Oe3;vrxf!pT4)DLr;-JWZaM@G4nf7UT%>(4;&Z;k?
z_$<#*hy_*r6rQD0x_TDWKXd}tmnjB7#WYqQ0v-rp{St>oo#)S=D=I32FZ8KvYMO^6
z2==mHU?A2)mtN5JG3~3-qEyLqOSyS@(o#}~g`-$=9%UJ-+XE30!XjVLP6SD8d_g_u
zewwg#5PyL88~Xly&m;lrq2fcR_L37UmeO)^)Q*oJ!*tnPSne_tGoSvI3~Q39k<VI4
z?W|%8U52Fr!_ml8O;Zl<OK&xX?o6FiE7QbK>E1YKBn7X1^ypD{zf$C97gBRrKp~@_
zkaR@Dg!$9nc;UVEaBO;dnyyyGQD?w_8xEOVhu1@-DHoWuO&~h6=Ge?ncif5hdQ~+;
zkPJ&)gk^Wu)x??4(Z2QYkxSqd+fTaZ6y~luyJ0YYR8_B#mc@u+52lqG4D+@D1(CXc
z|3lAp%Y!<^xVvcBj9`jLNJzjI*5<+$zp0r)f&%_5<ch5G&avdS#R$2!iTQb1jJmpd
zPEL+|VRL<bPGMn$R5F0($19%#p@czRg#sH@WWXKL0~M^`%~$wh<7|EN_iKR+&Tj`x
zZM$~-FNaehYke;Z@`gQYQ}Dif>odXuqmZ|J4ZK4$@{(NiQ(oSs#GX7O5Or40WhF~T
ztQe&Zxvn-@2aYuJs9Ajd<M`h83?$&6wOmDPCS)ExqNQ+hbX0y9?e$@S4SB-`TF`lI
zs;&Ao1m3ShS7YvpGBFK90KS|E#c!;zQ?A*Qavd?Stgm11ywZc>_Nyhxn{oBI>=d%G
z!U7^9<z~B(-!O_eE4L_RzMpYx)!DIY1S8v8Kg7C6oN|VW^8!+L**#!zxI=;Utu&wN
zkrdkE__+E2w%vOs9}M^N;VQ^qbReR^%YL?oUy^HSX`P{@EHvw}-dfW0!RX(L@}e7p
z%p6h{cijFa6c<zT^CITGyq1=Wfb{|G1qm=ks|V2`snM(ZMia$I3D{TITwPuL<x2wO
zFjNfctYMr`Q-R~^>FI&IBwEx>5Gw~e4{`B(g6548T!mZxMjdY741Y!!V{z`8s(q>Q
zjPA_EP;eO-8bY+(gt(01HK+@MEw&0yULD>&1CNB)+q@z>#|;`Q$Z+#5hYldTLZE})
zf=-0&6LfUJmW$KV0#FC}U*NlP<;uo<A0IC-#OMm>N+qNymX-S{3RgkI8T9U*agSPj
z^h|k0TdMq|<<|E0GEs%CG}T4qsI0fo=-FCqh_0jcer$y$cA)CZi0WZf%b$DTg9ir(
zp)sgYfSR)H(S|f%^V4I`s?p{sKKTTNGTZ?o`vvl`xp83zjK$HBU(>LR3S04hl;6Fg
z5IRusHouzq&>^)<jhn!aW~zc18gZ533)at{cf)Ef{jMgV)=hX;t`CU|DGjGqg=?`~
zs%D;{Z1LT47z4bL>6%Ni3kZ$D&2U3n^y9}5NO_?I$+wNbpd&?G?2EgeLz}QCpaKC`
zqh0Q7H*KG+h$O)uK78mswM2`6do2d<ni_P;_xQ>yc~Ow;#s|Qhie9@m4LHeSYtrkm
zFsjPd&Q5z6BJBSB`!tXYh`MZ;85@7d%geiy@Z8Q0{MHmmB;$gD6f7<}Kq3yiqNJ$k
z>RDeHlz^b1s{0ran>z6P&J1;ksH*rMh!VK1S}{)D3OZU^<KU4v)d4l+-B&PHS}}K5
z2(@%{bQzhMkiUxR*LZ`)9zYgRP*5=5ySudo>dB{HUXVb=ZP}R->F+NKr5X#U0$_JT
zm6xUU1O|%59e@hIY=5H<l!T#;<1iah{8svR?tF#VY&tp}QLW4$HIz37<vt%gYip4M
zN~iP=-D$Q{)WK0vnwfEr?LeM;pr)1H0WXE>(2?|y0%8e+u5WIpDGhohm8rU#mwA?&
zdgajP_#@E5XAN`u42xfndU|^s47kw^jg6H26%Tn0^JbRZr{Tp{7_y`b{Eu+@l_hg=
zPt{#qc3-?8hPp6f|JIeJC_W?F?*`lu((VK7z?!URVs5V87Z-p2Mw)Viu6}0Q^a)Z@
zi8N0=TKAKmzW;{(QCwWCU1Ixt=;?6r`z<!r@{7PEr!fVA<?^Dggv^MC@-hbl@Mvqf
zHRG~LReA+tE(`uAC*Aly+;PP=fXKZc%d94~v@lc&y0YI>Q~VG}Jw+hR6VJbX@1?cw
zA8V@TLA#o`zRqJii-#mo3g(a;LMohgf=Coz9kkG2EGH*-49Pw0d~VPxKP>I;<MZ);
zq$*_g$Co@mq&P&0%RX`p#~D44le>~QH#b)@4X{1R+q?By$3r3JNC$84C+eBlpw{v{
z!naSV2vtG8tvA(@Fn+55Af~sk-^rwakD+YMqox8@59Rkn%2&nZD>HeXw|7Vb@1+;A
zS@cLbf<isZ>LnU9=A9TD3+>xl3qR2mEs$4&ys%FE<n?b7Ga|p=&b`LeTnl|4pOa!g
zPq^oBn!Sf*SJuE=W@)L#mD%2i0Yxi5MH8^7{r&y)KBVwfrU}74qNm_=iUv|$QRNKp
z&;EiKd~ZDzIy8?~yr|_NqtLr3oAa1{>j=?o73m*#!yLr$ce_WYXM}KrisT<WB-@yp
zv`pSFlrkI|fLu^?dKMneIK)NQ=}B{Jw(5|#^wed0EN*v*LlKFY$5`r{9!i=$@jc<=
zZEC1_hVEHGZ7ra09{27Td*EOv{PndO!vVGV<3_ya(cFA_9e>F1_VxGyEP_Hkdeef6
z3l4smxdCrTBX)sl^GMB!bWnD2;1=Ydh@_{3Do!BE5TCiD!Y}hym6nvKri*a$!RmkF
z-;jI7`iv#1i!!(;=jnszSB;)>Ab=F=vEH3S2(|7l{XX{N8=uE-?xDH!Ei}2G(WSNj
z)f5YF6(CY#rfL9TSX*0TFg0M@%xCIBvwibv!&y)KYN-EeH+X7x`3W~gaDBZxk2U+X
zvxM98VR&$u$8r3m=>A$8vvhdIqa6K%y=_Qf-e1>yQN;jj>A@mhmbl0K>W3kkTcGsW
zy@1~a);WIAZ;xJd+WaZVSpfoYb93`ChIIeP%Y*`C^u{L)A~>&--&>s?=x)8s^NbE~
zKa}{W>MT(#h&dy9P0mn^PfoI2xzc$%5C-TUHP>v@aH;Y%Xtxt=)S8xB>HM6`;llZ1
zLFc^1U&*A{Bsg(n*LDe7dBQAqA(ek!2;0{^PBO1abUDq>67cAFWj(!wZf0S<Cr>Wf
zA6Iagxmq<;V&~@#i7GgIH9lmc@k*^=;}rI+{wGlD(^^R*Ln(PkgIIEwR`FYJVPWB!
zc<8Rj_rGEif8&3_n&R4L&JP3lr&+HPFuE{UUJ7Zh7;Z1b+tqbR031U6y&s0-mPZZL
zI9E1<24M!@GbIqYw|{u3V%C!=;Ro5y@i|1CbB#Mn<@J-~E>22-HZ59CyTwI(%Q_B&
zp3%_$;q=7ZT*R#V6GS6^e*X5`NWtsxTmNS-<Mz<n!FL77;2}+ZFm-MV>TlsfSpZEg
zzoVytW_M_lZd1{J8I-w%etZ9O#lb5J5d?Zfp=tosEqvC|K)`GT-rUai?QI&Ycz~(N
z)u;?ULe7O>Ra9$xE4TF><NV&Gr6njT08Ow3y%VIFN%#Eq)jMAV=0DCuBEe8a`B!~+
zsLFHGr;a=6KyaJ7p^@plc&fg<6jPIvaqDqtPDm<Y+rDF`9QMCA3PdgZcgt~UxTU4#
z*E|H<XZgKdT@H|M-}Fm?qDxOt4^9bOM#Hfeq>_O+{r@<73#h8rwQYFWh2lb`q+FD=
zlpw7lN-Nz+w+hlAVBw;rI|S(x5RgVuQUs)=5m36j|9h^*e$RQ|_dDbN#<#~AXOFQb
zbFMj`xZ}F6`+gvkgHOwcNVDF6tuKqY`{GGg{BO~J%fLR7v?oSeb44;-`#RQp6G<ph
zVur*g7r-@>3)q_4*l?kpoOWAUq1Ynp)vH(Fpe&&VK}e_{AmY*!x`pPQuv_3)w)VE?
z1qB2K;j<uu2hkdmNPyX8TT*MfK6(TZ()1NS*3F@MeM>KJ9I|h}%Zgg>{ZEFE3DIHX
z-gWqro0w12;r1+z$h)Ua&{Xf|$|DHirlvIPA3uCZL*2U%X><T7=NHTCs?);&Ga!BI
zii&i~9rKcsXwijMeQUw;B6{j;kS{>6^M%@?$9N?k1v#n9A4O76^j<kkt+l}tb*6(Q
z1kj<(%@d3F=HYv*FOUxd_}7xU4)K_%>&7)Yy7x1KKxSNJW1DT47o|@Ge9{79`c45y
zS63nQD!`00X-klCJ2^R7Q#kznPWC&X(2~1CCq6tJ(1KJYW{XKDVAN<C8k!1wOVpP(
z&}eIOmgZlQDc5J?wgFv6nPOfY-@sx|Vw?f64rkY94pq5E189r9$fh9#q2eDQ7%-<4
zVq^2eLf1j<M}k$ahgcT!!mPqVL%&&TK>%qnkNkuIeM5uF{9xln;`8Y4??s0F7lRF?
z@wUZTe1dlY5+`w3e@adT0NX(zyy%w+`^`oCD<Kz$&rncgB-0K&v^F5VxS*g9LKi@*
z^7Hd4C@BZ}`(M0xAz-^PKX&_89PA1nDa)>BopDbjk^o1^j%6@R?CFxjt(OEhH2JWP
zw>P%FtE*_qct>p!qDhz{fLN0Q*qVBKqp(m<@k+ESzj*`Y4)_A>iF?UEjhmweAU%f2
z01$nUww*e8GHr3WvAJ0oLgDtl6KBbl>>mmFuxJZ<81*#fp7D_Sz}j?;K&P6Vg>ry4
zxIt<>$YToR3Cy7*>mO?Af*1|DE}lMCXB{TY|CI6NKt*WVBo=Ei_~oL2>|j=;QCS+{
zM)!U8@bJor``SDEV&$nPuLL(>A?eoGAuRE_y1V`DP0ev;TH2TKMOkTS09#5uEU+m8
zE(>B;Xp9*u>~t!O9`-*Zjeu31I&~^E{+0Z_do+o#5R|8fkH375Rbo~ynKdtr!rBVN
zF4Sohkx>srtJi&~$;pM#TOs}9r%#TsDLFYg!P-5AX@_iHsbJg=6h;W<GC2zghwDtP
z6<4IMd(BJo%;<QfQBT6tha3LblxVTrUplMPDO#r#cw^U7ug>`AH1x}vri&MmHtUSo
z*B&NR=l2qSv3~J6>)xCGqtM?m&yDbMLpTS+1v4qq^bo}`!eeK<0qBK-G4ck&v`R3+
z&s<%v15N^d0T>kFiYMd{L%0B>i%A3=QDu4<aD7N=yG(ih6fKXZvFLJCc*8<3fF_VG
zE-r$Cg5l=^iu3cMu3Pm&r7K_)m(b`u$3^|;Ex_QAUwHbHx27n*!ITD5+n|1ZmQun`
z3nWqzoRJQ{J_XA-GaK-nNB8tbmR3HTTPI)dn5;PdJ|SVaphZOR+1AHjuX)@caErO|
z^!nq$o#D?%bajUSa@^S12(dU%O)UzO0;`NAv=Lq&M*rNob4c=icQX9jW5}gR_Q2{q
z#r@+_N(#3^E|8p>+P&d)Lr`$}cNIl&!^(;^E7nNe?Y+n6hiR_$N0eUZiZ4v&S|)f)
zwOqC+)JST(YJcM_@Fk(2AGS`f>XZBgew=wPBXMLzUpiLvtya|+iMwwb#KmI@GOcF}
zD<YG6cXHX;y~Gn$G8f;l9(DiH4u;C@$`z$s48cqnPoLSl!pFNq&4mU~D_ey{h3u@u
z`O_}f{dae-g=%XH?sHyt_pWemxMok(o#=cdZ2Bbvq(NmD35U+t{p3>cJnIu}w5+U_
zizDA)S3`a=2TmE%j^5(Ox%aPfadG89%CfiPRQZQ*$h93N#GJDK#}9ujB%qB}?kf{$
zG!PJ48{h+G)Hf_nz<=iVHy}+pe%E5Kd=rGm@q=-YL}cy@7;(Y6m1PLV?k8Oqh?HRn
zyti<;D?%Le`-o8GJ)&FLDn8qKnePBIa`ga^@^3zMxw(|{uJ_b^f_<_d6I~B3lYWV8
z?nrYNV{HOFt3oG$Bgp)FW23hossHFgG3%>a?^?|ZPcHD%hDvRsI7Z&w_S7Rc^$34Y
zmw%V=!RsoQW1Bx?HGSvH>v%tAs^mgE5flVt+G%e0=n?x~zIpU9<3M#_a-r#IY)lMx
z0!9IV52(2MQw2I%$mSO(2L=cCkfK>2<}+Lj!6ajV4$&`}g&l`UJl$fax}HaD3F4-F
zu(B|aBD<Q=!NE1uui{-K>X$+KA|Ynad2>~IC3t19?Y1E1Ri-JE3fOF!Uj|hwF@s)M
zEK_|rT}*nDGpl}QhH}sTV*hd!O^lHUyG#@h6CK?c6gBgMA{O#_m=<3N1MrXi0|T&L
z_d(XEB+|dDzO^=cnUXU0?c0R*NvUxSj;mKc+5GX=V$a?2-eMN%deID~L|j}P{5zO>
zfQXxjpYCi}W0qD{F5>*yyWn^q)tIf=eFBq-e&ua@nq89y$c|015x&tfkA6Juy-31&
zqbIaDaL37CQOG?-zMzb;hv&;TsRU-$wP&L4UV`qSWo@P%0A5azUjY^E-^od!GIj<W
zhQOZ(7^%m~0lvcZ^7AzaHS*ivPzgB#Qt9I4wAfy*x(lOMd+qxNy&6C@7H5lx+(4<h
zvwBz~+a3%^)$SUkoy|~vo4Jj++`zziKSg-eDl5l##cR!aufCxVp~(5y8g+wwNM2FD
zaSCDFrQa%PCb*DdeR4roIh){Du*M^b;c};CVpg>$K(-?hoS<N8YU=dr77(m*kzC1^
z1u)T*t6P%6457z2GDKm-Bm$_i+e`sHS%+++@7J$1rHq*wOaD<QYJ{ayr@N(-EeA2~
z;xPZWhNPus_l0$W3Z0HEf;&K<qC7GR)PDY`uipSw6I7+cR*C{&em(kzKX-I9qGbhQ
z8);x$Py?W$G00p6?Pmaxx@B&j@$%(u&=S=7k|~C^Z!5cI9gEIm6*OuL0c?Yk{|OB*
zZy``rKt<Qo(9+R;1=L*uXCA5i6``x&-uR8ZxV_{<rQMCWeGo>U4e80eJU*=EbU7vM
zo(+4>()Ad>?9FfC@k8;#?7WXlNqxRw*Z`Kn^Y!acO?sbE{@@0nhX8B>HVQQC?B%pq
zQ;STTLmCWsR0%b*i&VNZE&5>u=!hiU+$zDSD(#e~h%2e7Vbv!rzJLCFGqY3QE1bNb
zurRd2Q2eXAB$Nq8BLGW%j!0yglIX*P9CyW-xmW+B|KD%v8vK*{FS2xtoxK_=Wnp3Q
zJyhF3Aw7gzIpe?b{)nw>>KbR`DrdGszb7r7c@Szu&i`v=YPwT63*e8=%O-|~pGr%K
z0BC+dA3mp&r+x$U8TAbflCrYdi<6p;>}S5;+6SPKRi`{cetvFl?$Dt_M(e%5em$W;
zm{OpLR2_98k~-_#@LuTrDqRIPB5sH_MrN};;Y#RiQ4#0*tg33E*lqCIjP&$u>iMs*
zLc+oUl!9SFK>#2smWCVktgP6&ySrmFV2c6;d<<dOTRpBZ1e@>cs4v+;1yWn(Wa_-)
z>%2g?KpI(A`OtaQ&)Y=wL8jl83mII;gn6vNO=T!&^83Re3WI=QI#>1DU`<;aCl^;f
zK)A*QPJbFhaU_oERV1+}ly;`|!rsHQ5uCkN-YSiRP+~GE&qZqUf^V3p1__<gkGW+w
zH=rtN;9nvuXl?QG_)W;PAx08kDX!Y}tCC2(WmDn`VH9BUWv-j~)pIZk<VPh^m1hSU
zj0dk^^wqP6?j^kmM-m5*jqj8Kk%G4Rwi?X!jOo4wg)C>6Q7tecL*-7v*z$@BmdAg8
z|Gp{C#KeSq$FzOYY>5hoH)A>sXlvHYv!q5dV>lq?e!!E{^o8f2R+szuoKb9hxOeOf
z@%MPqs?z3!A1BW)*p!q*fA*4g3y8O)vCKP2=>-pijAS0Oel*?o6(q$l{z@9`6OBt}
zcBms+i2l(61oZd!pDqs9{E*^}e-=#kIDn{NTztv6na%7Wz{X@4MhL?|(N<q1NvHin
ze}m8qtH>-Z;m=!a!GX&f<Uf7-q-1pV8}oG*MnS<ci90uWPge9}u3p7r0oJ>8`SNzw
zbwJ!8C1FHYtafSQzqXj}*ya>vbliJ~P$_PKivTF1P^|Y%J>rF3ih_eLpaI5jju#c>
z<}R3lR8?PJU&5aPWU|05fNi4rfaJJal0o4~gL#s;C|{*3iZfUyyAP3izyL7cI#M1Q
zs%&YQ1vb3R6#8tx2I{t;fV(OaTo15@rw9nXH#HfM8pC9+t*yN{1MGF01Kkb6&axt|
z*T=-a@)od>@%4=TQox7Wd3n{|#{gJ{fI5oX#9xb@F%kH+g}!|Nd9$)uF;kFaxw@7I
z1OzNp!V64YzCK;qnUgJJjYdaW%}UkEaP{55;@|DoEyldjLZH*j<||!5GXq4oEyNog
z9Ubr>K0ZEqd3i{`JUmd}MZvT95xJ}v0mL$CD{lB`PK`+Dc|j2x$OrH_F_4;qLgi6`
zadZAKLikK^udT5ucd*wcFk%#&x}t_p#YoZPCaKl*6y>dbAgfjLk|mNhfl0>gxcl<=
z+7*-n1>@eYcUc7@34cNn(4)ZSx*&R<!3s^M0wCG#<_C9y-6KM@9b(UDq)Su0duk$l
z#SP8p6Hi5AR}h%4u=#NCFZhx)0GA<>&3G~okx+_!{M_hha99{aIS2ern`Ex83RVAI
zh6`x20Rb}!zwxmHJf>FO)B_-oOwcR2x$S|vBVObZH5u9DpFf4xqY@Y!d;3A)QE)r%
zEB)?t`BXS5SF~3>05lLK1Z&vbB7?_!m_R<nC=&s2Jt;Z4)V>R|2sigrP&UDv{Ag^{
zS4R{+kZ`UqFBgb4zzrDASfPxKSsGB)k(HO49e#NO$Lqk*GP7QLHQ3f>4J01qpqbIp
zm))87pwY-QB=we-tVk+44wM|UaQ~Q}Pc#T=nunSUqd&R3*Pq9iomaZfUA{?lrrZx4
z3rZ>Qn}}06s4gHkrxbM)LdS|cD@{lc!vq2w#0PxYWM^i0vNMeR`9pQqZ}k1%>;(_H
zRHRXI`I1WXBe+09dY+mGUvx95wG?e_ZF|@td6SisD^h}=MPb0L=GAEROYX(}{-gTw
z<<KphTP-z-M7e|g1MZ_4m`B*7;I{(<1F7A2@|2kUe0-*%rUBwMjtZT}f&TEEsgRy5
zAqtlLWg&NYfibIHJd1%s$%eI}ay!k#S=C2oXP?dW-@rr(IV}MN0a>aDNFl(|!t&70
z?RE!YGN97X^XDV|X=FFCFa+D|>-2OYcsA`Sb;CI-W_wxqrB+*5V(^!0rMBsxA3m|S
z7Xjl92U*O_&TazvunMf6TE3x|60@w76gmHs;i3IKhtJ2rtaTg%Mcp@vMg<~7n#S;i
zXS<Om&;rKTBSqHKLtau+kJK2L14#6iS6A7JRoTPeynY=E1_j28q7GOOIh^>y&I5s;
z1(VYuhamE*?T%k`lf-ZGbO0nEV33dDUj$YIwD_W#2w!l4hzJChw2+11>`rg(#K#P*
z)l9Z61{2n~ae}OD^*nAvqKKl>IJ8TlduUvr40|{<iy5bt@I8;)HQD+f$(6^>t+|19
zSBSJ1|Ik*LgtN$~^QQ<GlqKhr!6o4Lq)SAQe&?K(<&WE#5ks!#dQJutK0LnVc-R2S
zsQ?j=c;RaI^?E~+5N0H#F(|?MZ;>1x6<yHnEIaN7VG@+%RoeWKvMcK;cPfxd&ghzS
zhH(<0WvW-5?_Hv+$js)s5{!fR5-QVmJmSHafrTC&9gQDEcATdksb~7abB16^i2n<F
zqdI^pydO>HbYqHARC<Z}0@}n#n@H$$M}8>&BSPL>S%H~-pGQcYhWSD!dkSq0!j{kP
zj0I2NUp)<V7~o(Y?U|P@Q}vYZ5lzUm<)GpFIg#1;Cnymd0u6uE)wM?eD3kneYv*Bd
z%i{#h^gU;8S+a^~PI9mMjtoCz+pBLPRS>)7mN`fVbzdky4<5EVCG(<qxxi5bY&1O$
z*}z76osW;tYrawKN>{)+8UXB;O3nx42H_0|#==<0_D<klkuCqdMMi|MZ^J$Dlr{FK
z2ctJ$<9t9u1E)>QFM2gGNDqL{16tMw-~&J;h5&klKon86iSBR8=Sf?{oF9gO_f(+v
zo}v?tw+sV@%YNcL8E;&06OKrTdTW^d$i7xPpvmi;yN>n7awAK(Vr5ffSHxw*2zRho
zz~P@gUK7D-00N4~AQn_pgCs@DQ}4@{F91R^*gy9>k9*Zi`Bgty+y(FbOL^;E@({*i
z&a&m%iqwdo)eR>?g5wx2n|>od`-Xw-T!s|GiB)}7RVz@bbeQrSIZp7a$VwB8oYy_(
z&>x^ofJFd1MZG{`=J4e!M?oQON;GCN8giFVSaM!xpwNxh%jn<}E*yKkU;inSJ34+a
zG_mv8z$f+JIa@QJcM<0SZfJ7f?TUzb$5jk$=JTmp5@(UgB^Pz-m;9DKwAtrLHQ1~7
zobIzA4eaXl0B-7txMu?{-YUg6)*onh(Z+yT02SiW5_7TkyZh$@8w`n((j;;iZl7q0
z6Se@p>(MCKvCl)lK=#s3$TR?`$Nf}(IqMOu&~F0wme#V|wX?c_j|0^2Q9G7P)XWwO
z0X?B#tu%d9W1ppCM7zS)PMcnCubN<2fOIuP7tB3_uwzQuQYF(G2;IQ{feUl)8iW>A
zhgM-xshz813fN`dUZ!xD-~Wjkun<8gUX)*p=S)8g@|S?%tWH^}s@;C$6T$-J=8Ai+
z=DXI=!35TXH&$e0_(-0G*{f?F=V1((3e@fR*QzFWLvP|h(8ARB?+rl~c>B${&PiK1
z@}mcojO_Mz{O+&*s0QI;iG3ryz=x~Bl1?;IQ9<{h1yX7EP=`wu9b!yd`OlgeFZXlK
z5}vJ{Eor$SJ#>rttevgnBZt}PKZn!6;3L86({X<|^99feZi8BaJQnkq43c3+*(fDx
zL;|gL+JKA6fcC%~uE~ntUXQ~W6F-%N3M=^EA1nUE3$I^C3LFu3ArEU{qY!!2*ZbvL
zXC(w<0AE3c@f+-mmoGujIyW+cv47yFQW3EO53W`qeR=azMRY}|9HLhCJWg6xNrOZq
ztsR`=XZk*FRr!{+m_Mw`pZ{o^)fB{vp`|qd!A4|cB%tToRc_B<Q+>-x3#{P<(y-O>
z8X)xY8^J3-^2zrfAVYUwUdF~YVldI^by*f&&@83)R73xgvg-Z&!_(8zk&!n*8vzL{
zO-1)1+(M=!^^+)M)CzKXgXK@pGW<%Kbk_uDn;c(U*{UrWkl-_sPfoN=3eofI>)R{y
zad{S}wx{j}Mtp2+Os~e%s{d1<?lz^l=8<iGNLCVV$yf(6V`7?}?up09<dBi++t^<~
z+<?cDgy1hf`wwMFeEnUMmxyYmlRpOWbQEM<zHUjA2&h@4Ov>rYA7+-r3*VUEL~Ppe
z0I<q;ErW#`L9(3Ae2U!S(E%dg(xJm{!4dAI<^i#PjncLFL1?M<w<TUT)EN9O_(u$e
zI^J8=Y_=o49Y7COB@UtIh+gGvG>B8p*Ht9<3@{@dDpAAtpZciaCiQ+UsMb|JRlI-y
z-n$tJ4|lUJ%-?sx*{egrV}H|Vi#YGR7XPZQ9)9#q6s0CEoE8)QGcxkUIwm<vIr*#_
zG0yZT&D5L#5<^1qmUv9Iw&cRmg4hbYA@3l@4foq$C)Sk$`^fw}vER=%OTS=za=jT(
z<|Gg!vI()88W7gqnKfK-#&8K1T|E_7t<GEC3FaD<^WXFjiGT`>mv^;9_2$ObmPW>1
zFhx=C`4mIj(7tV8nN_TKl5XD!mBKl0lzHs2!>;g;sp$kH$(cV%mnmQ(&j0*)(dau>
ztsw@ZShg0GU>$So3N9K(+1)htsDi1u{J8vG`y{IZNd3j=4ZT!DtHgtQWWes{L?v*>
z$gm&$wKy=TGX*UhD<S|ck#QSCG<A=TNcti~ISZnO;d@F{HxL_Kdv30`8tA`BE%=6n
zQ|PU>acGz%evy>)6X?05J0OI)f2c=yJhY1XZ+jV#-uAyv>Z>34N5vP1h3c%K%1@jJ
zHtCs8<d5V2DT%lpKG$39c|pPjJe^oO?Ny}J+VrY3B6$6HAJQ|_{_X#+;yEA(QRiPL
zaFc>W7~ujy>&=0C?gBMMD*u{=nREC0Y%b-}b%fwV@u-8^{@%Tl=xm*eQh?!a(IXLQ
zBo30`i1{-=sY{*!qJ(~L_*>Lh`W5mo%238L)58s*It>jCCubXm7kCjxlk1u1=EJo&
zeGo~(OV0_NWi%owK%oLp#lA)&d*H;7dJvF%1MLA(y*p^oz>9LXkdBZN6El23W_nD9
zq4W7wGbs#6#wG?AI@jk|aG~RVO>M2j3{e6TVCcI$J5rCKYyuQH;Eqd!)(~yM{ZqB3
z_vd`x4ocxEL^q%i@#KAi&Uy0TP01N}lUD`=$2%*vEo)_w0Pfsny@eStTgW@EzHUY7
zs;QX<f?hts%iyDiU5A|HUrNk~n#we_q3t~D!JlDUkzGbZp{S76dx704;*NNXQVJQ`
z(q{JR!C&~as!wF5efaPMNF~^++0g(NBplN7_5pRYnn(pp2<%BDHUYA|^MOYIgeLy}
zdw_vWrNCsEUX9qPgY1~T2g$ZGq$K?K3J<bv)N0@{&sO+sP_N8E^&ZIUTFl?Z#K?%B
zetpXk9@uJnkLphb?U%?GuK2Bg>Mt=)f(^eP2gBBETu@cD5B&@?e)7}Ni6ap)-10Bg
zc2rLo?EKkiyf{E?7Z8iOuU{7?p`ydCHQ^=@e8Ae`L}-nFY?9g{-oAYTs(K5HX>gg4
zw5ul=!uxOh%G%Bl{_QJqTl}waQPf*6V?mdLb>0p98<;Ahz&CH|HkZevc`YQwZ{EBK
zXcXv3A6ITdd`y94MIe6`xpGAhxaTh{$59@h{>|k{y=B2|W?uC(!)innjXgX#a|<|t
z9KbZ`>FMG;ps^CExDc3|%Yg<CY|#PWxjg+lpr=&HQnLp6EFvioyIqIaok#9LgOr+K
zz6aUs{YwBQ0X6JZKQK^X>k}SMf=+(_KIG8@=1>@DcG2!wB==0X8Jv=ufhp#87Dh*E
z6thur2NvYxs#f3y!K45u!(9BWx!Dg3<jw<o1K`-akiskk^6>DKj*80i9c(7`6F0Q6
z={hP;6RLr?UgQK&kpzoG98OA8DNP1WUMhJBcw6U+O53EDFOQ%Hs@!#P6D8I_l$)l%
zb`28f$QW(jq*`k=O$6^q5CA9uGFw=?8)H}^FMw=yaz$;%?wJ(;8I69>6s9&n0<?XQ
zc4>kTsjM+KnoQ8yRKjocKo5a3eF)P69osHE0DuFKB+#$dJa&aHqgbSJi;!A`E)jTW
z;L2VgNFFXvIqUzwgbysXXOXw@U;rQo+H5WW6zz4DiOCGaRy(y0gHCUZdLUy4^wEC0
ztGTl?RP@MxfcwsE+^2o02Zg1(==W3ygK(0MH4^8W@KBD}OOTx4)aRg`gT|#_t`Srv
z!$$!#NRRQ8cskq5o2?=(4j~(8xz7&*LcwzpQA*%|Z;&)DuRRG;8-4~J8<k*|Uri$s
zDO2?1H^iwVQ^0ilL4+{Vp4tM>270y+q?O?5&r*wMArNQlwI?wn8;_UHm~6Uig&#*u
zSuqCrKF=RPs#loz97Ob}k$+sZj-X7P|A%q{#d^5?B;bH|BAYX6@_t%`!zo1)e5TmS
zmL~mfh#djcq(N$f|3t`S9C1lzf)kJkl3WbZD%X7uuHRGQsD|pYz$(s1{rsU)%J$Ni
zV2@>M%tCJQ?xF8-W!s}J+<pk4xwUn<^dJZIs7-P{1pPH9cg~86ii%W+KT5}dKlYID
z;+#-^P}YOjTut-+9R7_*hERT|JzL>D?)$1_4wl`vnP>m+@FPUIva5$(<>bz}gJ|=4
z)dM9Z!{6Wp5^m~&wh9e8DOMy+m3@e9D13>J0C;W@)LH9yqj*!>C(TzkK=R8F4DuaA
zE^@!ZURB!KD$nCDZ_eBzXAUz;R$xMt5gae*VMRDUEp|XT!uD^oGrw=1p8S0TgGu&g
zkV#r&C!R}mdka~B*DLIbv+$sR1>~nCTK=MjS2pNfZuo-~T)+G0{l$5v!rfCD=+XVR
zk~4&F{)FWoHx;<{s#4E@)Kf1)L)dNd>|QWS2Z4^*6)AVZMe@_1_sc#kSZr(rZjMtk
zafD;<C5OpG-IV(gCgk=^VQu~Vp-SV8?(_SEY=?98{1&U;I3BXd5^?4KDynDN2qCoS
z-iE>RU8q7hC3H9hk~Fo<{V&j^6DnxXldH~J><<ip$0hQj6BX4(zP1pQ)R|ss@82IC
zmodjv0MMC8ZOS94Tnn=Kln)+Q{=!fO!xs&;VdV4kCGZ^4sFZ{R6ysY9PaF>b=y{BA
zP9Cr-+yn;KqZi{845X5OLNgZlqbnbb0eFawrG)YH1WoV*^M|A<5aJ}-o~0?BI<^5F
z5Van<BR<m4gN{+Nf-#1PnGTZ^X@$gY75~hfP3M^fe}Df=kAN7vx^7jfk<IQBSJ%{(
zd6zKjI03BL59;e6sKZ5dMfQkc*A&DXcGH~<hu>26cfgg`>7v>}+6m<&OgC<*C$&Mj
z8K^rJfUa6zT)dJ9>g<Gs1f(pbF&H*e=@;X{2eX4)bQOa9VZO01aA@LvzE?kl0%Kx?
zAeYQmK~&|Cp2@|F0GV?;U5V%c1Q8(awSg3iqiRPXwOtk<8>rX{O9LG6>*v9e<lcf<
zkRjA}<l8lTYKZakJ6M^BnfJcL@=;@!CAWMD7&H+Jffb}yYhm~^x!+$%y@(OG274#U
z?RKHb34_)Pf$4UAfK$jNAJ};Y2nNBjo<whMZYo#$euc6nS#hYGyM(*%R+y_cD3}mz
z6m0w}89Ka=FQvDoMAS~_A3$Eo{J0B_S;aVSEWE=xl{r8~A5eK;DTu`SKB6pqekDd}
zq7SiGD5Y>HNNA4YrA-9ypgcH(M3ZB~9(Oz;k0kB+cC?eDP5!6LUBcOuxSv5;10OIr
zI4J%RU=4tiO$sHYrTajJV`hNF6l3Efp?3+<|E5#J4K>X<kaz)XCJuv13zW<rkIM~U
z0hZU-Tbi0^5w-bgMDsP}iv*)%ZH$w<uLnVN^)(-lI02d`95^yRzYpC6km3PQ1OrV5
zwB-U6TJ^uXeZo@@ns9=?PEo*GTTHAP2=8MgP(W80FnW<Ops{8D^APun^Mh&bkhnYq
zWdcCp>>1xzpksErdGj+Q;@^#z!7;_hna3IRKy5>=c9{xdobOLV!at+B886Sk*Y<Cf
zyKpEn^(}dEg>+jm<dBn|JAK+5s@o>*J@tBvtOUU^gY&S0g79>u)AIlvn1=)IS-xL2
z`K($DKUvOS^q->bz^`y|1z|x<tEi}W=`<2WoColYoCp%}C-JPC0!V0j{A}C0lln{)
z^4i0vn&cTQ4Gh{dRXISz@J--0<Y6Zuw_cwQ9lVEu;|yk_-8@O0y>Ki5Y9}1xxMcvQ
zpk`nRy6G01w!JV?kCc%R2jZ$Bh#oS5uqlvhnIAFWNlUZ`{DsKN*c63})y19}M56X#
zn4O)R*tAPmAPxitnsGto&=2c=K&S!31iV%s!F8)gNMltlp?@F&G4NCFZ}T(>@f+4-
ziS_(&)y9Yqh_5U_7HR|a#wSl+E$gnW^$b{qs3!+uGOF`|mZB_)Z7#npPSIq3Sb|Kb
z9G}Lam}zkI-l9o2q-y`a<R0<Ykc*l2R)PIZAUk7d_oc8n;J08ZIbj6?pHEQ-f@ms0
zgQ3#Xz4SS70fvT#uoRn(mmwO$1s!2ufQE8-Y*0W=gicio^F7B~Vt)X>Ix@UF1TvmD
zEOb(@Dk&`$$@J4=kH7Jh0S$<sIuc_%L<hZaUG<j@iUDbFx)Rhp{ryMUi(;m=O)f@C
za{`Nt9sV?fkP>^9p8gJU8%+!y$6ZSF4@Bx^Vw8RA9Pf>yi+I4-@!UI;;XgJv5Jjl!
z;0$o!0MGa$yIPD!^2hDfoNU;i+<ifL!8mdS*g`009~g+m0*t5PQ~-Lc^AG52X27AJ
z1H+naZ)N1|gRQr;P~|cX>KVkV0#ON45St|=&|nZPsHNphL2rzZlMRfE_)3@NW1Jw<
zoP9I}(*2(y=L1Vh3>rQ$A!Zpt#(`glm>(37vtwi8NSwVq-qP~@J8N0^=%<K?2&fc=
z1FjhvQ-M=FiR^N7Pic802TNmH0g&rjYIE72HnF0`r!@D^{y%VP9_GO*$)tSQv=#nX
zVCT#M24FsbjKx;LT8VdnT?Z@zYStDm^bV9feKWKN=U*Ed8KECQ7z%haNTpt_!#?Pw
z$I(u=C61sYd&^93G2n{%Q9R<e!UO-W*)*EdldDaQxAP3IUqE*^D2`Hpm1J9vhYl5B
z-y-NmYe6ax=C<j@2?)@x7?NB7;1ld6y1gWm$&@v<8=n)QI^M$81q4HwPLw27Xz5J+
z;8~r4_ZL#oKal<cC?3Pk?Q0|#F3^6cf?BdOLNE<N;s2G%2MqRJsp%kcH9(5(TR$js
zRG&D0{Paw~DF1WFd_5W=<ot?u`a-rB_paMVBmUhhuY1Vrk&6(d9z~goBS!DTPRsI+
z@UHX`+g!b4fGI+Bm*{y&CtYc4Ys(*WnhgyN`NtsBIg4cP>rFX2BlD9d2!5`*%i^>5
z>K6$9Cl^Bs-0a4pLAJ`}3l}b2zT5|e+~AmDrIC<?YcQixsJIqf2ZQAQ*}(unLxt79
zJiSn<4IS+uI}kFc^@X(l2?Yr?HPp1sgG3x;bw0Y!R+XWPAloI-@V0z`K=)TYluIM=
z>J|J*iFC$=s4)a@Zf(KY4SP^>590m(I?=s0s3b$6;|(Ax!I**OjEtB#EFwbfOy^;g
zn3M141n*uZSy={<3G;wVc&_5WB(diH%{&b%5GX4K%^DCIZ$t^_|NfN_2LaL+w4pBr
z0Ru!PAod)Ce}`p4n})p;N$nY!qCBMEftt!Fwc$+HZG2jmprZCr{h_BGQVFjA%^SiN
zNESd9@vFT(7z;86C=zyFo0*)O+k=Mj40LoS5qBcRso`hdV!s9rAxGXL<c-8X;$)9r
z{pToM(2Q2V%S45Q*aAy~G+=;ckTi+lt#MWHB)Z6H@Rne8WQh@Qq2zT%yd>a@DW}K{
zZ2til?)$J(3`>KD3Z)ix2j&=xW~AV}w#NGUv}7<nQ5T{3zQW)_^zvIMzR#y~e9_P&
zd+<s2olht?GCn~a6`UWym~HdwA1y$funUA&R@b8~a%mxfyQA<nx3}<4(0&9GlN#Hm
zl^C%cxD5@HO;mhjWW)Fxz+gZ79~l^oj*s&J*6-tR0vMosOEyRW_!oQuxWt<%&5sAh
zGVKiUnFXMkdVnth!w!nxI_sTUP@^6@deoPkx3Rk$S}$m1Z~)2U^=v~7oR+B+77M}&
zu(<z_=co?!yF+_mX93xUe;_VoYM!;Xw+9jjNM0%~L+q`0=e6hnxq{+j0h<$MQeaMp
z%IWw+ER0zW?3jZ&=`gS=-^SOVL@Df%?UN^@;BzMSP3C`$j;76(5ADk-b981uyQxEO
zdGz}6+Y|;5kH2{SaOCJoQbR+6>oT>$<CP4-zyC7iN;8x%(M66mJPV2@Ia1SHp&s{E
z=EWh+hX(rJ^{@Gy;5O#IUi6t?C-0B4h4fU8PR7*U1Bt6atkT(G8D*<Z`<=?qGPYB@
z9;8=lE06cPZG-S?$Iac7`6vqOw(-z(^|Tjs*vU!r?c%UI?Ng2u!Y7%V9&!W22n`uP
zV^^vQ6;kYKA09$J3oY1VTliiijU1}}HP%<vtyZ9rtboz?i#}iW%<IBwCtTDa(bu|U
zZ#8QYOrWZ)JOB+TVMUj(^;+0zx7g}I^*(xNaG;i3o{5DX)AfEzZ9t!~{w<R68a#+%
z{QmOuFf?^V2w#ZZk%LpFhxvAp=9?>nXjCoDeok!dQ;V<p;$JgF1nv9u8P6}20vwKV
zq-{n~2h@Q$KBfoSJqd4${?gCVMb3ChfBJ*>ak|4NS*RR%NN~9~Zy5RvNTi)ZiOKJ+
z4{=^EA5K{xTRHRn^5aX7RKSXo_IVNyLyP@gbjA@>E89X>%5d%3xzRCSpYN9+p=Fng
zVBS%V{+XPIXv&UWub^^Wwp+JedKw<lJwzkrqzSqgRQ@3#qVhdQkLtQGEzG5(C4ve#
zxU2Z%agi_ya_!|PX>R$%&{9!Z$fupgjsEHKY$PgGe+`JAdUerlj-1umUw_%@E1$^j
zJRWz}1T%@sN9e6`WA)mC%x7hyL~xLh+iW;dxYnE<!gtpTKmxuHl+QZaBEPxj%8E%e
z`y2j9O6TjVShzZI0`3(`3TkML3f)4^Qi~$qUTTKJN7rzpXUGI_yW{OhW5?AC8lzXx
zXzhRjrSK^G%I-UW&(gS{hYwXBVJ1E!yc%IYgT?ZTxLe+=;4SQ$PK0N_?Sy>qTCb9-
zNt#~CpzSSj3U?a&K1`-@2LA3oxci|}`aPFVzaL~n=R*qGi`kl%@-Z`Z2pU?C@Ft|f
z8ZMfqf8sx`T}NAgIhMCYzotHlUyCIO)sQ4^@#r+4P=<gYR7+ijrJBc*iNAc~C35+u
zs5NWH2d!_3z*L7wOo#U-8|ffD06esS&h<d2QVS9}PjO!I@xp5BptE(SE?#oMtyCL?
z4weW*_Ce<b=D_plvZXB3U4_(yR{WQ_AYb;vb$gxEf?SCimg}ced41vh@806Sj1l>=
z_Zw?-*W;WN)Wfi9!5Id24g#MO@RwDQ%i<fNy?R2$5|*0cSyN1TrGgah)%b_kz{B4<
z?GYA{PIr5aV((4GEPp>%EGV$83By8g=0phUt=|7!q1x-X`5F^~`7%@(&X)b`ce%f=
z;S4)hm5q;SR53j=U7|p=&=?$!);UvI()KJShPN(e{nCTRz93|>_>f;c<$uA|?%8lf
zaewkxBh|0DapP8Zkh$PS=Hhza^{(9I$L>SgQd(|Cu5)&S2n*@4a~p>8ciwF48H;Cv
z#a`H}m#Um~Mt)YUmLkCW*cEsqf|o|G<Fse{cA?TwVUgp9{P)W5(@5I#HXiv^hEOST
zdA^NvGnde9JWuyz8zKwvlj!6FlQ?<5!Ej>o(iU?3Wa*<pM+=*+WB5sQm4IKqqL@Oa
zv-_&c@t!z^HeZ!SxWK$?E>N#1u{hf5Bh{^Grm7PSe|o(}F(Vs0d79JNOojNnckUv0
z-`9@)%@}UMD38Tn;L*49$(HQIU*<&C*URMIzAtB+3X7mv#`pfmIXRaRI^4sn)qp$m
zxY?!=U30cc`MtJy%#)j0Zt2fdLUB|4kN`-g>PWL@+25Xbl}&*2`$pfkMQ^YE8h_am
zxlA@u7vLGEm7~vz&hwGh_n9k_Zo+@J!8urd#}oOePG>vc`X-CcYhp=Sm5xc&Klp8N
z`0e(egp>r?ZnxkZEPYj#JJaF($6g=2xi`G|?(%G(flzw_AY6I1ve-el@K;~(m-CQ8
z-Cda7{I#*H*a!wQO6x+IRo>B;2Y0ibI9zTP1(nYr6?_V(i)3fVLNo)AC79{qqDU)y
zC{&04U<qYRP~`xH)+s4)Y|W!uUozo7I4)!<ua0WE_^&Q>+zn>nEZ`ZY-@n`EHy(%X
zBSynf@-s#Mwa~A18M$~oHaFjbGigAQ?S=+1Vh=<UNUs8@>VadSa?{eT()Jk_^#D`0
z2nRu_h0$Z6bu;vRfW9Lx;DR9PvxR@z^MPtmO+thkMK(UP7Pt5lIh#2sBi48@zqPI4
znJQR#k35>b04b!oEcDRPb^QGrX?Y5L(*h`NEVak)*}?+@pm}(4<!ZLCW%Os`+1Xh_
z;gz>W>Z;myoe*h4l`LBmA{RwEf<rqWfE1QTd^qeiOAqZmKS7lS2lpu3ziqNTG`~??
zZdJaN7(Ey~u4_8@LF~9g1+X=!Yd?8%&ydzbZB`NVVNp|agpze|px*V6J|j(5CrbCG
zGIQe~x7t9ySem0E<Q^TgUs;51yk+I(TH?0ay3jX;t+;=L{|8h{^l8Uk9?s0n^b-kD
zh}@$&L?8vbzSfgOF4m=LSbw=zQxZsOR6UhC<uDZf%c7zDMy|!8qYl^)VEF2+p}s<B
z$BMhp5=0Fcn`Ykbl~T?<Z<d@1Hx<w&iAzgMo0+BSk>1`CL;1#3U(P+o_vi^Vqe-*Z
zz!M)+^@=yJ^{dr_JKv;&yy;RPRFAJ~%;#9gQ~re|5+;=lLyO2RGb4xQNzm)Hj%YX=
z>C-4bv<aq8a6={4S-;3FC*8^Qn^PEBVria!d_Lzjm?*KZhAEd}m=g?Uj(%FW#De_t
z<u_GU#?jB9;&i?^qIl3*THM^iV&KmD)+PWuKmtN%a|?KAZdO)|fStLv_5ef)?ei=j
z3kxf@=gOXS!{-T}C;j|6uZ)Af)h;%Axjn@!D#gsz!)$u)&76ZCnQ>p`c$cZ>(9`fN
zmcNZ8<KusFrI@}^FFi4h*9k<QJW5fwSQJ6=$Ia8NDn9z#&%bWAQcQbc?SGdy#SO&?
z<CiqoMUDKeQ1XU$9#FML6j+(mOhQcTeUFEnau3F$D;F3QfTpCxA62ZXIQU&2Mvn7w
zS~iYvQPt8yI-fxjQw0J&-R<7)N6;Usk^`wvaNi7fpDaV#5SUQ-iPG-4?PCp7^9*g>
z6gQhG88N}RrhVy*)CK)3IU)14$3AD~G>=W2M8eTqpSu=HT-@HXU4)?P){k}`zkQ~f
zuP69jsi-KH*qCecv(7C*%fPK|dFpd-?$A*M5W4q$zdlV*6#yj{Hjoq)Sq;H~Iwab*
zfn|ed7XV&H8g)WFj;n~Ss%kf&mvHJ<B#7eHk+%47ik4t?z;POGR+h~T>KRIEEcWW3
zZSz>SKU%ps06>l<^N1_3YoKOhC7v!E6dn~`P)&W=f4!jD8QKiC(*_8^Vf2QiwXnrN
zSunKQ0Yk%HG8PmS6*XtEbmsJF;6|Xaa;f!bt*56a^vJ2|&V(gA2Z#MkZnK9Ot!6X7
zEamWrGfn1-N$=gJIe%a3>`v%~<HO6F{qq|ayu{blh-yM*Oj)Qyh2g*xp5{Gs?UCiD
zhIIQ;LMBKR8QiHu5|jz2Q@I@bWEK^(BN+e(fBQNLJ?bm!E*xWC*LpF^pJL{$EuHhz
zPWb#u--+LfZqp+CA6Z!f?>Q<T?_PTNy>SYqtD`drB7Yz#Eh@@Vva&3ZT?}exuN~+-
z4qpmIuD{RdZj@DRmt|yR#GOsS^>pc|<ZX$^Wl<pJn%?dJTd=<wzh5%H_4lvw`qCbH
zWPQ8r$PTvhf!ovh=i^6YH2Ep)kGMj^@A`Eb=E#%TI!y61B<MD~<&?glcH?`M`!8O6
ze#jL`u@Od1P(8j}a!7>SFG#dKX}a|Cu&~fXi`|@UL6Oz6D7}@Z-qFL)O?evmtS1*Y
zUE`=Jn3GQ~kQU?%RELL!dEbNmpSue79i#Bz0ViT7WW?nF<%Yh+#Q;;Jr2LH@+WEUV
z9xuAk@~a)G`E;<Rcyq|j1Dz<tT|Q+!Td#+<wjC&U(wpj^#5opJjJk?Et*W9TEe15Y
zvxRP;`@;O3mz;U`D7wcOj+cY87xbP!U55Fm0xRv-?!rTk)UiU5%9h)FUq{DSCCMZu
z=@sVQOB#!HijxEL0Ru%5NaXCH71()bFTaWXLLkybYWz+CH1mY<d@eQ&7L^Lwglo`u
zi|_jNACcVSu$qMjKE%AcpZ4zEZ@#L{W>{d(!mIF?-t{Y!Ur#+0Ps{nS(0F$>7}B3p
z<at}kT_JmnkHFla^6#sh*g))PyhsHonjr7CznkM-H!8Z_(*nm+LBbE~wNx_)`M%yl
z1U8t0-!5|7bw;a0-;HPFn8c9M#oEhv#@@Z5vXLM5!J0({2(SHu6d&p+i8*wH662Y#
zT~A*tS{QY2jgo_o1F!Uphm2$oLgMQ4X{^-M$SH|P$5^Pv<EO|;Pw#>HT1;RBzx4=*
z+oDgLK^2GAr)z=t`k%H4!*Nx&2d*nNLq%3*=H-WmNzm4N2RUm{a}WCXJA-8Q&U<jN
z7f;6uIsLSb0bl+YI_-sP*)$A`#>1hN=-snhyL-U<pgCm==1JX1-zO$wWZQfD`YPDz
zRA1NVIW*P+U$b&NuU#Xm=4?QQKjZ+vU;Ii;Rx#=%D{kb5Qw>tgxUky2!0Yx)XKH9@
zkbY$sF3kxK4ZVXwAJ3E4ymRNZLSsIw!$qq1pRf6X#4u3gMWca8d<!K7GaO8BYwM`+
zQo(%q^9K*1u+Epiyy~F`60TD@RQtGtd>zhAgwyZbzFku7x2~g{Yjc$-YE~+MCPe3!
z-pX~aBByQv&gr_=W3oQe1Loj|;Cwk8nu!#h$Tp$4?pX+sx!IIV&74Q4exlQ;>R@NV
zN74`L^LGh^j@OcH#8i`Q=#l>*bKKo%YYlEvW@W(GOPX1>4cJRgVwECS`HzUo#5|`K
z`B6<sZ8H2OfP10lE16@pW{=x(D%8&uTo9-AZ4{#)t_m55fbg6TI69QgnL|`!ZV<zr
z%+`1pbPAo9QCl|!83gx`GlJq7vK<!k5E|=SCoyOaZj|!E^1A8U{*J||tHt~wu&w9Y
z3lQb$5@v&tsB0`rZ6KEDq$4hOxuyldy3@Dy-}a-MCSQy0&_O<?^`U+-F9{wjCNlhq
zSZr_M@w0BnO{ii<OLod}KUSZZnE3<0`j(OZ<FU7lhaXpw^?&*-JKwO(!Sh&+rgRRM
zk%`qp{xJ5kS~L2^Wxh`V!6Yr$t8A05{;`sTTg8kasTt$1J5{-7GaS2b6CM;_)lojz
zgd;V^sNrk0{_=lT)giES>ici`1IicFq#iLweYHL<TMm|X;Y6xBlvN|<k69pwM@c{=
zEcPg$)nK3<z85({*R6Tg>xTO%F1M{diGh(2Zp`;B%yf0{p@;po`E9eICmfGAT+4^-
z!p2vM{<b{d#ECU;tf_|Oo~k0Z=s50ss<Fs8(lbBuR1tp1M@57;qYCZD2jfVN-0jI_
z@H>^N-Ji!Bn>78m`HbSY+`$3~K}O?vYC+BV#C*phDBU)8^N#ssF^wPG?u^{-(T<kf
z%iUjBa^~}$u=1Hb$SzR6%<$1!-Ca>yd)W4V;3TeI)N~9Nc1!FA850VX<cA`p$uwh=
z-3ku45onJN{F3-<)<Hc~8E;A*J>as7nJgpZCdrwK;!0-fBR_K04-x>BBjmm`(vy%~
z#JkGqKR`eDb7?QvoqLY(p3t0~Hay;g4`NYh!#(jv^i=jvaP(koW5Le;2Ai_ZZ`{au
zg!3OEq?w+GrvMLgusWNB_hm1I<G*c~6W)Giq_3>c@h_86viMhWQO=@;ocsLv)ZJqp
zK0P1n-Yqv;!mF>2(N~VAgbd@t!2*4I3Ay}wzJ}0ue4n+54U?We9(87ll6o5fY=$qM
z^FL&GY`@FFFFc+E=t8Agp5?GU%7YA-#s`Mokz-*kWq2CNaLKrvFB4;HDUb?`YF$XH
z7SwRrZ+yynOY`;Iss|xUIADh646IK_B<!rX=;%;T7P7OG{MSyG0fz+d!ZKRG0xaWn
z<-;FJ2PW&y(8ZE&Cmg!+wi({G{2?q)i(SE@5@|1%TrXYD?mk3`=D`X%<h3c*CP({1
z_w~CNjMTT>U4@aLrV3m;ih6t4iAhrGVT<Fbt}Y_Pklijz^D%I2mE*EN)D4)%hbw7@
zTpx{--H?N=Qoj9l=jB~~il61N<1oq(5=V?D((6`f-`o+m6}E4;N;l)iCyDiB$kT5|
z=CvvYOy9kH=j-BTv*xs-qIm@ltE;Dn%d~QL%5JexcSeIR8C|b?R92G_IfdWM{xDXg
z>dZf>9yuymSfSB%*Ak1_iWrsg0LWVe;gEPsY(|*9k<>NFfRXz|W4}#QW*&3!>R#~b
z?@ifo`R!fEZ;!oh>kI#7qKw~pZ!?kGU%8E_0rca2=gwCqCO-gYq?-=jO%L8J5<y?w
zrmVf-GJ7Vri^JMm`ryl~;4)bXwFs$3#Vt7CCXa;~(%;xG7+m^`vY`T-z_7VC&)b!y
z1n0u4jWT?Fq&xEJ*1@nHfnoDqoE@+dS`Ntv5K=b7zmv`Wg(v>aonFA^;NB^-CvV9{
zHxZ#Tw`P7>`NWFh!55Dm$m`!8`gic<X|pD*%`y_0Z<=8KFUPNG;lY<D%oXv15aGN_
znm+j9%kaZ`r^*YNK3Tuq=D#tDS@zTvZ!c$q^b<Os!#7Qyv2+&R_&W`#;W#^IZq?0y
z7N;&O&iCK`wK!v*MS>y2+0KV!u;2+||IfH!y$d|`wqf5plI|^g%ctQ)x}O;Kil0{C
zKts|qFbSSt@_kKIx9#1L^Qv}wIq2C-`f<>U2Lf<B{19S9=I+*5JZ^wS%W_xQDlR3e
z<G;rD5<L24;(tO0aR>)Lw^7vZPI*(mH{!_RRMa-_-8{IP4emBc(9dhls4GnJ!(lJS
zR=YYh@Kabz0Ymj10<HTsHWibvu>5;sGevh)nw1VNW8iYTLg3nynj{T4CY)d5sz&jz
zIQ||SK<2>-07ruRe?$gJ%iz1W-tqFij(1hEv&CW+?~`Wh0!VV;6;fcZS5o5niOtn=
zz&ct>Nx|xx{mFIk+XC?0Rf+!z8H7R-52}@#6Sp(t?t>oaSEva<F9QEqUa?p3X3>xL
zsSISH)ic;dzALb67%wj2&^c5pOeGfkpKw9c2B3}}{|OdE?OZx6bgk>Z;n&<G{c{sD
zxZ<;{8a;t)DVsy?b-xx~0QS(%aqV%rvh4tF#84~X5q!0u7hlI`@f@G(cN*Ce%?87e
zuy)w?<U%#+rC%!=3(bD8G_!8}EziX&;4eNz9d&#-V;wV`Ielcl;txALdl^P#T>vKp
zc>4#r(V(w(%WiLWzD(MyDc859;s?$&q6m&3q9T>H6kPeDG4#U<XBy+eKdn5*o5t#d
z^TW?MYw~Zk<gCr^-Gk!-H%P`AjdlOx{5tA5U{t=Rm_vEu9xYJ*VnHzLo#-HL6Z{N%
zTmkLcraXP{yHwSQ6YD<MiBhN0?1o7aym#_=2d2$Q?!Q25$U}C_wWrFL@y4h6EqeI5
zYf7T_I|TRh_rZ=X*FDNEPU&#Qg-;%~@Mh)7r2;kb?A>y4=h2L&U#EyQgER1Vi$$GW
z5OY(QwfgP(M?(|8G#yfft%LB^<sS7Rs`}d+7EeEc-TUZ+y}*v`<S1;(I9LER06Q=F
zb@g8aqibh>urke6<Nn}mGRhM^d+gG`#hKQUNStXFiGO%KEWWp*E48=ZDa)sQV#3%v
z8GUr~QdGiwT@)8g^4O`}*3Yd|j7pd*7<(gS<nsLq{KL7Rm*g1qY89^KKFb_Rgtq8^
zH8JZTdG(*W!J>}|4aEiciHat(N{eST@aFXqOJDovZeO^3{Xao@C_}14BE_$7T}ke!
zuc%Bhn>zyN#{K)(Y#xUUxr)RqnA3@ftqy9M50FF1X`Rx1f7RjjvpwG`pK;epMVE|+
z*M9Ai-cAlVee^JGX<hx}(ws8FFTX#G@(&URC-u*nka`Zqzj4WE`|ai#m9i**!)k8G
zhX1>tvDwJ*`Gv5Nt8}T*%^odGp(<5UW>#ly*+*5v*Z6qOBw~@Ub#TYC!oDziu=87d
z_v0#Qk+Y;uDP`2XMa}FiR;4-1Z-gkKQcAWpoAS$Z=-*71{s7>|);&;MR3>^`^flo%
ztDYxMCiu_kP8fVooE|TNVqLT_!P53F^!>&V1=K+|l)H0&OPX*%L5dk*hGeeI|HQ|N
zBw7?@dlJyea$`#!jyx|*m4Z{~M;1SI{~60zTC!>9fmZR|ug5YVzgjQ1hR!NOFW=wo
zUWG<HKYsjxz8iLIB^il{x0sC7RV*wr5iFwS8VG`*#r+0`NmwL4_v1$;1gg7x_YEWO
z)nsO7W+5kw>ZeB?Y1WSpI<jr(ULL4xP=ET}Nf({t1H}TUB}!PFO#cS)|52?S*TF;l
zC<$PK>M4t?vfR6>o@JJaBXyjp+KaH_Xpv1E|4dNy{Qn5QQ4)l3&#|ye&i0{?*V)nN
z<PqH+8~wFp9h}7Up%QE~%Txv6mX1H$*FhZ2xI6ESf5MxGAK`LVN|0CUROlAtHiS0=
z<&#BYAIb0@r1}VahznCYL9Vlwpp6;=?myOv5O65P?}Y_0FZBul27nQd6=KKu4^eA=
zsE@MoKe!C0v|b`3le_<emW99-j9e?0t-L+zz=NP1VX+%hb`BCF!H>lyblu5s#DiwC
z5o7qRA_T-jzE9Ur3S^r?XSh=&Uy0cYnGdKXB)k<PZi|xhjMvJ+F;DDXT^xl?jRzJ&
z1*}ugEm1d~l`$I}wWLK<HcD0Yz5{QI4bQ-sUc&9-fJ{d%@tLJYgj!0gKe&uOEHo^<
z+xomU;}Ub^d32_^YnB^OsQ9?dV+JgU(BJ=KUWjUiDO?EJA0_m&llL)#oUw#K#Ub_R
zVf=v7Tn4|_vLVv{NNC^pR}uDoN>|hw;u|7+bV$ar67Uh_SW~sCJ^E+-o3%KdvS4$$
z(YvJOb%cH#{&<qW<HUbR3sL#xVCk*?<y_*d+$JRsXeE@PD0nSz3dW7CrsocOP1;g!
zdkV{?9KlGnH{M6E!Gbg$ISfD!{)~V>->Sok2a>=3`{J}Y2|sbE$g2E9SCv5~aHC68
z8<$jFZ~?kUKMaWk$n!Ge7Uq93AYo3_^o6H;wYV?`HAV_AkRM_sS!1(l`Ud!?zu;(0
zz|kNiBgzI=FzHR^Py*orn*W_4qHKuZG7}DHd*^NgwU^z^(T}8Mqkmf?)GEq`7%rn5
zeP62z=TMqmL@#cB6r7Wji#RZ;HW7dA4vA4IwB@vVi<V9fEN=hFsS01_SiwgFC6MG9
zModi+<hm+}IsrXegB`3CH*Z<SU%n9_J@H)TCua6tsQXXkG^SIp_0P=b*U2B;z6sfb
zq3^#IDe#}29C`mNQn5H#g^}-gV^8w2s%pn#MlObn1OI9C=U_3ogO?*{4WK?Nx*72`
z5x|L%TH*n#@Q?#1SPo<ADJ{AqX}WrWD8L7+{eTRiM~M#{uq$wWo_F@g#6TtxDucx;
z^NOfcb)0&QkKe_pAfK9C{#OL<7f-6`2<ROQS@lV9&Sp#Vy=K6DB5r1?z@V7T{_r7u
z^!2tF_LhJvwcA6XeUm&Zw4|(Xx7PjnjSU={eC-96qWbM-aivB1PC2xdes7){Sqvf2
zfla6f+c*4C=`P&|gUl{RwQq@XH{bE53plm0?)ztd?&kY+Hi+9eijVa?ko9uN{22fr
zKcV}usqcBWjzU;xMQyQz!;^N4rOBwVapUOjp{r6(N;L3y9(lmOuhdlinH}~d*>Xx0
ztZ@IgRqRX7zI)7Bu`Oq~R`s~)@Wu}0j(EYHFT-J>*68_BLS-&au_K{UTWh;U9aG54
zN5LpOgA6@3&(D5m;;IsYb)W`5ouJxaVGeAKlJ}<N8a}qfOhuiI<eB+*zT2<a+16UR
z#3&@|kR0*7dDIaWqtC}_7m>C3ZJ)h{LwQ-x+sm-OjvY%<fJ{Y9g+JonUce}0{QQJ>
zUs0KHDSAi6X!FZY{l{mHAV$6RT3wqwMf*C41_!d^`5FAVH)OM;m}pf%lO$&EBuK95
zy71&{zDv%|FW>#q-${?{?@&kZ#Jzj=yzQn(T||-s)9eCcWkuO;nW#c6WksBw?^|W=
zJAyl2#IpheERd0@1-DTB=7LYnzZGr7Cj9FcONE2W$HLuPw4bsPEkiO7*TUG;p1w5u
zbA5JkcH8IiP+aB9?|t;i>Ahy~Kb6&<#OdC^+9x&fN?bkbb}{$s%k>UDt%s5Z&XILC
zE4DCRv6m<0uJ3C*rT=I{Ut2o4Q2$h;{nXH46@=Ga0Tgr30?Sj>P5h)g<qCep8v8}J
zE|a;8y#J-$^nxR&GvNA~*5ghY&ha6`@UVbe2=<l4q2D8EEx;yd+-F_qrM8w)R9_t(
zBctA@zu2Q^V9R{8mg}(4QmY9`!@g1Y)Wj}F-Qsq3R66py%PE4t+x_pD#42CX&X*B&
zQ?d|x|4p{<q<U~mQA-B-Z=`Der=EbEW}~;D-1nkBtG<s`?bajy7Ydx&rM?&SvW*v*
zlwIzV>$k>qnX!k62H$QNLl|OOe^kIsGB4kzgIhv8e+JR1%!Jk`8@Z_$K9p7awzJBt
z8S}d^aJryFNE46u8S|0ON-soSh^a7f`eoK<a=v%n4NkCpWhNs_NBs>1<f4sA(#z@+
zzgn`)Va?pPEFDHeE;+=U>nrl)BA&muV7)1HjTuH#s)(}x)0Ks}zjHo=yy*S8fEd}_
zx<{p*8IipOvtlF-EJn(H^XnMo(LA<2_FwEVMYG2z+56=9<-Qh;7F{TqmFtreko#&s
zQbhSQMy^SY-~R4;q7|cMS&jXOKV^Zif4}f+UbX}!ubg8cr!6$p6b~hbnretC>dCq!
zCrv+bKXPPM>GDZmx%6t%yB>G1pRU}}f7m4WtUEL3M{5RSqR*UltzY}OX9xmSHEwXp
zY`*jR;%0V1YKnPk@{icR7ccrtHGn_KEE&(dLm%Qz6%fhUCpW6)kZP4$F48@qct2f1
zn~%rHH0oVP`kiC~S#F+7N=rS7+gbCuBz6O>O7BuOk5jnMiE8xjwb0g`Fule`%0G0{
zc<arPR>H$VT$0~LD+5G30;s|qwZ0sKwc7ABm2l75F5GR$ba>Z6ew#R5$+2^E<QFs5
z{A!3IoS0-Imth~po7_b<5)D=1tRaCc-<P@UNKB*LKT15SD)=}S7*?chJb#-boM9<<
zl#K2qDZn>|zVkJ+wVvalX7;&@Rr_)uKGE6N9L{t;S&*`AZ#nynen(f)IIHeOKBJ6G
zr<^>jbW*)@@btHV9Ro&vk_ry?Crzgoqa+<Dgvtd6U3{-pDeQN@%5b`qN_mMC0c-=$
z?Y7<=dS0PHH?m<non@UqNwTpv&#9x<$LE-N&t;;aPJdMz5&;v(2TcljX<eO@f*<C6
zkMWm2QjL!l7%vBTRLqyekM5uX9d7LIp#1r9j-OH~qk5cWVY?>mPs-fe6bF;}1cfWw
z({g_Y(gm|VdW5;83l?M?*$cNw75e#eOUb3~4QD$#ePOdwjdX9kfZ6zH$u=da@KHws
zKVmb!IPGjcswtgO5^Ehz4k^I_v!_PMz9ilD&t|Fbw2O<=U&yGC_4$~2uZ-9$4gSxp
z9b=VI@O1dXyfC7W8p&sQvp)*3cDqcSH6!Zj=t}xG`Oy^<AC*U`bBaEtA`u<4*V(&Q
z-4u)UNP+2WZf=&lzFU9e%KP-kDq-#xuF4N2CHw7Hl=733YNiuCINzWGuH28M{LQ$R
zG@oGIn1APZ6pxaH**Ve0yOQ)D{juKPj+@@NC-<OXK^GEhc>-QL2M&!Bmko-ndhJ!Y
zI?Xk`EXiwK#TEJTy>pb*R2M_K?N9~~V%$FOI%{;v%t8F;Mt27F4W8-({~W<P^m3A3
z+DeIcdROGn5v*uPC41T^3Zqap;r3H|GwOW}EgiG%M{YRk@}b56FF7t9$+<sAnEcq8
zS}MVTpoX2<-dj5M|EulG!>L@`_BHHeH%N0bwJI`85*aH~3OhqWsD#LrF>^a*D#OZD
z#&$A?ovB!c+6rY#hE?XGVqr;U%kZ6#-s$%qzvF%X`Mz~@Xq7cQ&wXF_bzj$YU*~z=
z)25_<;O8$44#rcy^E^q5qlgB6sbHk2PjKx|**%T==NkS)86A0>KROq?<X(KQ*2N^|
zx~cczx%Q^LmkbtU4?bU74Eh*@@yy;LJI?Wk=CO-ji~1p!`qQ4LN{tP<D(JO-9;YCu
z{I<3wf$Y>(sAqZGxhwBUTie-OUmm6#%s5pP61Gb$ZW>HUH5I!*L03=o{ex5DZcI{h
zy9dX^15xpkdwmu*XdaM_S7&0{`{{0m%pTGRXLWMZmA6NYX|6lU7SG5xC&YKXsZ^H`
z4rwv|+U&RMDAo3gp^6DJQ+P7*`5j{KyW5UljX#{>aV73__K4w>mPa4o+|%THYEsI|
zZ6CjV*q>rWMZ25e-ftyN`&}pzjxhsoHM-KaAFkKvN)q-dvCYWwTu26@TA%nl&-r~#
zlZ&~JxWl)seO2L@|M=REW6fyOIy)J;$9^r7d$x=;W&HQTj-t4%E`Exq&_BKSH+z~$
zlZ)N(8m4|hOiJy+yDbyzyt#cG+PAB7?0s^nDRF2xKeH~k6TOjQpC?x+y4I@ELS-~N
z-c_h^N}o{gGM6_szecu=m9HhatYQjB4cPxhsfEeR_hZ?Ovfp!FrRejPJA~Me+KpUE
z4vuqu{v!6gS%SCb(Qn-=cx0T@=-Uq^p5AAqlM#_%e5RJ&nr8Rx$*}atk=9~6V`}Z&
zw<NOK=ZW#V@Z2jf$*VD&&SO2d6Ma*uvLm65NzWbaKED_2y!Q2S`}#0Q6O-!_h8k*H
zjGqj$n`W(K3Xj9+u;ZkxWWX+aEPV8ZiQIeo%&Im6J(Q()?&5kOW-j2ODpWf)XE~@s
zL(aKN>4xgj;ghXFdqM?Rc|^aM)3%f8?tiV<aQe1NrdI5)61yw0(duhnUHir|<kOb$
zk)lr%c`uRR8Hh0CE2?D|glLI;LKYo#uM4&(Dsm~6McG2Tt4j50i$WGxGd=U@<r$3p
zsx+-G<l{aSrb;+rw2{dwx5b8T_6wS*9IKd?H@smT9oHm}wxYh!h)&=1{bUJS8BrZK
zxOpim(Dy`Q%%sxhpzStY^w<vX#Wz3X-gUAop(x>PYRgFX9~#8X)SN+4b9OJ@sL}x6
z-TplI*Plr|A!yF@9$i@tJk}j-wIWTEn<At_?jA~|=d`(=ntpAQwKs7<J#|i_%eW|^
zuDWI9+xa7ShrIHQ>hO_M`+T}uN>D{s6WFFpJHGe2r`1?O%KjA^oa+1Z3zWO>RFR5|
z2b*&btQ8gD7a1S4s?d+I^}SM+TqO5$R6%s-&iVsp(hWb_rk1W)by!t#vKePI-o9D;
z_LRM-@SZa#{To~?FVb9>-Jj*15_+ng+grOMp3G5>X~T;)XPR%SM-6%081=yQh^@rz
z#`+gcjY@DG;Z)^d+s|K(P|y>95V@87b<<R`aH7tvQT3?rwMT<5Pxjf1h_KUkALflv
zkZ&uHeBZQivu~l?H&bZI!+r2TT^2{-$o?dWMayrNJ9Zm|=I^(*3*s``F*s)$!PkG~
z_)ubAM@Vzpc7>4UBw@o<DfDzRJ{xUfT*TUO9XCS~jHe~J6b^5lG7U&zz4Vp&VE!mC
zIdY+ABj<*;^jYQGe!j^g7NztTf3!umI(>PnM62MI8}qQFa%b^{IJeE6?R%OySeEvU
z#Hk=lXC!UOR4KTT=pS3+DRi`}qp1BXJ(Yf`@&L;&|I}Z$u*gNt3Ar#PnDp=Ckn}<w
z&kM7)J`ThE;e2`8m!uoKRYOf>vM+=i)H=sg6s(4$y1QaFlVfB`7DprMd${$b3&i~|
z1?=)IDKopV#`lU9L&e;^(D$K2QeuXYzTx@D#raoHr_h@dhqydw8v`moIh<2Edi3b$
zv6-5A2i&h>(cTwke@?yH0`=+qX7Xj)WW&Iu@-Kc?rH=|197xiNex^qX=SQQyD3?U5
z?C&~aY(z?k9?2Qza*-XBGA`<Rw|&=)vQYJGfEGI+V*w_%*((3ZEn8L2wAvz_&;kSR
za*yw!S{cVR)>#P<F&BffLpf@|P_g+}3E{`OiYJ3|m?g3mb@rqr80uy!Hvbj-RbTtj
zQ2ZcKGb)9RUKk1D^Up_Z`TZ5I*nP@MJ6aQ7P|~8N1X%<}My}ZO#steWa5A%oP0zNE
z=&*E_oitM7AWvzHJJq>vMx%9+6j>|ZBh@D8fzceroCi>+V9gM{oZ!Rtxf!I9mXpJm
zjMrT;{FL$ZZOS?7D)nu&*tilcVV^na9aoP&gAasYsvlF>OZA#RbG}G&Ru?=eTAvUv
zZ&t(Y9jo><hdD}ILhrHF3XN@69>4LQ$Z*`dv!&pn`7cpB)wd0LC}xlrT5erbj^%pi
z!RxwgT{nD)-7Uhak|w-I<CiF261~2AmjJDxq>;vm+<GNVb8?4OtjtYs#f<Nkf>FHc
zN}Vhv%KZ!Xix*B?QkR|u|4=r6{W`WS;REx|-4*HSYRv^G3Z9<7jO%2oIsl(P<g_RG
zI5n=h>?Pwt#^EH^U2@ck(oQ*RnWXMmy4TXco0y|VnSBbG`9dl(d4}a(?cYaNlnePv
z6)DxvGB$2}r?_R75q!^S<?o*=kjkM-f(XIyiw<omJ#%KUX^WMOr@NHIP%hHF17vMu
zH|`i(Qg%gJ>E&J~h3-*Ga>0iqm&OC?H|{nS5%JpN*v?KrKxi4Ef!%hOlyT70zM(JX
z>rUNacnhTSFIjDNNzq#J<<;8h)xM~G=2Po!xvH#Dr{aaMYW<M&9&QVc?`L$V>n7Z%
zvwp2qNRqNs>LIdtra2N;h{XP-o_-eA(;WJo-+ha;`%fv~&TY|jceSp+t@0<8m%4TP
zliURHLRH2gZBm%7)=ZqkK>io4*~q!EnM|Wo!S5H@)?_GAlM)Qad#d%$GAm@w$1C>w
zYI02Xe~buWTbv`lh-vQgW{>qDn~N!>&Q|V&rk{hVF2DVaHh<w^og3}f8U4nDA0`r0
zCUrK?o~RpF<nhRsX+7q?yV=<EP~qI0snBZBOTGSlrXz7++F0-0_poi|-W7Xy?pdMH
z<$3mXnWY|wMryka<v2Cd|H$2CAl7=-jjJ~|O8~lA4*0U#K5Zi-8K`s;fc+Amufoh_
zwvtC`HPtU0iirt-V8}b{7w6-MZT9}*wbMC;en7aqPeOk<ES1HlXLat?<?H`)cm9@1
zWm8MbL`2g3WsnMyAY}Df^VMPKq>XxydMYbhJ`eQ>@33&nKV)rZ$JNyHuxC`1pC7m{
z2^pD2F>zVV0IPm~ehxT}i)@1f1A*0jCNU9NXh_v)zN?Z<G_Rq%X5ObFu%MveL(zzf
zwDig4Dze(Xdusbo)z;V3)jg0RRBLT(J6RHa-h8LZ#fzWRxQ~4~h$6kAp#W*@dMN_j
z2=++rZQHgbmLBdcwy9uV1ZQz4r1Ag|nqi>duSu1AVKUd(*x0xX_+C{m_kB9J=0nT(
zcpgorZfJP|(>ywOh$g+fJZ&Xj&G;EV?s-M(Bs4W9!75hSBN-RBT>W?_oV98mRL%S(
zjKZM1-Yh!E(Chk1ssZS7G<-@PAq<x3p%EIcKo%cd1MIP9wj1?%<3^_r8#}@=F)2QK
z!gYHQfL_vPlgr%^l+>qBuc7JADL?U1(*U^289BL`nLk4MJ1x4BKMSU)OoaT1(+*~J
zkdhc_(GNlC|KP@e<vvgy^<q{pd2YWCDb))`Us>H}W|G7?>=Y(uz1ClfNwkrY8t&e+
z%IvqZqYD6*DLo!dA4aY~690bgE&3Sa;wC4ER-}a?)6wO2P5Obcy~2dlYqG=&&vM5A
zo7WD`tg2xXAOYCvPu1%${d{RO?OB}_*h)-d!p9{v`h4Gtg*2e4a2*=*6i0X6R%JFV
zzaUggJ2epT=5gSy<7Ir?+`@X96E&;}EBQI%CFYa6X3(z%p`mj^ZCA@YLp^$D-!m6u
z<c2cx@_rO`*bPy-i)@=E7&(nwHJUp|&kEN%R=6p$!%(z?-NT6f)y_K;_uf1<Hz3Rj
z{g37b(C`^L?x@GJ$sxBiq`Ggp7gAv7dfv?}^kZsqo@^m~QeIt2lE%~pw4-v#e|aV(
zE!BTh{H_7!zqYm(Clg0yWQci+jEdUei-~H}adDXhnPHEcH(1$wsBd^GdPeEV5pVj5
zmPSDVf%g(2IAi(b@L|LER4d22*dj%qaJ7AF%4gTE0(Hl=RbKk~J7y1|V~pAFShK|J
z@k6#RFZ{-D-XXc9^WFl-LZfK8{T%mvV#Z5;x>qbp-My7HN_+KnT#+|RnWwM*zqK*u
z7ZZo2sI;NYk=oJ%>4Iz{&+MKe+ql5f(AXo)=kMd|a-(86uU4ra)Yv2LO<+>~2q-(x
zrss3Ctba4m|JV`e*>EDCkh{7{rK3kW{hwcu@!wndk_-R6l`k$X;S72t0~gmU_*BT8
z$+U{S8xtYvwuv$|<(j+hjb2xC1zP6B&~Ufx9xiX96huWv&Mf$_0nmMM6J=tc#N_PR
z?LkUt!?VR=SLd(I>%8U^zGnGkNp)_mK-jF8yqi?oby@I_emc29oqtWpU-wg%22>S(
zl=%M+eN<zo`orkf#Bv7WO;}Kn0%E6%@q|$bOikIey?b}T0BHVvk>|ZP(b!pE-;h3f
z-uXewP>D>9B_tcJ3)<2jzTDSXFuUM$L~8$LmQ?a7qXS#fgj>h_K(bQI*?0%}oKV#@
zUk?e5b+?a?U5{7jFF>PVN@Sl@-OXs+>f~~RAkOTV1ES5{T^!OQ9V<RES3R^*SQYmB
zwKej54r_ewORY*`Lr)&>zx3o)ql@qJdfDFWC*{ms%SP}VZ34ujf0-N-Poz;%C;mfm
zBWmZwlkipl%P&cy{&L}*ghsxtXTM2L(FkU3`gd*X{_X>#(xH{i_luNB<kP&$ukwWC
zs|utMIuTv!G@7G=++uuypO4gOvYOJst;#B9DUQ&->pl)OxvXXn9G9Y2E)aSm6=Nax
zfok5~3*OUR-&~}3Ci_4&nSi5jBPT6Rqyc+9UBYqqOwUZ!k!~M`Et@;N_#1kWOGTK?
zTDIw*C#&rvJT_GPJ`Nd*sd&$P3Lx5P13MimHz+KQg_Vgg?{s>8YkBN$a46nb=7`{7
zMyf;%)iPX*c;d)}TndkWd5KsE#~Pedf{%F_z5&sXG2%iXQicZ!J1mY*KM@y<SA5ZY
zDTn=5k*AmVk8byGoYmJ>{{8|cRhsjQnJ!w;0S{3|A1Yj5+mItsllj}6^=>9|nE&qu
z{NozsbJr<OiEB845(sGfP$g1fVIla74WB#wkuzjMz3m0j!T$<waU1^$ZqcBbJkBL;
z*fn}#;>7w^$Z~aNY|#AHvpCYzAUA#fqHiF1?5$w#{2aBy2V)70LeKQaI{Mte1I5%$
zy-gK>%_PnD^Fbb92N_xiKs6E^sOXCBVRTPJt)nz&WAO;253l$B&MZ{BnT)B5j*iCO
zTQ=Sx0|t^3{6&f5lyS8Y%T+N|_7LxQIS{auMZPDaPhtvBM9U`bS;a2MM}0KrmfKRp
zvr=|$8Y+qqQCs?Jo#=dnTe*&bfja~z3F82M=6XL_73ddxcVkY+b5MOd!76vzpiNKT
z7-Z3TIIphFYs)ca5MA0J_C-%^DInHe(exEVt5{`!1MXr-8Lj8jW^)4XBR+0aQBY#{
znWpqiQWh&gaMBd!<2;xW0}B7w#+8ek9&8Iy(SLI8ljgVX+@B+|<{GQ)+*5F4jC4E$
zdZHUJOGt2##IJpR#XaB4T#{1(zY9RgM2TYmpD`vL{A9Jd`v?1A->&ZN#b^L2horEP
zUGeK{@sK7iqNI4fFf=x{o%kctvE!XSxoEws*xa7?&Cf=6`TsFAH;PL`pyE51=1a4C
zgX{T${_q}H`!u?yV-H`|r1Iud<UM%94OU;D=&L`{vy}{rfQix^tIp@wAgS)aVF_Z(
z?kNBGXO@jzY{g$=uX=RNx}pi$SwWBlEs&Iyl!Ag5vAZiwLdt_DE*JSVv67E%*svjh
zm3tBPK=*mgKg>!#1Ia@OMyQGH6cQ1kFHWV!{2UBaW}CwP->_%pAe3D&B4});7j}=L
z=f%^kG>}8Vr}J98PeAKn&_BM0A4D%+w2U;1#zgOtLaJ--W#7Q{a{<0H0bi1ib}f8+
z&^6g~==ZwMT)4BJ!PD24<>0BwFE(7X9BNC`l*R04-J8kn?S^hmWhRI!ok5S03!n{H
z^8^J2AxCSdukY3%{mt24`R|5T6G^xl7%$q+<@ECuULB|%&IwCoxr`&yg@I*UOi%6h
zpSYOGhrIpJZTeSKjOU9Igxg?GWO$WBhyjoosH@Fj<mcO|<*s0=rqmB)scBir>|EHi
z>nXG$VJD_8t$Gl687WR&*rQ@#;pRPYnkRql_W@P$2V2R%9ca1gEyb~&d|j4)Oru6;
z?7A1s7IY6qwu~GKy=36goz6-6imM2P%ci?Q@xzhSNNsH`1k~OV2p7l;FZ{COd~ZX@
z>3`f|pu)nSR&1)UN$GQ{2UeK*--eR(BAFXXqqerTPtQG}dQ2N>TMR3u+O!u8l2y5^
ztgXe@n?W!V?swG38pt2C*F)?;V_Mk_At3ZB!6*ySEVC0QDiJ7&PUt-r%;&u86ZokE
zNC@EKb4q`_EkAMV&wm+FzYl+K(`%kVlDw>qa`o-ox3cLhv-<%xAtYF58-2TxY}^+0
z`V><!Z{9c|Lc$<{huXEr=xe71a1l;B2LC>km4v|u)d28Ao#y|STf#;;cZNs}0JSS?
zdsb8wtcFH?=EHl!;(;@u^eosRoYT;#C-4)Xb@%<=5xV<`CQ*Iks6M0Kq)#>&biso2
z9|WKU9t}n_B%?je^nv@*t;?=d)yCE0cMf5qG&Dj2Z4ew$=U@#%YPm06ZtMvfvsxX3
z3Y@@6rDM7($qMtISmWa2zSYMO&TA^JJGqIFexYU?eUsA6O#@6w>2YdmRYF;Gd9cLg
zQY^HCr$Ke<!0`=;j<$ksW{Y9ggLC4Pp=8UI50r1HDu_^qp=|~Lg=eS`HYI@ZxYn-D
zDW9_*RX@?-Ir}9B!ouI`9ykG8P~ucEz3tcA?Ck75<8eLkf+<uT@FZjgo(2_^@jrck
zGvJAPNKl1$?@7ZgPMH7g7Ma3)>VTXA9U6c%@PdDu8xTE1zEk;0`y<`}>9~FC*6YSR
zjsyTi`7(4ur>xMQK&#qqzm*)o<x#WXZ=!U%*=zV|bsNa_K$#$?e#m>9Rn2wgWbR8+
zv2k&%oJSuI%|Y*mI&Fa2^ZuJbYJ2C_tW@g%PdJfvRqK@7mti4JIvy2^OH*{Y18->?
zdS~*Vce-hW^ZznNCy`Pg@n5uJ3psizHY%n-_uiXz5fLe8Q==Y*w(rZ>X~OH|z}~c4
zwD)YNPDE^^hROZ!?Dx}ILLY28^6|@b^#j+bXX~;kLNv*qMpMymvW!HhZg<NEmW`eb
zFo=%3ch_$tBwiua!qzl2xwyD!Zf-s~Iq6Zhh(&)lDCnuzEzfNt4vY;|-fLqxWv@ej
zymgJ=ML41jYrcC2S`a$}a5kV)B9we3_2>(<zJ}GE>93i&y&NhSH<-yLx!mHH)$p+f
z1|LC4?sxO%=FOX-pOM@=NOz<5zEnST3XqUK9>rWL<P?JgV`B6PCQqor15W~83+%TD
zHL9zsz)am}QN}SXaK3kH_g%&&k=i$J_(eo~J_qq(YMU=G^#oku1yU(9Bje=h(*+eL
zIvK6i2`taq4*iTTuQRp%3@3Y6sQnC<zBBFn`>!^G<r)uJCKlP{CtNpH|BOyq6<00?
zHaD1geqRo?UaSiLj~{d64hVI@HFY&ar5R{C8qK=;B@$Kk5S8Snef#!x#?iZWfKh$J
zIP{Xsd%UmRHp!)4kpOe$G-WSR?G?l&?Qr+>4BK@`MG5Hx-q+WQ`EmLF{*>rpw`vp_
zevFU%VugIQt8uy(gtHPQ;yqN2`(np9!I>i+s>@P(HVb;^&Q8-%R5L>?1zu@ey1(p>
z7_<>-DdCPBx}es#i=AD71l(wLW~Q~!Ft|{FzTHBG$X9rHxF=Ns!Q;{<49tz8($Z4s
zCi9Vysvu6X7aBH^CsZlEn=0d@uH3_Zl(mXkt0`-BcSnaO$U3Lfjp{2ei$_IAcQafe
zv}g=nVN+Az(Ku8LSU2J$6(c5wsJSp4QXYN{u{XvI{Cf#RFQj|n;hQ)(Y>X$tRR<ii
zxTIte3={^5$_GRf=8)#GUmZf;1A-eU+Kx?6`y#BvZd;156?^5W7chpy?DYr)bt%>-
zW(o=<h9V*(Bf&4Eq@qG|K3Zl1O^FBh?m12k8!*0>mX+bfpqdgU5kxg0W&x$Fkk*#x
zz4YoVSHH&k`t0=d`;xYXP=Z>uY88Hom=|#(#D90o`a-4x^p+=1oPhQ#$l3`p2mt|L
zl9O{PiB$8)W`33AYrvKQ(;Qe);S1jZB_jlt=ea=e>fnb+Io1NS9W>v!;DQ>y3P|b`
zYbZZmGF}7;UC~u)!y<QtO0oeo5*FME1m+{<bqoj>20{gG+c@~bI+&k8g3R$=d^|@S
zH``Wwd;1|V<*e4=K&YKiT(kH807BWxfuBaB>3)Dpj!kAfd#+~M7BaSN(2{m6?kz7b
zhxGi@*&1+bm6h7~#)O52775wTU^f}P9;8dTvqLmd!Pm!cWDyc48~riqvD>gj9%i5e
zQt#$8px5nGuniqo0`D?GEqv-CGTh&O3!iJjxEvA}hhJZu>abUscy}Vx0o!jIAAZ~s
zuy0PSN0bfu?WaJ2fo(J6FWeQ>Q;crzfU*k-EU=PtAIm(S1y7HSK?!&f;@B7$+~;s#
zU-DaMNoV)zt*ihoDWK=K{QX6YOC+$l;U+eUD4VgA=J)9ZIRnA2ybs)Ja&ims!GKFP
z%1y@ykG5n)xPZwrRDe+N1(UWbz4JkAY<s??7TXv!K{R_gO-ZeG(2r2wCm>)A;Yo1(
zL25F(kQJ-iNz4m`O-~31V)qUb<ZOg)VtV?YBBG+A`Ye|(U3%o=uC5M<Q(-&iWatNs
z1v-;#u+nFBS)g0VC?2Y*sllikR#zZ#5O_L|Ar!j3!6FSHK~T@?f01maf@My-`3J-?
z(zCMiBePbUWVKdRSAz#&+%D$??oOvKd+7*yJ9*Cu|IQMZ+PS@0Dy5kv4xerc9E__M
z&=0$NSBRhgsZw-IjKX-`q3GFhidL_3e@0rOfT*bbpT<MHXaFmF#g3)DJ3V_C7;%IV
z*|p<8egL2NWobG|R73=7T${+qdMd?Jo<KkKpq!ks5?nC^3)6pXZf=Hvy3%E^WWkW0
zg)R*whqcsAq1eYcEq3V8I2Ir&4|DV3!?~eM@B;z5tfl0G)EG34@4xz4<Vy<+*Nf}X
zS6WtI1+nZMtZz_d4*dmF$nZmemgR>m44ZOb4RFp#5QNA_6GU(U$xTun7ez&LL3?oS
zo;!GHP0lbw0Ni2f7VbxQc}~O#med37!^*iuUeN#Gly-caA|@)TH6nWi+Nlg3ImiTC
z>z5+1&AJC+^|MNtpwE4Mwl@n6N3i{)-fSn=+R5Wm1X)R(Wqzdc^6_PPwZ2bRhU~qr
zZe(AC%-7-T!d$-dV_b3`AGW#<ygSC(vP9u8Yqt=Ul_hZGk}qK=`4v~(B(A*UVeB#T
zIhtX42v$SiLip0%w*mrK*(KDJ2w?*O0o{v;S4~V31K01Cd}3Z39o-A!g&;N1i2B8q
z1O>h|Y<-fq-$X9O-X2SZh|Hy~J?i7ioTWjylfhtgP~p&`lsmLUvu+P)e8QEX;_3b8
zf~Jf06)+=B>5)GFmCNe(QG|)<dG(CtdM1Z3U{3?p^z>SxnxOjzLX)V9?O$4&FP)Eb
zFabqg^V-6WP9VFL9)n8Z^XogL&0JjFk@O4|LJ;cQjm$i)-nb7rQqm?@p`TRD7(X&K
zq-geKql9-2HZN&nYwYYpt@<}gxk2J1{l-RIB(`-!)yy6DddMGbazK@kovkNR3AQiE
zvWSh_Z)G)PZDX_vQJz!kGR6#JI=}pw%*+&EerbJP)9Sm6H(FZC_hRRLuBxV118*is
z!VZMR298)Dd)df<?RVG&v}Ne<u?2;Ml=Yp>GR&3)6}g0ju2PF^X>ChZ6XB}Y9S{)M
z%r?nW<|%#HWJg1;f9lk1z8iSJQk}8S%StU9Px92G+?*VU=#GLJ8*#~N`m8(K8Bt9J
zGJ@S58|y*qdYU)?<C8_*Q+XpXfr8ky!D69rALpiMUW`3{%4wLSt*7ULjm7HntlJi8
z857G4ZceyLY!wi9zM2Nk8&-D03_8GdrEbdx0B%ZR5*Qc%)XFxybV;cnF|~DYcNN6Y
z5GaA56nr0gK^{XLSm9p%HEeMg@;{aWgOBn9__CzvaYx@ES;MwrV@kPLxu)>EFSHZf
zxY8oxIdk)2&wL~pYCzMKf(+Q&+}Ak!^D@F<M|!tBPbz868DHqwps?+e#4o*C%V_aH
z9RjzsgABq?A24II>7g=%yhT`46hR6qMrUr@gFYeaY_`&6#BdNZDh$=q(9jEO%SxNh
z#^?+wKvLqd#a{~Z-z!1=`lf^i*9&*Ioz3s%cccO(63(i`wee~tJ~7czaHF3enw&yH
zjcoqU;*D(_y67xq4i~&S0<O|JV2sg)5AHY-({iB=atfj(_7GmALden)jz9wzG!|hY
zAsbj(i4Rgyse&}H@^Yp4eaf9+){Ps9TOg8%`I#~B93le1Pwc)>l?S9fNbUUjc$cZH
zv-jXx56Q@MTo7B$%nZsCs9oX<y}*Qus1*X0KFHr&;^#q=2&q+F0y!^u_KAb-wi2_3
zP^tmJDWRoB#A866ui40PKQfY_GX$%cg@r|;)U`{P5$qr&cp0-<HNTh|PBU0R{b1;B
z8bW|+++B+eDksh*NFb|7tQ=CCwr{ss3VTyqE4Y9EYk41>=WM`nAB+;9?;*7~zE+I7
zZ3aPP$9}rW&(9B=p{=1V`Bbz}G2@d_5gDq`dOB5EMGHm**nNw@ryXqQ;V~|E89K|@
zFm72Ni?A#;6zgz#gWs@mc4n+kU<NW|6vjIAzjg+y4y<6J_xJT-PIkJ)-@l(8{p!d8
zd3l1iAI>QxG}I0$m`Uzv$Fos{VOoi@v!JYRuWpQ&KBS=F`}tK65nW7zAnR%0yFX4K
zMZ`9LFzy)zegMTbc+8hn>JcF!A?Pt`AI2hvh7{x!cCPn_NHQJ>1Ry8^D7+&gNqX^W
zJ9e0hMyxYjE2dx#Gc%|X<)=QFK_UbdN`67XEo>mU9!gT+I8Q=M`r>`dk-WWdcr&bO
znws56=|MfBg;RwJ2JSBQ)-FW)vK?@r$;%JWZCuhZ7f0#ohtgxjqSMJZFMN?e_=E2@
zO-)TG4&Z@YDL><FU7aq4#jx!#GK9w0dm(jny_9XQO8?idb2BqSkTXCj?B&as-@u>o
z%)SSr^7R%@6+R1y*2_v~TH$6IRS+)UsAk|~j;IcOe#OEN@Oybdf?dXi<~-^Nv*^5o
zwByFo%7`}w*c6eCCn)$XLdOZ5ozTQ$M@WE4*zL7|5Bo0lkK`q~a=G^Vf1B6*xAbU+
Z`})gmx9?^0780*KsjR7#bNs^1{{b-xZPow)

literal 51560
zcmce;by$?^7B)Vjh>C(Jf=cNiC1uj3QqmwHEl4BX4HiR*NT@U@prmw%N{W<7r*wDM
zeCwUr=bU}cxBvV7u8-@SYjgR`yzdih-D|CT-4A{WauTObTs(n7p-$bG6jef@2t`n+
z!&S!!;9uO1h`xgV9JiKKw?m<5e3Abj3LalJM4>LB?u*`0c8;DKa0w${FUMjxQ&J0O
z-s`>8Lm8eG718>7ihgMLsXs9R$+cszE(W*tyV;c$-<P8ZYG^7pDwVdhwJWb=tG)6K
zb%QOZWb`h<Qz|rCx;sp))2F8-^=<0Tew$Y2A?kP2vD^0Y!gO9LDqg9xw@(ddXD!TJ
z9p9~hS)owg{YFzHxZhB6!33npZz8nuRfmyZUK24dA^$~<9XcU^`^Clk|KE=sn<DWJ
zq!x1T-rV0P#Xh?BP($M?`qk*I+qd(FUh);bd2{CbwVByj2cibGHj`TygZw2QKYlDg
z+eO2q8~*k!|EseBHTE!;l$7-Kk%H*V%#1)wI}H=(`{bT{^RAaKUy7m=_@ygmIy0h_
zemD%!OT3zDev^d`k+iX~aa+9Opzk9RpPnuwI<oBi`0<B`lTx282TDwPZpG6K{lD6Q
z|7TnM|LH}S;`x#D_jVt4sb%6HAPx4Pl~T-}v5BB#{}pqbnT^~T8BQVVF5xZPBm9sc
zv|cIgFMHEvd1=dmd*|!xW&@GS3AimaxxcjUgw7J+tiW<hYjbZM@Yhz9Z)k7+`s4HJ
zj!%nKgds~(@8ZF!(QfrK1U@2)WWJL5ikD6+NG+oZ)Gt}jg<qv6p^FnYXGyx7pio+&
zZ*Q4wn#`2g_Q~rT2V1kuPNnXZGsllUE6E*8TYGGB{^%l$MT94Z#0SNJJ1al_+^bFF
zjO`x8zAd)0YKsg%c$ZB>eS%qL?!;uFHIIAQ<H9tNOV;j_=Q14J<5Zkpmr<TP+OueN
zGl^V&OokdeM^8;MEBj{A8Eq`bQsXA|(ix>$qrF=&Tf%-)VKM#q5SJX^jI^_mnAqhH
zRg;AuB!UZOr5|29&&JMvj;zimGBUE)>RycI{ri4;=hxTQOP-CTncRvzNd5TvGYf6-
z?5ohMAB~MPb$&Ja$k4A}zwGA)Y{xK@E+wO54DGi{9p`(yyWdG(u{Ul|8z^yjAXS{{
zHdaqIpuVDV<qQi4M`l{uha<7YQxm~*tJ79DY}UT7TUH*>oRJ7qf{)L>vp7`zva$XU
zoYa<<xZUc7K+`$zD=55=3%#Y|+n4v<yH+WM;v*oWg%4Ft+PbX?G#yLoUodu%q=;`-
z97ewD&fzrmBISa0bFt_B`*qI`ouI~l`ZG%%N%_7J@!>t|?&5<9lE%a2Jk56zL>VSm
zPVY&L(W4f)gw{WA`|CE(9!F9(Kcnlpf!ogypE!g6zMt|vk37ehHXr3g_av{V@~VWL
zu-dz(>{KtG1Rs}+vs0qf86jdO7peQReH3}B)%kUa4Bqyjoe$^{iGaC=4xQjw^qaY^
z%1?Ng&IysOJ46$NL#$!!d0$d7bK_6R;$+JK{*Ob`$`=IfeF-kz^&x_1Sw|ORADCMP
z<yep{OE>DPbm*uTFDRg1;cfOb+4ar2oQW1y8wt^}70c{Yx4kzUCHD?(3FGZv-!rFO
zV<}LIQM5X3iYzU4Bf#Immtct-x3eM%#83LpsNa$=W+Ag_nK>7AOE&!zJX=l__vRBC
z*6lCr?TBj}#H#1pTVi=+?_M5AKt<z^fLZG+>0m%vo{YP$sM6tPJ4R(2Q>RPD%VS4i
z(kGOy=lep&IjPMTCr{bb%C5*haAHQq1>hD;p4b{0{^wj{dwW9yC4XTM<@Z*q<06gt
zDSbVj;VxKaCg#e=3#MPIDaKMBf{LcVodffg3OdjEq}EdNUv_?J;<XY3B5p)0e@*9n
zETxN$>C*Sjb~`^Z`q2;Kv$98qf8y-I``d6yND}c=O)h^7^SR1}<mz*3v&~21&*5J=
znWDv=<E-;&)`*NODLSf-k@G?(Ke=}&eurbCr5H0_*!_&@6B9l8oz_lbCQ%x;u+<E=
z!SOdAygI9Wt*<**TD1Dse0KQfulJBq3jAohQnnGCoz0_4%$85QV8RcYYYMOt7WfV3
zW9sN9kI{UL7@y}g12s9D-BYbJCq(-2vzDtj`BXR|!~Keuj7(U+hEZ<Sb-fdY#qrZN
ze*Bxpzq9jG?NPGj4`PlTf;yYDaPsx!xV@TE{&{#eqxZs(SALkGq+6;E!Cm=U@M?7t
z+}ZMx+6@g(m8t93iB_|+=ce?}hDS`35MJWPKbuK!;h;NGSo6$rvgHz$>ePkh0BJDE
z*SG&#^sy&H7uA*1l)!G>VlCt|H`t}&***AWt-w}J)Q`CQK<s9qE0(8&R08?3A@0kU
zDhY)&tOP411>{BhJ@@B-l<qx}M7}J6o7l?3e}OP3`m@ionB`FS&=~i4yOH8kU{_;l
zt(vUgt=Qz{nOZ7ai!GkMOCFR&^^)S2TrZt)oAo>EL1DSB?R&e``NS|}>>PfK8%$?^
zyp=z#-LBz}8s7)2g!02NwD>VrFdY~5rDyrm#nqUQWyKHcj0COM@jG#_BryAe!ROR0
zm|RF#P1c1ZZZdC@d-%CNpVg42JW}w4oJ{em=YEr4qr?=L)u;jjM&1JPO83q6GKwEZ
zVK90;k82q@FD8Rok%-~GYa8Rw-?`T_5vz?xCr<}v$dJ&c!ssh;{Ai@@+|-_W?Vuof
zBr5r)rP55<WjNbbe2z25an8KJ%=kPJDF#D#viEFq(SAz@%?WQ>{02*Wez{{Ze{xhZ
z2D23+nQoifRhk7}orkyQ;`&d*W=-!cf76{K6V@<*vvXbh4PTbSe|hqs!6^O$eyi9g
z4&*b6Z~rI|tDaRdlP~UKorP(m_VJrVf(Ij|<syvp__^#<v&xSF@Mc@Qi`Ji;VB*gp
zZ24qxSG1p2a6Z+|ofTQL)*bws?>_2_s84kHA?t^6Q_brBXnNO;Xz2oOS|XLh$)`dl
zDg0;sF@2lVrM9Wh?;^I&jvvh(CR!BJ(O|3qo3U%Lt5w@R<oNqgG^UFXY(^wyxVMnL
zQJ1H4u!U+QFlsC+8JQCqZb7Y#c}9lLH7jfOD^CMrH0uKNrh<-KGRB7!!Mk2Af_oGG
zYxcI?=bD2G{gqtl57pz%IR5*#=T`pPyNj9kTNpAgUZ`};8x%#^(&3&u_3(ed$Yk+f
zDXjgat{^t;)BQP$qN1N617XN4(r{u#ZTsM$(f;&!Nl48qfhKnV=KM>$_W1_sm$TrO
zR(O+o760bG`5$V#Nk=lW?_cWrLrSKU0FlSe;&R~Dg|Jak%AZ^tb7#n|%ep-=G+R<Y
zMlaw;D<INE0@!qOqCGE}$bYgWItk}4a)39BdCA+{Vw0VCn|jrKHl?I|cLv_fiN~T1
z{ae2^?UJb^g2Cjn_0Oibl)EAm)5Q-v>ir#yHp1M>{Bc<H7;jX&fJN!O7YzYy7L_ih
zmDJ5@APBpMH};&L{e|3oRuSFT$(E%W=U6sflMIo=b;Ex->ZdlHCb5w;`Z8Ll(EnkF
zpr?%=93bCe+_GNZ_L<kD4=_`?%7C_P$(?WLs4xT|ix0&ol|4w_m*X63s|`rCna?D*
z`dtt9#QrZ=Ats>9D2&y;63c{!bi{F7+w<B-n3&uP+-VqGg&Z(MEtYbKGURvF*XFJJ
z6%Dam-fW|Wh+Wl7nl$}b@*kg`9^0nmE0jIm{u{zr3LXP{7p!`B8Q$`84{=CM2OWve
zO~j>`79+cVx^zj(bNK|{QhO%d+@a{@_ujyxiAKvRnv;{*^9z^Y+w<3N2$>-U#V3eA
zJgHdY@-J7u=*3_#N&rx2)wDnGpY?9WmqmFm0f#1u3Z9%naOnHHYD2YYz@bZcqq&*%
zk2rJ}?{&?Xf5agIeDWbD&+BM^!1j2NmTXzJ{Q@_O!{BMyK5`1&tgJ6y?R?VoUXPj&
ziC-r2??yI+)d0!~eR12DCwlw0gt_r~i5b!AU{U?>2jznv5E~URIIyUiyfG830VcIB
zAyyqs(F17@tspWSk01Vg!IwmQvh!bA5#&8$q{yo|@UK4ck3i)86mL>;1`7w(8rL*y
zM$seTYSpQ!`?#!#*&b(7!&dL&=JF^6KD;>&RuOHern0N54D-gDOdMHv0Yi>StnCBz
zNNfiymuY0{;Gwaj_}9u<kNSp8Qu_Tj4pm*p<B;f;qJ{GM#>VPRu`3P>F3nq4g{uZ{
zcvF9qJ_81G>1)1~L`0=^dU+go^#$$*Zdq8*Fy6HlrtX@nzEnd38)s2>u|wJUmk_+H
zZ{Xf~;@zU}kTb*DM$~li*caAlDywrzYcC1BAK;8CUICD(gYw>Dd3daJ;Q0%5w>7NL
zD3e1tBqD48qvAN_x#j8lYJ=p`CeB?IoFU<{zhm7f4hhdl*Fad&h}pn<)bK6b&NH+9
zaG(#EBrLF`AiI8Ya-uJklC+EF#9sjBefN!|=tR>k;laz0FmJr$SEAHT<v)vxCc{tl
zg>CfM8dGq-Q+r;WoXja9UD#Ki2e|DaKXT5TYqDu&!|Uqb5eTJ$!#?k3c|ag@^WO=F
z(IdX}8`YmJ7b^e;L}thC_&8CgBqtlK&vcwRb?Pt?2{qS~?vM9FW!is^l=NIw)z%(p
zZEd|9i`{F%3M`hbiW)RW^WHRW78rl0i{0reciR*Stwfjv<8S}TsIn6`s!bPq${{oH
zR&87-@S18(sH5|8uX{y<R?^TY(fj`L6ctyxX5m7J1T6~-OT&+n9-7V^k>T#c6;?wP
zssae7V#4}iK>;cQGe`kgIF1W$btFG}=&tL;filV0;#{iSs^N5;>$dsxg3@y}YAF5F
zr|wdx#bB?!QBvW(rOzp`J1Jf}De37^HF}lU{YClrV;_QrcK(ohIgf{{Z7dE6mtWBy
z_BvRN+njeB$=3X1UE~l<2Co(Jb$Jx0qWNL&9cD{gYQU|6S}5%;wNN@gB`WVl0>6kf
zmql-$=l;&f$cO};M_yv$T~XKdne}IMdsVlZx^&#;hrd31pA|Kr7xLh`?CNGE*2BhT
zS*>!XV|4i!A+kEAU^OYj{SZjynMEcl%2nqB?cFw|8ugt9Cf_=>gE91Sq!90=s-%gI
z`Lkwj&0U`nblafjv;VZXo2Hi4>^?9@!k^UJ+dEL|B)m2JY<zrtn!!&rF*7qWJ)P~0
zUv$GLtirqVhLZ>2N9Gev(K7#qY|~9Q9Po>X+m=*W^%cm9a+J>HoQ+6%`_Zb&8hK<7
z9FTz3U|Ay!?965fS6E{Y_A3_S96Ht6$Dmjei6Shp%AaUb_Ir*g5^Q-ju+C7qB_pqK
z1jWmm;&r&vX;FvlK3&d%`ZR;v##{q#nmvo2oBS=S3#E${EgUj*!n&%eTwVuzUAkT#
z@bGiz&K*8_@*R)ONbzKB-#0n{e(&}ZmIur^bLut_Q+C&KSyOhRA0HD*dEHJ}QCzGp
zDkoIuW@D3$sn*Z_+10f0<!l-DV0X4F3p+wt$@8enyM;rZa;pQjNvE9e^fCMCUxu3t
zuuqFR`DPvan?u5u{Y9aRX7j!Iq(WQ64UKO(b#*i}Tz_O3tL<!V(|k#gn;RqbZq;C5
zU@-P<BZ_9b7$go90?W$a^3c5mixT_kME<*jH)TUN!zhZ(hve9`FF|yxt-AAvQ;_^r
zO!r`^!|2<k_a}V6m&ldR*%2Nh{}Roy)L@__%GSuI=i*9qi#FJ}!qU+<=F!gD^h-UN
z7{2xA<I@DXT3UO-*sWkSEgP{*r%fy^Jty89Q?P5S`3mp-c}mP7smqE~<{2ZI2hwJL
z`qvmn0uv|+(-z-~AQex2B)kNsED=nbZ?Ox{vnta(Krfs`HmF6T915EZO^H|3Jt^{W
zXDv&XYpFOZ_g27yX|P*mpFU0b5Yh73j@A30V>L53r}J2AJFDs^{^6c4iJ%K>#bS!m
z=g_yt0ZKn^$&xJ#w5*nG^eIe6c%Ibm>Y%xzd*P+t`InfY%tqe@3yNb$fFb#It7%Q;
zZtuEEp`CPP%93ca=%v%A$%Zyv<arsqYSIP+q}$tPel`BC{>-AdQrOb7=5|7UjEld*
z?$a_hVR85CrMSj;k#Tz;PEJl?*EN#|58976wo8{d&PQ$oMr^3J2yRAW0%-+z#-fg%
zJa7H&&bsO2+CcSe{hE#WKKTmM`q!6l8vW#P+5Yq0H<(t?b!FuV@Iuq~*3SA>H+Ts(
zEBu{?871<p)6b3|Ex$vIqhL(8>m*F>n3|fpdSxdh__W;Vt-aNhsa?v&&CPV?I6}A-
zS`LIM#ZKnWmU_jW$WXtalPw?^Q?_<vGPy4)dwEPRMDk-%QBg%jEa{tG#?!~b!o&R}
z(IF-R9rXQ&!j$%o5fHMb5WDYmBxGhTJRgTs32EPVp2i*3C<!&s7a$_xeX;H%I}~ek
zJvYnG4Q&=*1QNFD!t(yoET``vucP|V*b-|@Tj8-%B^2L>=!!tCO3!`KF~q#}wt?>v
z9ShUdFLXqF+H!kkykb$;Z9ac`)Ki$1B7j{z$Dk+2kjZPo4Ip`~0CHzQ?rJ2C2@VdP
zr5ESp9+Yw;YE_35sGLa!2#RqU+@8CW3$X#*jP%7_KsbwuStF-z8?Ws~T~UK-3>gzq
zE?Zhb!A4=_{=C~_xzjh<tn!UW%*hKkKE9|Vh8NU1$F6^SGrW_JMoZOXm9hxUkXPy)
z@s^_poD-wezGYDV8aUd$8MR8OxGVIu!Pq%tVQI$D58^>ruU_R7vV|E-N3ivjIGCFw
zHl6p~mvpf=zwuVnoo7d5_9R%6EGzdN$+pKiz1|-X=!#O;z2~)?kmf8`R#LKEQJ;l^
z(VAeqr1Omnl$xTVVg5tc7;VNWk>Iou-d^cL8Dix=N;Pz_QHb4}6t3J|DBa&#@2l`|
zLx)Utq;<i@Vte!;uF!!Ad29}B_7H!G`2)tth;`H2ukty3FQJtM`>7$6sjpyvF1N)K
zyF0|m!jjE43z48=@oiWb@^ne;AmYMt5=n15bMq&fEzszfe*%l?UD}e6KwNs$YwJ7R
zOmpiS_?gWt<)8?l5pZ5=mlp=Gnw8}qSZlKK%q=R~%KZ*rs@+8cQJRoatF_l5mtRJd
zf<vpwZnBxvb2$L^MtGdg@O<skpYJXFi)FR7wIwBdeFfk*{haQ9h=;bDoiq;gunlq*
zIogizlP~y*ie3qqviUX4MeO@sibXDN(CF)<YaSbmqUf*Tp}jwI(v`1<%P=Bbe2C-~
zkMS#g1-vc}edk~5sjCXmPWts5wV5chKwn{$rzCgk<H)PqRuRSKDHx0?pOcIXN%_>t
zu2V+{5)?j_Z^rXo|9<Tkzi;O3OfRppR1&BBVe(5gI+8m4?pxNS2gJ0MfzhKpTH={a
zMS)r`U+IcxW@m?h69zWc8%}K*jsejR1AQMk>APB6tC+r7L{tPu&z4Nj4wEyEcXyVl
zPo%hsR&&c5_kaBDlELGpuyez9zMQDTkESwE3+8&mR-~?R{oWM9Al&NcH|Xem=SzYs
z^PL9AW7|tUP#(IJh0Eez6a5b>*)cnSWZzQ?xNK{T@56$%y=9QuCrYHwS5t8taD##l
zg=ZbTqm8Kf&c|qeh`cCP&3W~eii+Dggmgrm!sUSV+W(kdqo#00gUG8JPG0)#?=nVg
z$d-Sw>Z@#SF))CIcyr<3j9v|m>3nLeeB(0OGB|fG)of^FW(alSA}%B4W%^<vZO|q#
zo-A4|nD$n1+Hxodd2ti|#q3{$=)|ES%Vb4c4%yB9`0U9H6&Y#@#}0{*9M5#)Tv@rX
z>_K)vhQI!Ta1ZI1#uF$s{=Kh;-K@lPJgyH6hTh|_wyb<0gRqYE=kTM?|HV3fCg!lQ
zq+xoQz<mZaMvhBc?N|4gsIs09dftS(%<+ZYXa<l+3aHoSxJkAC*!-SVao%+?V@*rH
zx_;|Zo9o#_B9^%9TZFD9pseW1)}N9!>UkQJ2993yX3V5gFS%4?_yH-`W-X<(Z1NXm
zP_Ikzj}u9H6Y?V8ho(;y7{%i4jj@{VQ22O9<7yWXi3bfI*XM3d)bS(36^2?f-w^K2
zTMNGhzDlpo#t1idmuiZ-x7*kLE61)sj9aBWH9`%U2R_>st0p}%&!5(Ra2R>^9&TPD
zM2bzaZM_#qTMv;fe<yPrQ7&}Sg(BoHexa2t@TIGHv;F848QC&@grI4po1+yn9EYnU
zTNeUL?zQz^NX=Fh-AmLjP3_H?N`SNZi(u$|`=6vV$q>k&taOBMtj9CadGvNeI_85i
zhJ8?xIR(gOZj^>0=IG*w)Qa94`U^YVP_QwaL64}|2g5LqaP)S^v1KBq$DUx#4cfHA
zrTj2`t54i)^R3hoNTU6Kl+~AZpGFD$v$?5isfqRr?IT9Rla5S`8+W)lY)frKP_zY1
z!k4ouwO!uQodk*i<$|l8y`P+@z{-t<Pu4)ucH8>PP0u%0IW5Og>Qw<O0ERjLCWvgO
z#F{P6R$p5aC-9bghMV==u?%<mvY8-YkA5{qM_E!PA%o4&af5+>BZxRle-!JT1pCOr
zQQ+P`opr{N48N|M|Bb9h#o)UQ7ZAp=#^)3r1|h3AYOBL_TVsx+_?U2u>6I+DG??D2
zijYMDmDTysqPeSZMDgwTT994&hADgcZ70jKP`erbq-hA?0A>)ycN6zz4cQtR^FcPd
zMN2X=upL!KB2wHaFMd>_r{K0%#P&24`-krs-&HZI5<wolireJ)bINw1Dw>aJ@x}g-
z%Uic*6f%sb8LX3)_uEV?f^70|JrXXP$gZ?tplKACaW>CGFFx>Rs{dk#0<cec^_i&~
zC7ZxLp-vpbfkKM$KcbQMHyjO7W%_PcIH);>#ZgAIj3Wu?rShCi<oLhhxT|=Ei3Ry|
z!dwPgO0ujnG*I2TN{Mj%<Y?Togw!6@71V#@y&8UotgueqQ6=9>3~>npoM}~kCM>Y<
zV~b4b55p80ujeog%xgv<^De^A`^9iADSZn&dCOU{!bM)kab;ohU&y@iV8(PJ(~Wy&
z#cpI5{f5u!^-a$lx<{2{s9U2i&V%iE(QgfEhN$Vv6mI^TB7*PYT(Qddd1v~`tcS6&
za@y8@cC$R!PeP*uWs37+3<mR$6+em^hXLL&s_SuiZLsfegf%J}M-D4232?5Ayczbz
z6V|m^#j3ytR5jpM^P<Ut9ti8g|0b+aJcP(biTqLeY+#^g@u-dq*hOp03PG0=yGxLK
z;k5wE)^OOU6dXn|D_nj5h2rOR;T{|bz&;B7FJ#Au?QoIlL<quO_tXv!AndhWUG?wy
zRoHP<9?mfotdW*m#&wbf2u$bmn3Rk6i4Y{<Ifa{rp5m!>1fuHoa1izDR-qal?#;NR
zGhY(FSAN^L#AXeH*@`ffkDiDf&;*pJx_~nU@)M&=2~++57eo!q;3xJJ>7^~mAVef5
zdjJGI!*Q^qzZi(t?f1JI8mmTE`zXMUG@sb!4<#YAkjN|!{`h$9y5!8i-n<%enk*~P
z4Tg2KkVVIL;Py5~{n4&2J@MpKZX6N$ICPN$VYgB7?D%2T|7xaXMqmB&1<cHw3+K1?
zaz4RA7=g-%&=gTRe3b8O^46h^eblAkWi;>>t;Y$wDrnRXo^DlOJ{~1;td(Dpb=*T;
zs$AwnMtlBkNhIan#TIz#$Q$4K;tOy6!Tl|5P!f2h^8Vs!E=4J;D<~0F{ECQ&NGms#
zrUq^UuVkpk%N+<!5s?&JjGrP0JNor&c<1l^G*mr~eG@sb8>r7QeikK+BOK?0{8%gF
zfdjgP@8NCB0z@_JRq1m3Aqe&-0{BSefa`^D2TTj`E&oerf3d7+^|k2hPFFZfaJyWJ
zqu<6@$9X#U)F;4p{?->k#h2q|yz(Z;!u)`3wH^8*L{FPzR6gUv8!C?;_vix%Z>knS
zE>qr^t*@`Od28mmMUVsB#clG~1r@Jb8o+>OoW=Lk)?L@zz#m%|a3{1~E$kdW^*%-A
zUO8r~4GrOKb`Z%QL?rMNJLQE<8GL{p%H9}+ewxa>!#uQeQ1Q?36MI&<8~U5hbF?l}
z3~8vPEiC^iK^8s5gLt&Jz@vo&p6xkm8){rbjd5u7ZP*ktj7!$MwJ5u@Z|8hbg@#&)
zj)#7@(>9Kf6p_F|?^s?{0egtpYFN}6vW}Fv?WC-&<svYizmm0L)*>5^v0||i7qosf
zKl4_SZ@-8LEJ@)tZe6@rixvjz8(aJq10-kq@3ie!q7z;hWiDu;P=dG*6f*;o{87`D
z{1;4NK%@X(VUORn=zq$aMEG$V&H)wQzw;&>pAGf}TgHLkvXfw6J9nqb3;!PG>S=I0
z(m(9d86ZtSmnq%CAg)7IIiXSRheGk<nD^I*aRlX~zX-~q=py&$kUx3%<HA9!4CGJ7
z=7;$U{g5WfYB%IhNCr7Zjm*hg?r+!>**x#W=&KbE2X%=O2LXACuV`X~`~FIw7T&$P
zXkUT&AvG=<wBD<Ugr;fjSKuIto^3q8T8KqLAgUM_w&Q)wJ_UDno|hmAz|i?YkEFPK
z2RX!g9BO;NGWt!I5B=JiV1JC2I-XrbU~h~Mw^zP5Na*~VV-hY>AsKf;TIv9<M}k^G
zAt5k+6k;6IT*?sR_I~9&RLNIUg&qm&k}%#?F%Xu)^8Yt|s;45~+A}&YUe{j<^hcB9
zfGSgR*vO%u&y}tU>YYDI`k2ufl?!~wD3+=1>hxyV-2*QZ_Z)fws|cwph$MhEZinH+
zr=nAv-`%7bf3@hg-|`shiecc)wmUYJ<h=aDykaqz?l?+hVU^25vU%&-$2aK2oi$>i
zrOm4vke9!H+kHqJrzn9M&avF9{nm%JPnavA_v>tW9W-O!8LCTte#n;0FIjQo#oa&F
z525b8&ln1tku;zt4Q^ZruKP<rwXgl)!Gp6_F&2-{iz_OISoD2XmU4=CW%Ok0(ii90
zqD;)z{!U7A|E%wKi=sHtUIEXORb0&5NJcDv?R=mpS|Fs4O~R3+y@fCyWpm8KQ~Ju8
z`JIikCy%zD&792{8Zt^ysK0iSwu+12^EvJQ&9h=+VxN`$iM}*};%cW-A04tJAnz1i
zW)NwlsHk}3z08#}CAs#WswC*1$Lg{HfA;y%B=6Z&cQVuV(Cnh3hfqrjwTyE~d)ZM1
z*Rd>FoIQ5r7vbQlfHszNdxz#lAd&#K>O3_cdd22-2Pcpq;vZ$&W^~g9#2;Y?UNv6(
ze^5Jbwa54nDuxO%hD*D5S#?RXG(uwKbX<$=+AY4>RHD3p;w&KtDMrf4U7ciIIcz#^
zLo>*95yo)Fol^%a3Hzx`FX6wnkp4<mUo${Wvhl=vRRYIwlwU{a1!IE0T4CO73NU;_
zc&Q7YR~9kY^n@cIz;_O3yvcvhE8Tw5e8%UMtv$Fo-AJUaixh6`hnl4NR?NRd5Geb*
zI8QHlP+-v%WxCME3bg-Jzk2wLFhU86O#ap(6bhFjByi`0WQp(QKZ+qxt)>@xN<IV&
zA8cjU(Uu<U%mP(v8EJIwJULm*Dy`dGwnblop*br&f%k3)T~PB*d-euxHt1yXoKPaA
zKjeI{)AC+?_EW8N*sZ#(%7dMRd%glK(wv(f*gD~Z<@h_#kcXr{IR@KKdEUV{zzEyG
zqrrq^sTH$AS@`55MvLOap1kKuDA4J@3oOVk+%jn`vKbX}GrxcTeo3v@yt!_pyC%$`
z7rIuWaV{FSxiE<Af@@yd5~s>Q@m0zY?<LDd{Tj+7y>Mf!Lrr%e$FMO`Ea26US>y!p
zf{R6?BxEileoR6;Qy)%p1DhR>lY-rBSFsDlI73BMy0BW}h>-6Jr*E+1dJFDDob2o#
z&|S!cb%hcsD=Ytu;$GZaZP^r9oD{~oyo`eCsI|U-KU;<W9yAM_2d?TrSOg*;flG3K
zm=TV`XVu-70<K%wN`{%-bt=JEbnF?aP`x#1LQy409G@Vl1RZ^ZP?s$S+6al63L-bk
zfAzhh;TIiSY2>tW=>{H2%%CxX{emT=Bq9M-olVfk4udusl#p&Cx9~MSd4x=*4-j-{
zC0*U2l`rx_+dy?G;l2_-O30mANYNzNIk>w~n6?0!2#QZ9nb5N}{tkL&CM~N#;0g49
z(7(RvZNR*WS&*QX^8#WXl6VWcMADr=YC2}o_Y_-xXPNx#Z?*{JIxmm3P_5QdS^bO^
z)D*QWU+=_iY-`9}s|^3Ly=cg%u7HjtR$)8}SUED3cR@p5mDhTj?tFVnFh(YtXIXP-
zr!!X2O~@^6aW{a1JryLGjG-PoYumsi#{8LpnefRVZD6Xbk9}@Xs7LoLlD8v;jY1Ke
z5ui)}CFeqKzWLSg9fUj=g0IeYaeB1h?$WZ`=rNXmcMVupIhOSl;r-9%U6~h%Kt5*T
zg_AiUt>t2%QRhLG5HkJhEDH6${@ZRVaAw>Vy=zO~NmJxa3gJ6zE8Iiwc5r@Z+YS}A
z42^uVwW+rERzPaT4{UpmA)Uc=If$`=_LgNTJd~U$k;1t>Od2rp(D2EsI9SgT#_p|p
z?a%50aa`hsJvam%4=_<US2}J>>d81T#qF=F>6CAe2x+~A7GoKQXA?6LtR)*K)E@%F
zy+n##B2_QD2f>#FILu#!KI~Ltz+vio*^R7_mE1gpvObxj2(6RTWQv~+#(H|I@YG_z
z0GG@1H6*@n3HPup$;!%7aa;Bm+oeppum?xqFqBd)0Iz&4_Rj$w$}3{R13#u%=dln8
z5HL0yTtl#P?BaNL+jeeXIgVzVeC7A~&Lk)mP*r|#AYh<wTYr3W&%k8jM?XKzRpI~6
z<18Y&_oy}YTFtyyCX!V}-%Tbw7{cxkiB(?<tlTZFECsR?dLfUk{gh!kOn;}{s$+SY
z<~t++&QNmf?F5yk?TlnD@CJl<Hll|pAVPHgKW*+pHn(2`3+hPE5o#qvjp5nZSuulO
zAMf`ZLmc%^;CJ6dJL6Ba|1AVW#WTF8T^KC1(k08b;HT0<j1CH!_Nh5xTJuE!22<Tc
z@D^L$WWtJ^93|0-=9s9UYw)T0=KkEq$eZ_i3P6rC5SilHz{-kra=qId^QwdaX7+bC
zy;d5S;+#j04{heYIS!i=^RCoU!_GLhh#svY7Tm49xy_)Bn{J@N4%HTic1hXJOd2rZ
z?yyMglCN;T_4iYg(ROmgOo><^$0Pqx90C+%{5!{|_T<;C>_jA(SSQ07pjXC1<rn%2
z<wZN8jR3;Y2%-nWhfv-WUuCJ#P-%CaBP)~_MkGGpk?*vg0FMTs4i?^@md77l%n>g4
z(=K}1&aqGqMNi6nS~;w~b@4FD`&HPqW*5>0#z<Z}zxS2Y^P1_ekAvlIh1PpWfm=tM
z>A|0~M~Gt&|I>npUM(QuL{pU5Sn1II!sB4Uh#>els)t-7_ePn%3rG>y^d}3Jzr8($
z^4<$m^a-xf2U|tk&Sn%Eu^?yM`CPSvRwo}V81rxPP!fg8Qx`|#8%vV=kg)BTWB~!}
zjDZi7?rtn=3z2JDBdMS@;|6g7%7y*Iy?W@}*{*k}1=Odj=jMcg(_MOusE8V{`J13;
z?0S_9XS|}FTzP00X!ZwMS^{<xKjZesm@;RL7yukZ`VeC1DHLpfxt{xP6KLwA>YvKU
z>mxV;ea=#3Z}!R;J*>c-p>;9SlWR;oO6Rq=e6kQcZOJ8URy^V->7GL<@RrS|9>003
zodh{3swc$MhmPNIZh5*XiaSMKYKQ<y3QU-;zR#m*zV2j631riMk^=8^Kv7CcN@E73
zCCeB#`f%0wy_-`2=3TO2!HwI_h{P0S@BpkeJNr5-FYI)$45Zz~w<mn||GTWa)Cl$h
z<F214l7T!L9R}L++2vAO|2`{k+(U43tFi0<Q`g~jk|=LSTqV2x;}bu129kG!427ac
zWz%`U4awR?+)TB7yk<V9{5sp7ESt+Vn=9SidYDw8HFX(YnPNcAcRC{j3QF{d6HAcF
zz^DT|&UYMtmk?l~jl(N=9a{VLKEf4aCz0e!Wa6iX+lN&nKbo^-S$RKWZL`!gkOC1k
zAAqp=BC*$<yq>T-izA#lNGktb0g9BGYx@Y>Pbs)2mhE+4HUrC2mxBaxX`6t*K%Am8
zRar{3Ki0S}G|i(CVV=BSwU!VJWoQ<jjUYdF&XDsY?6EWm#v~|O9w4OoVg*PiJb4gi
z0!8zYohh4&*LG}sGHyu71(qep{_k3GWMke+zO0e-CF0)kqj}?394oh*tdNp0o}O^X
zi)rP@D#B}B<mtX3+`!EnPKt{EveXP`YO6h!0RrBv<UY5RJ>r-?MvN&^>f)<a?`VpD
zxsAgghGzCr9Ak<SAx2E5Ugdvfm}H2O1tmg>8@*pa=VEI5-$G232q$h#5EM{FSC)Wh
z*U}$zST+}*mOTPTY{9Z*Rk3Y!<yWnF=AhG%hlM5xB*;&6<wtNFd(y*<PE_98i8}i>
z7@<&6`@KrTah^NVDRG`97JW!<?M?AX`Ip8+^M!6kkh3&5JqEE0l45yBS6yBM7gYOX
z@J-`d-JXFj(kv<*dQWZZgIL~}e*yW{?qa3ann^Pnusi4e--S}ekM+%3w-NMueGgtK
z5qREabHK@15F{ha4?#24(R_`B_B4+DlCjy1OGD18_yuAn-bSejLK+a1lpq7WawaW$
zW_<xS%!|`h+5fX<>aXI~o2&K@j-daH5cDJ6V`f=yh-L~^uZSe#@xK44Bu9<mNS)Ry
zBv=S**3dvtr?dtnztSZ*H<1JT1aXQNCmLY?N!&Gf(53(Ar<Sj}BJ!$4ecXE-GFX>L
z1(rnk$HCSJ(f=F;UK8;=hd^ylz4~GOPt9uef6L2J^`yvTTkrH2suEIk`U7iWR^`(C
z9Hdm4fKkT&PO>mcMgVdr*X`oR^ZR9&n~UyO;4M%HVFA4zDZv8s{sY|l=RvoLdhLWW
zZI2g!VfDO@W4PV-^Eg5bwStp_=288d5aV6#r5dHr3O7&!J_EHvE|CR?v+>6zgffEK
zYOx^=t2<m^<Gx7(nv>;I2pAGU{q(n%kF7-DZ!kIUy@-5*N{`(QaM+y+$a$|nt~D?Z
z0L$WL0U>BCWp|0Xs3^<iL^dcr?P|eAZ-l)ds5I+Hg+?L`zhiX6a66YfWF1{d0T%su
zp`YV#vMVr~m^+XogJy}l-};^aZHv@J<g@>%14T_;0W`&~aAMzBE{2DPBk~g7$>_({
z>S`d^f8&2!)6P8bFureeTxG%A2N!)^3^azj($(qFXBlPQ343{Rtw5Gn_AxRl5kCEl
zNVJ;t<=hVz%gRHzPk%O09`+a-DzJ#c2!QTK{Rg<lN^qNk6$p;01<`2FYJb`2iS?Nb
z#NuOQ#Sz5}Z-^vD5;Elo<jDg6ENTY|)NsvhU4dCyON$K+!K_IMEa<g7=t@829@wXs
z14PS~4kbZ|Y!!l4K1h}!3&HMWVa?3UekfgmkJ2INoPRS6ht6z{T^>4Jwus42<=qf`
zyoYq9LGWeKUt|OSgDmu?6eE0T7m=RYB}<o-JEdXZ`K;_XY-na{&uz_mm^}i@Fc(NZ
zK~R++z*(^sX`Xoj8sh54F-c*3<6DSW={q}Iw3{dBD-@^b_`i@`j-#DG7=p-Xd*<xH
zb_*i(!uEf2M&NFn<BNlx!Lq=333_B>!5nyQ(CZ-C&^Ba^?M&IQuY~BOHrLs0g!UWR
zn6*+H94ESmd<0h=1_cGFH$c8&tNVY+>2J1s02ErkydmgE19)^gp-@#^$CCq^rjxoR
z!ErKfL<+;Po;#)vM(KOJiuIsTX@4K*I>Q3e5*bh`g)0qV4>tR(&Xz3_Kqg82Z-L3!
z4bTdX+sQp7M%7b7kv^DfEPv^A(cVg<gYnl34H`a9+D;pDJ*;Y({RL37Tl7)aA!Rfa
zj#(8uGa$;~K0O=?yR{MuDUuiOTJy}BmhDz}^OPniK$g@uhrF;DGJ1{kk0OnHU+~(2
zki*Jr?E{^PvNE-Rvw{;vbjaM=%nM>lq<&`7BJx}uEVn9Kt__7MGYc_=4J#~{=iXNO
z6R3>rsiA9I#<umf_wOBDl8b__YXeS08T?|7VQ=3qA~`BfPem*TPKj;QJy@C5#k#td
z6csr`Sqj-R=@9}*i@HJkaqr&u2zE{IktL8<$i2IEd9?`ugJi^|lUSO7sxxWw40f?(
zixL;LOsu3nLs2;+=kov(ajV|EX(%%zFz`yRMYkb0sKABX^mzLUyf#aR1P``GeM#vK
zR^C>+LwN@!a9Vzp^*_oXo0m?9-PGa^_&)LT{Y^*?AH3XN&+<Z~W$Am`kdE^IZ%{0>
zw+GSd;VYpH1~*LABYqUd-59r8iGDM!K84+_#PWwry||lD6F}(&+7O;UIZF|S9zb+x
z<`~>(Cg*eCT59p!$n6}&0-;i|8neizQ?><4lRwtZpqvEt<=(+&r5EGdEdKAH|6udT
zyI*>M%z`^yN<7YE;-MS&$;k+QINI06u#qY9OG2Q!3$@bDGf~{>MChKaZ@}{<{RWl<
ziaT2OrLRz(f*el_9YV}mwgif>?qa)17VfpLM05gct@l6{n#YDzwTQm)th@jSL0*w=
z!7|*Eeg}K5VRM~^d=yotv2}6$>^E7C+?`@bIC2alNk*v3=FQA-<Zj3-%$$GYkHo|;
zpVBadeHxoYPn8Kxl|&WD$SFS_A!a{LGku!izNFEO`GbQZ_guCkGPX>`EQ?*bH=UY)
zx)?FvjB{%ioHQi6FDu)z0}>o3u`8^svy5r2aXdEqbKTkSlaP?0r>6(MJg$2+Q0awT
zXD3!GK9H9Y{ng%{BJ<uB^e?|NwfV1H`KX?2bo}`7C~m6?MNV~fb+BV`sFbd%Wl~d6
zbOLEqN9)GMz%VX84=TAvP!EEbaKN&#D@|<?P{b}aPDHVlp8yM{>{nkOLeFnMW&ME*
zL^@Qr&3*%ewKiWkM6tF$<T4opA^nzl=N#04fq{Vn4l}=he}1qDqAv5!beqvy!m(l>
zi)6yB!a>(Gxfs5#-)G6m$zN|msu!YUtM$r|xQQtk()y-MS`UFB{UDALs2}<1>FK2l
zC9|NP>{uunSb)$s-&eQ@BB6mIo5!$-Ov}Aq^DNm|fu6}#QzvbXuI<%HgwTLvbl+d^
z0{Zm@$K&xI;V?59Y3WA6H_^zC01zkKgU+K+?zgJDh&&lAD>iP4rR3E4DVVrk<i5SK
zJX%+|-l^G@hh6W|1#zd#NVQ)-JyKr6b=$S+_7s@E%}8|`m0S6sah%7-3?;AY+En0$
zn-eW@ULX9reUhAx=(xpsZ50<<4G9(6956630fO+~x@8abc}PeIS}E74X)?}h@Clby
zf6?*;E5pPp><%NAa(MRl=f{<D3tD7?i^M2zbW&0DS83snU!`VT{G&T>1`kzPpJkHU
zgL%OXVWjMaTp&|DS6q``fQe}`vSt+uc7B@?(YO{h5JDb5%g6=_IPhTz(D9g$A3xg9
z_jb&#``3d+*h)o(Nj{bh4f9EbpGaeZ7Eq0*Zf$Ml<=u*E;({eXN!RlJrZI3pTv1J<
z%YhKgYW&Z+UM0nVPr^n_j?@HPzI@qhw{HlMWZ91YP}B8LDUH|$qc{N0C(9E2a2o_>
z^2Di#^c5&8SoOhLyS%TjZ(1rJRMh51A;KHok2PGglt2G^cl|NUfA<G#7I3Uzr|n5-
z1$gM`>4z7mF1kt#tEe=2<+^h<=NmRs0C}D)`|#nzg|9H>n%de0nCfh21}KMrZ}(Sm
z*_94}s_`7FYF^*G*jfdICO#?0h{wzB?b~+fM2(E*twR-4JK~gtsHCdeAS$q&AX<yg
z^d3<EeV$w=r4w=o|4G-#zsh~bezueS@?||x7J_<_7ld0NmJ_QUw>5n)z{TYV7fu~!
zJ3B54uhi3rGRnF_(fvS4NoaR27gB(8d8oLS&7U`0Fp>raV?dAEN>~H_3O(um0;0*h
zygWo?2U9LHQps{xrDfLjSnmKF5;cHLj^TIeB&n^bSqJ4d7gy)JBd2af86Z;eJmTHU
zBVqwHIiUCL_h{eQTog9?`HqOz`4YHs(p#O2Jl64C7J6WDaqcUAm|K?pTaBER@Y00N
zgZIX<7l0RoYJ;wd+m;@7HUH)X;r+FCAk|N0#;JJP`I69_2~o|^(6-a!PZUf=?9y>#
zPF+O>Q5oCgj55*SBas(ZUwV>LsbJ~Q>0kV)v34LD#0R?JbTe1in~YX{={%j+x8g7Y
zV4}T2&xlCPxFMF(TU^@)d0~tAc>xPZTXBCEKz1?mt&Zzwr6ehI4(!=o@-L6RJ{=n~
zAag;u==WX&5E77mLg#>%96~Zh=88z$#7y6l0=bw2<M#HQ^;uv(^+1=Y7RSj0MhkW+
zHU_?3WYo1fH8lkycuCP{xk<>N(rG~Nh@IZt6*E}qFD{<e7L{Ra2d#Ero@skk#an^t
zQ%l+<_GBSoOG8(_BdT#t?4?a|Nud}%2Px6K`x9#Iuc8H9l(2S(`(&bobwPw{*ccH+
z!>`J3hcLr%!=Fb1*`dToQkCRoCYjUvk(f_kFE=0hzMp@>v6lv=VBcG4<>evtvpiDJ
zDLJ=b>LCC@re79XKl|YU)J2sWz2=0c3@=&`7J)ul2#l&rII{FY&pL6<crL8_=z{^v
zO^C=@E+#+1Sin0wjhIa!<6*VXs}y*98P>9^udhFr%Do<k-IEaIb6Rk(qv_(9OJ$Z?
z7@QH4S|r~Az;gWa?FGPHU_gKy?6G*zMThk{o41X(0jMc;-LEq-)tTi39bX{A85<tX
zfbNRUAS0XSlyGz`IG8+I@Sb8AA$J{SRRe0267IeLJc@>}1av4|M6lM<(n=WW0tB4|
zEZkp);PUg2_1O0DBhMS`$H{{`%*u@d4vodt=qr-HApO$H7Qu>r&@=Ct)?xPKdeYlO
zX09mqO}g7Bt~%X)Bq=Fb<z;bxagW#V4|T)O(`3g_h3W<KrS5-yl5@AU^J#Y9JbZeX
zQ%BAuE}_9-gg>b@&#MveNSsnhW85?NAsfX3t&YXlr-WiQVosT`PHnHHovd|;?ifj!
z?y7qT(9`54D2q>J_d^+7l7r#q$Gdm$bUnAncY3O~(*+d3xHbWe3=9o}KwDiKcwu>Y
zd9&Y6K8njCQ%6DI?S@SUxHSX@mkJF4BvMc`8@JGsHT?Yfdt@*u^W{``_T#F-+90qj
zh=<!iR=RF3#6(8&L83Pq>y`<xZEI^o+d@nSaF%|{Vah(_JQ^f`U9$W(x{Rq`_Y!;h
z?3vgVCMF4-(e0iEVe)?8&_QRiELA|woAB^_leWZgRy9wcUS*@WIv6#`M?8{}lR0!N
z_I5^*)Tb${kd!>P=O^tt-?^4J<sPGajpIwM57{F6b1yUJ33wd}Gn;7`jqSmz=H|dP
zLUKKiwJfP|^GOWh>1l0YF~u#pHa}7Y;lp0k5nppZZ7=z%P5WcIW;+}ouug?HZYK~r
zvrmn)@bU4rpyJ|w9A;)@Gy?cSIBOCbzA)I%ckh_dckbMA8gNj%d-ogk$gV1XdfoTX
z?EyrevhB{b-qhXV+<hziwu9SUY~*l3dm9rR48A2cS{o!b_A5zRv(OSUxi&~ufNo)8
zF9-bEf)c4{`^`(2U$viN<R)*|<zJtfmH0xu4A?*0U%V=e-EP?&bX#oZU;ORK;QRc!
z0JsvAz-znv&_wCPLPR)r3ETF4;liTs{Y8~jbvi1wkJI3vJlyi{E*AD>WL(p=pNzCp
zdP(7xsoTIQ98RP<*&Kshu6WXuLq;!dl0s^0YulNni|u|K4DkmpT3jz22jEy;ed<{`
zG`l|Gnab@d;aI9o6~eCMbHRTaZfSr%A2NhWzzhTsnV2qusQHWDcQBt*G<<p1kA=v{
z$ld}i4k-w%h`#sYNxnF$Zse*GVDI314Fqoou(D#cZ<He*w=Mg5T3FU}vbddI-BYZO
zp$-oXwcEcyH1@#S`b&dB$$>oe_D73qW^_+aPp^X%^Yatut`-d}M6taNh$239s-vRu
z#*G_!gin5;+Oz2k7~-zd7Gk#2R9ak2iaw#mHuCjvyhjcyJUppM^xoy?zP_yfBb&Pc
zdMSxHKa6q-iHK5CQr@;GYVH3t%H`@Eb1b#;ykp*`NX{T2(EmC!ptoZ;jq<LwwRKbE
zZTT3!q@eFas*uYE(+PV*Ize?Cu3%s=CX_*xQB)rKU>=Z@L1Gn*q2{q3hT!qA!hIXE
zE%;^!B)4N@V?V-K10*4(0mo^Byw)P>OH`U}Wkn)$YU-d#+w$sU3*~oVfA6LwNL}c~
z1J7UA7me3Z5WT9J&g`=PS*y%hQ8bTKdlHa)P8?h|Sx{R=MNIL;vf0<KU$e7ao_u%v
z^Nn!+mL^$)VEM;Xvk*t+0@Gcyl|vL7mJe69-Q5FSO(uhbG86ND>}M>lyEck@7gG0H
zX%;IlyNFe*l^Zz@-O(q0lekBf*i<l1m*eT4GL;bHx9VDc#5v~Vp4-Dc`Vz6rEtNiS
z`-@MvK=Fe(B$2-|dnyJYNmfgLs0h6_^NqVL!EjLwu_d@8Hg2o!Mdsf=-CyZd*+1;R
z>0Lj`lpIrbjF>n8gK2$Ljr$3_czXsKtmpY1pO~4=Ku;L(g&p0?wgS3%^Xv0FU7OqP
z=tO4!27{cwdC1s#0EVDkUP)KbL{)wYpXvzAOFNc{z2vz&&rQ}~mtGp*a_Uj;+}s>u
za_YGgJC&NBKY#8GF!5qvE8|3=)^Zsa9n!UI8Opo-Y)Q^DzCy>X6QSypEH)f(7xu-t
zHCQB=QkBhjXOHe&Hmn@NF4xg9F)~tEMiy@T0Mq`pL0JkYdNc3+oi5H6TITJ&YG|jQ
zrlApp5^z@cveWi{Kg&<ZX(>V=xMgbO&+Kf#`@^rQ<UC|+5yqA|j=|+BiV!Qf00G57
zS798Ng~q{<TtfI13Ya2IvimYJWv<$VP=G+K+V_NOezbkNAH`+ILk)f(3;8EJ#9j7!
z?+c3F<Fqjlt5GrOrJe;|#aiw{WXqjaQ*-<e4fS|Y%oHlA%55A9+K`QOJUFkc5+mS}
z!$qteoD%CQ6*a%RH5wcTPn+>gmWxRZk^??rze_-x%DAE!(iJnCi5g1XGR6udOhTMS
z|N5l(K_E1jmT<osTB;0axLQ)_dGQ;xC-bR_8|&+8Y6jo{&^O>(5}oJPpP5?ALimd(
zA@AS6_fHhqTl(Jnl~Tv~PLwQ{el^D2@dPZo7b25GdB+}0H~CI~JNvVvr+!f=wyYlc
zDLYndF&xxy+sGb7#VvRLzRP0yW^FK?6<3s<tu0iuw}b22A;(PH#6s#3G_)8T6x0#n
z(gM+?TShSq@XZTa$#BuJGS+qGGhE5SU>e6Y3ZC>pU-&Eq1()@3)zb>F1A@k#Mv`5p
zSy2_S>b@`8teIIq80k?8T98mdrC4s%6bT6|!lBCxu7i{SZl?T#)pliQ8sB(tuak5R
zMB)8*ErJjcL2Y0TnL8bjUvO_r7yKTuy>%mRrZwT7C3QHq^b-;TblsQ5pvi<rqs@MQ
z3YlcpWQ%~>JOh{^$TN^DED}K%lY$zo;OY!iJH1t7kkS!U49VUR&Tji40oUVt$8GaX
z+l>Hw3Wq%1z@X$}1&RU7p`!Db1j~%PWtw<fm3$m#KM!x#(9nQ?JW4{{4ShvEhZ%q+
z<O*9uLj(E-R4qqOT{M}8YDfSM1nO(JjFxZF=LmPPfQ%A^ng`0D^(OA2BZPz<S-QgD
z^IzhsZif+QV?*8as-U9En=K)BJuY>Jlba8K@A>hwP_;jM_6+vHZEH!duumJ(LZn2U
z>F5PBpLHySvLW{MYlYUyqhH>)OhPd~2$|YtG*AE}zN35FtEs6kU%UvS6V{1JP>h$u
zwnC!pzjlHnts^yqY{a(s;j@g^Z{NPjFg^grsqICWDxhAkR@Tp-Kcg1u{NRsDV62O^
z#Myb2j{?w}m!3~fN(vt|AGYY@BC^<#E1fs(?O%Vy$&wS}oMp@v#j27T783H+EbN79
zv8|z)qPtLBfhEa_ua<?rFJDrubAGz<Q$O3}a2YQzFPQMl-|@^p=s51EJb~ISErTEe
z>oE*rU=T7$Ky6)YD;yPYU6;^M-30mj|002vXjizqtc*4EGH-S$DJkW#m4GL~K}^tM
zL8|~&O~b6$?E%L=3owhC4$ZF_aQ%^kgJW}{5)=r~#7DETN>i=H!mX&=qggm(hRk91
ztuU<f-EDJKclW&xB4S0C*b_T`f>5eKOl@2p+JfS*7~W$}ac2)e<|QeeCrrWoVw6lW
z)DC;)n}el0@r~m%U0Flx0mr)MH(XXH*uh>QF>$FF7h!38FeUa8l1#|K7of;lwB~{G
z^&!*<P*H=;@e8wX#{szB%<Tb>CKZM>l3lZaJOrQ$QY6SO-PhVA)KZ)fQ1ZU}Wq)!P
zo4ncATU1g*Y3xdbGMV}PIW#(Y4^qeilQy)|f`zm57Pu4?r59dv&_df4Oz#w2L4y@Q
zIEAbddM8y7N#WtnP2vksW=8Wln1L%QC@4U%LDOR&F|`3NGY?72&v(~@F5W%?G|}<n
zra(v_H=UuKt{pmf_)~mbw|Q210bIBG`i!XXNzV=N!CSZTVDWGlS_kY@UU=VrpW-Og
zNWJhcV9Sx*Ryt4@eEISP%E%yy#9f)%T_2suAYVsnNTB?nen?f#Fw5<M;{7({KoH8{
ze&)HFzDn`0;=`O;Mezf0RIRmxxj8u}e$%afKO8`|rQ}2A9CO}>^qBat!_@)$rPPag
z%T#^MtZ5S^;CjWS50=%w9xL+<teUMmRwfr%Mbr=49bhHEz8=hL9`*VZ?WIpxD*9)~
z>*^6(6iVt=<<P<U+fG5X0wJ5Bw#z9W4nYF=ItAf)GPV1ru1ba49{5dbKOP21I2Qs|
zH8Rm87bjE>LZJHkdUu@*z%LV2P3C$P_m>FY7Zp{sU6>xRv`cBb4P@=zUsS_&K~zlJ
z$CiMIfJ`<GZdD$lMR|`TY`Jf!NV5|?5QXYf?d?LWrshR$=wL9Hp|V!I=m9IxZzg~E
zKzTe&9$c3_Vz=7%9gSuT%{7-_WM>ZqUJNy*G>oc*1A&r%IPw_;s^W$R?m|OJ(r-n_
z#G_6Cm5x%<${Zdimc>(Q&9ok6cuEa5r810y^aTWxS-j5sAp%g=l*A9fgb5ULa;c>$
z-ZjUZBkRz2sNLM`Fn?sPh*aq(sE5troWLRbOW`VQ4o1u4%AzG(1f1^rTu(0U>w6D!
zIEXmaAMts~iT=J+dQ&exOO@tB0Bn(u;unwMRp*LC%rPp{i!VHer3p8Dy8Jwb4?p#Q
zgFLre6?NkmO<c=O?Gk2m>Oe@$wR+v$h&Hi(%3U}~p{1Q{(lWWZ!=&e>7txMLSwWx#
zZR6JOVG$S0oPFNBzl=7s#@JOUcp0eR-}jNb3T`A~NhF~o5#RcAvYCg1Qz5zsov3=O
zWkLA2F3Y#6MCfySLrn&~E>?xzPoJ1$1Yy(C-nT@b(Ajy<lPgwDr_;v!=$&v`Vw)Ij
zPE|YA@P`i{h=0d61|gNh>hj}P<HR-d6l7lG^>YXJI$E#c-Ys_5${Qvpo=q)eqFelM
z*%ea5jF-7Xr#t>!D}mOZg7VtYEnubkrxJku2@4AYLY$7*w%XuEIfxOVeE!nB0FA3!
z+iNuyW}in^{V-XM+QxA{W(u!y^FYJeHfK6!Cr;@A@WBnt%I%3L#Q%V}dr3l*nk#`q
zJ?bg!%XUoK0{KH4)%@J*YqR_dIQ!B_PF%Re%xuz|X9~W3E(DlkSAp2X1(2%cH0_ut
zj;5W{aNXoG8JpwBtU)i*MPclajwrd<1tIq$%*T`zCI*Ic_kvNVFcoOY%@27pq+UhG
zGB!n*?|e2=%{zpO*J4i2LS%_%+(0gC3BZ;rY;+e`ko2y_IQH3@vv%zdh$#lLA4mO2
z`T0}kgLDy^EwD-AM~ME3C4l~OA+KM*hV1B^Jn{~f)QfBTg$30ay@K+MecCa!m!KV|
zkV!7#z3EzRRlXj9xdn;9`I5t^`1Ix7`Jv40g&ncd$%+s6<m%*VRYCLMA2*1qga98U
zkeWIJNC1@=cC|$~Kr%&cuNN%Cd(X43q$8rd5-t`<!Iu^a=7Tb!nRQ;o${V^hRzjMp
zpOp-O`Q<Bxm$)x)Q|of%&dF1dDTYPwmXv7b-!6f+-lM>ghcAQx58XBw_?;F4;hxn?
zI6`JB@+_vU05P>+r43`Q@l1d$-YYVT$MzuhAAFI8^g||jgy#X6;roQKBI-tpNO`z$
zbUjL2aL4&#4Gs^ky%|L50#JMZMp|zy9L7sIEix%`{unc^(vclkDlRU>q8@$G?-d7?
z<XGq%hn{R_AC~oK%qP{fo1WgBB^8LxphgLej7qy^V<2UKmKO!vL+kZrh&F4BLtdTf
z>OygxmuR5RSHN;%`YZ9|yZklFLgP)UyAW|uk*L5bI`|~CM<nmxCkyHB#uf`Kg2(&$
z`N92USW5F#LgK{4#5E@(l!&^aROQ%~do0WT8yx3#aBokgNE!anmR5q(5g{ffChl^z
zRAnFvc4j0-M@REp4}hXdfdtO8QrFlCF{|X?t6$@qD&7p<yPHdNay<mZOjRo2Fc)qb
znM`{Ca}E_a)wM@e9l-~qbXXpsh#^yCMk8$tC@Jh6VG{2RgxvU?+o;EFlQCPY&y7i4
zyXn+BaQ@zq$X&&$Q}$34@(wPC8RY<m;N#myx!WLTSsW-W#H8z#YvAS~IB~X;Unnkx
zCk8zdjQL5s_N7vv0q5QfuLe<ZVPRL$5GJ{;DCp?o`p(1F-@tnQ&R)C8z$78+u99X<
zdNJXp3f|q%0LH-IT%ZV9T3U+YCMG668`A!ND0}O$s@868d@9yf5to9XlrBX;K%@mJ
z6%eJ7PDPLgr8`U-LAnL$Qc@ZPq@+t)rMsI2-<WH;&pF@ue(xW@eZAMduKiBd%<+tI
zk9*u>JjpUMv!`ImRa+QTJekt>s%62JPjlt&kjPPgMG0w!T%g{P7`);cpjfjVsrh86
z)E;0S0Kjg|t7hWnZrO4PZ#=GCy*gAH-W{jLb#CMU^6R*9J56l^omUGJwwzB6_{LYT
zg2!I9(1&u6g?3Ex9W(?~40ocW*fn{lfxWE14)Q=l`2JAZgVzgl8LXF0ECFHw81Sxx
zfR~vIlQ0{O%$N)HS7=^_l|HmhtXa05ZqGtTgfKh_dSBZV=_-4QR-u#z6R_x2{<Ltr
z-&%qCRDuLkDIgkPbf*9`aTdSJkpZg*77RFHFc_F?unlL|9DW4+8)Rrg&f9r|X21!Q
zqK=P&D5GMwPHG|MyHxZNk@x~4OA}UTTF@N*xG&a<uFP9}n{FjUs@s29xjB|AFYnGa
zfFVd9Wg#(us#4>mTEpQoA#$}m12(;eAfVO&mRkeS3jiIER>s3*X<6A;@On@<&V5pQ
zDs!ygykan3^u*v{a4p5XsHd_T3o1-DwCs6VR{e1Uq~vj<9CI@4+1gC@%{N-a3z?&4
zoM#u(y3}T^r*dG$L$UEF{vYix&)m0GL|KrD9biLIQWu;vTC4*S9{#Q@>2~6S;26jG
zBH%+MUA4j5b-=+@IrRI*<~hay%HtDa@X|4n1AOp!XwiBIsA-WPAFGFdl(b7qymGJL
zhM>M?J7j;aL=#Tb*{SekhH>Q}$7A!&#otg^*_=%9FUPI-qvLFD+5DJU>rPW-x_UJU
z3T9~jmH1#*{{+G?tiWYZ)l(2#XuL6CYgIS(v;L<Sf*kljbt_(D^X0EJT}>tVmf=$0
zsfjtGcQ}%gm0Heo+*B;t2W@JiZ9GeCg29bgmkfUW%A5!g6?*1y>)91~^IM7~_Z*Ib
zL6`23cVypMtMId6Yg+E|5;7S-zwPnBjWyd#x1Q<*6DZ-{HU0;<)K35C{tj3PDjx*#
zBvcLGhlGr8>?kU(?cg9~-vz)t0Mwzd^&v!<G5&u-D2OToO?U`(@!js{!W2vZk9mSd
zxk+F7*0^csr)T?fTQ{rJQ!Vz4@18#!hU|m<TRNP*X39#-QbVyT^<i@Pasv}uy4e~F
zC2A)}ki^%2mTs1-=nC41h*axZ*vw6py(U&g;|!0%72*rH#~d^6SH^VunzWL;=E~Qp
ziImA3;*eSOgfUB&vRZ($H8q+z0~M99b0er27UvA^RgK}g@?zj&2nW5hv%R6{;tXGi
zxz6Kwlk>-Lf#&nB7wr$s-DIR8DYef*D>i7|duL-o8}h691n^Rz)CL4k8US^?N5kdL
z_?V<13y{rwJVF{mN0(01Jnmhg+*#C7QPdC5Dh+FHjFmYWJo{u@*}BVo+uuUk+|3n?
zNI{X9ghU$<I3zTV&MyJV0&e6sY_DF}K#QLEGmoNZ&Q);OX)nVg<+hFf{-tt<3|J+n
zhX<oeiSDIXC)8Ncd8FR9aZN9nzgN3$+{s?R9dz6w+kH4k&)Z`c^e29Pe&uqMgfb#j
z17LoI1m3$oOY*U*omi0!y$+GgR+s+da?dEwdGKJaxHo&-#KBadx*xr3(%vZlAYQ4Y
z1r+U1#f!xPBUEmm3(1vd-5zeo8g=#epFWTv*5cQ$|3Ql3y3*O}>LHYqhDN*xR9tn}
z6@8*^(2|iQj&g;&etY!~iji(de)u^_ya*)&CMG5U0o-_#%RbH)Q@_#zusr9h)4b1<
zrGqu$ELfx0i-75H!B`Y*yYXv$d3LE~1X$G925)o{cq-u&o4_YP$xJ9|Nky0r7AuOF
zu{Xh4<eQwvgPu$!dv!2k_ZGdt5AXgaLrd9i`xv;HG^4l(O*C77eeL++xD5;MSFa9X
zRDs$CLei{r9Hl+KJo|uGJK;aWA$D4!@lB_p;8hw`I>*+kj}RtLKXV}31!7;T#KshJ
zUtHV{6o@ilHERiJX;~UYIWn<=H2%pzDTBtUwHF{To$uJ<?b>!3vO;L`?=sRA;=Dsl
z%z;OvM<6^wxg=dFGfg2iuxG}3rHKbket>VE`v9f(o)o#uf`Wo9EK+hokV`A1DcraH
z3cl-NcaXCmSuB%Qw#TMXkM${vBt9xMbRi8%$gD~$s4Rf61xMRw!j0i|+5KX(3nEEA
z$Mp=4JYY${^Fz34ABLqy_f~KC`c7zL&BR%Z_Mn0O@a=z*E<1|$BjCN1gO$1ho`>6V
z{8i}=#5t&43be``fZl>s*IGk^YNmw<X`q<N_FP{x7Df4}xPWwJ?JlY?b464Ic6d2i
z42WZ(9nV&hA41}Eb3iVXH+5^SfPVcP>gi3m3qaK+J3bYCXCK51qq&J4Q%=Bwp=Rdx
zdyl@X&|qwOC=wPe;^jVs?AKlPhobOYwdYwL^8zSYg2Mt5YX<#fmEvhe{fld^9gYB6
zHH%M#dP4S~gK-m`1%0|>BTFGPkKAgx2RP&=z*mqFQ1e-dqw#>^`@+U`b=|?G_ckrg
z_l_!09KF8)AQd&(mGLHTA0J5>8ORhy*X}Msl?W7B(OkwM9-W!$woo0wUzx(C!nQtH
zkfn_2L9>9S1z%1Hst{EIYXLgzePH05ljpJfw@6)q(}1_PHq(P5xu?1IyPHcW#RM!F
z^56&(XU8ZQ>)NI<+yWCLNa-fZv*!Ez3HgKV!5gmy+1V|iqOa|Wvxalzeb@x70%W8W
zRKc+A0R5rFQVLE4jB0xagur2pTK-SAu)T%%z>_xEe+n&uL1SoAI3|Z2EFIt5Sr`VI
z7j5?$1+?&^uI@&-Bv8BSKm@86n4C}1f+vuE;MWMic5Cty29w*%XYi_U%amYjscwhu
z2oy7_pPRl6h9z<T{yz#}SfundV^+*toleVNJUnW2sM*^^P(=IvA4#|2GDlmGazNET
zG&oohpaZx+K_Fw!dF#QO2B3hi5!;jywS)M#=5BrQtjf(Er|pj@tY)_gN;R}3JVC?|
zvS3kkRVHQt042y$yz2m7ug?!gzI(UWF2QO#TwVqoiCN(eP{Af`@ikD3qwsif0Aa9r
zu$}k|h5)Bdh=}C(=k5^8U8ttOXJ`umc7bvQGzP=d&~FR&jnd$-IM-$2>CT=#3pF#7
z&nkiM-=i#3#qJnu@d@~72I1}3Uol4*7l&O%v*DEwzxULv@l=7Zt=llv58@eMXjE^B
ztvM{CZ2Qiqxl-ZmdZ3#F3BIzTH4vM4e<L0)-vVx5re2&JTq>tyrL1<a5&m6Yj2A-v
zp3^>V0|@HJ1_qP9eR)P**Vx#WpcU?IjZPctYD8<D)~y!fc^k$gF-ywFS^zr#^~sO~
z@W}12_nQM-NfOy>QV^=`QUur^E-Sj9DgUn)0EzPmli8hDDVMFU$6#DVbZz!O7^#He
z^O)UG0N=5zG3lVUz~3!Hn5)pVO1b#b7s#^ni*xr+cVSFW#sb-Y$vX%6Rx==M=P~)3
z2{MfKT10dC%dQtuG0+pmKy$d{$R5B@R9OW#q1kE%=y>GW&EuRJB#1EEr>k?%|GUH?
zh8hmt!><Q-p-B&{rVz?^Gqe$~{GhB7Ke`1{n+b~|jqA#fjlj7WsJQO*mXDuU+7hW&
z7y907`^hL5zkGy;!tRhD49OWn=7r{)Y6^-X{JxRJ#T(~xd(rd+G4nv+u)oQ_jH{_b
z`pnmxY*T}3WZ(zHF~1HVV?Jp|aG!<>M(F2aCJMN-nn@anuO9&QCIj!WMR;Bj#h2Vp
zraSf}s0(=;<rnRf2_ol<v0>MbAq+gbE<44F5J7~CtY?3+nNBUi^gLV>I&J*|)XEeT
zf(FgZL5QfKF#!IINSS@_D9}~1f&jI;8Z#VS&Z2p*;{rXT;U`BG6%!<mxdC}m2$VV?
z?27$8NAN>Xt;>hmfT(2aHM!P};=CMB^r3c0vul@XurqsCFqrrCUHNZInxgTUj=#JH
zuRrarHUnk(>XwqrE<?Cv=bhnOjzd$d!H@TV+=N2jq8&epz-lYNX{eq?gokf+tD7@=
zu>-(|iHa1OkIst#;}Z1Tt9C@Cyv}0uorg#LsME#fj?zTbY8nz`ST%MZw!3eo8tU`I
zERwY8>&eE;o*lMyb3)@bl$x6EP@im^va*?MeGHBbBoL$+?;3W{cEZO_P1^<+MfvYV
z^t>9$O6|(Gequ0kO;7b(%7>m7qN#cu|F+=emF!yS(M?>BP~LFx`)9wzV@QH^fe;`K
z0M=BUc|Z5217HTxD{O3SVL1()2l<h>AM$_UTl8s^aa*^q>re5--*oDPbj;G=X7L6t
zD~gx=`Z;}pZF~M0ASP?Sf7i0)XJ8N)0m>aC8vOiqub3GuV-N%OQ0-17Gf_^X9SJ=`
zbQ`b^yxs;{^V%&Pd^pXEk$XAY{K~zY;1y(#-~#9PBi>uCa8HyD4d(ziz_FrXr0Bgx
zQc}Xi%-JrFEg%BuIVR;-Xi|UJ*j+hHXcIESmKcE@G5=HN*+uRNAjly%#><0CTl^h~
z-q`p)YFJ-RTt=v^$C*Uf?|s|t<iusa|I={ZWphKwq##KSSMn{3T84!eEt<o?Men+Y
zG6H&5EPAIOLa&0JFCz=rX0A-u$BlQg6P3SFsN<z(W3vuXme%f<JIMCXoDxd<T*dM1
zDUU8MLgvaC!47QP$(14wznT-I`UM^-zsVqxUQ+f!({Ewb?!-GkX`2m&CYD?rev^nz
zrK>RC#qMw3O80A0a{J@e2z8}ee@Hntmq7Rd6#Y)yJz6+@PU@orE|6G>D!|I$22#+1
z8K>}^4@f~dUXVNG-}!?7N?$e~{9X#)`!yp*wk621qU(1rh_c|k9NnLyh%O=xjW!7>
zDQXza7f2@|@)0^Ap~&Fv<;7l{!k#X|>;k~VWqhQTs*(Jf8Pwe`r|WJ|#=m)kcNJJH
zbewXf;pTpNQgWpja0@AM<=m7w?g2_q>&e#NqxJ8>W+U02h(X-;%}VhZxnKtm`-s=6
z^7HCLma?6+t~wWkL#Ub|{-nPTg<_&b49V*-$yjcWYnO~0C$z=H#2^d+Xaqb_2W7gz
zA@2j|TAx;ZxX!D@tW>InS?bD~p}HYyeG?CX_dj@)Y8G~9I9OBmzqBAxBTZCJT-@jM
zWvoBZI9x{`>jX30+9A({MBC7@A|WL>#}Z`c8dW*wq0V#HmY@wlxiBWT1?f^dqr3s_
zOCsB%3bW*LOJDj|9Q|(9eXQo2y_D3>m&Ru~E{Qo0WI)SAW3GGeZf$>Gf-vgZ<bMD^
z>5Nm8Q^YI*K{?%}%(c>W1Yww({+VHIqC|AU!^=xPh`+ld^2XzDAP<d>eG0BxLR$J<
z0AQ@KUQqz+pv^fQ=kWY+1#tg~!KVG=aZ<yEg#w=op~51ON?i?J_NL~)P=Cv09VBwY
z2EV99-=Z!o*d)^RSl{wXoBodY;B*PwdSegp*`r=?Sw$|{rsa+!?xLt4;{IEdFa_E&
zy&!9+7!;epO5H<`U300xMRIiyRu~Y{@34(TFmgoiB>L_hHgSH|>CBF{C>z;47(G~6
z$bs^UZaQZ^A+BJI0DSj;E1m}CmRP|bakCIz+L}!dRXugZ8~uT^=3r(Yo;DO-`bXFe
zvarT?m+pdIF);Rf2J~Jt9I6?4o}pqQ=pc^fwAOP|MTxKRVU?7r?7$xAdLa~QOG`@&
zrvaPffI_ZSj+NB2SnSi>^n0{22`YHd!(aw*5(W)^Ox%FtH>&LeiS5<(Ux$%Het)d^
zLY$upI?IbFcz2Nv536Up<1c{nsZg4a^r)__WmPX=KD`G0Xxb0D^=khd9`7WmF^5RD
z;-~Sp@5V+dK%SVMW>nY@N;inI0Itqe=Dd?JD^zD}$n+-@{j9=_1tE&Fpmv+JLP2=b
zQ&G86t!%WWdG&q03AeXL%f!kKPAv4Hnrdcu`v%EYyXZdszD&Eks0h_5y`KpWq8Sp5
z;3R$fg<9F6YZ}Mz?MVM7>Z~%#_wq<fb4n(SSdWH{a(u3T%)9oF&?!a5{9cncR?hkV
zQr65rCeO$^S@~A#>**B@nqfSBd``-OJM{1npSIth-KU=l6%qM|_@S6j$TSz8QLYp(
zK3LlH6ht>=pzMJPN_zo|;}*q54vt*$TDA-&V>>(Oh##DX@vP*U_w>|(I;huuAS1(5
z{I31lP4piXk(}IIRrcIgsJ#Kw03lx6>rrpUH7DQ{P_QKjjOjM<`mH<Ft*#=!fB)|L
zr7tFSgyZ({S`%)RbiH*;{nn2EofOCq;~88kVi$t$<o3@&)Hr|s{EtJ7R3PL7y%lE3
zj$e%UHj|e;unu>NkDN?4{>lp({G&<)V-NOm-xhImyAdluonw#{GK3@hmU~oS)<44t
zkGZ~gl<4fj{6))t(?a0gYHDiW--kb51-zjrxKuK-V*lp{luJZ(*W&{Ayn^xIG)|<o
z<Twt}hO4N8@J}_QblfnJ7g#H(My6?%ZTl)IMY8L8HPb+V2p+4uwByx6maUAkthfG)
zFBPUy;XC!FmGYP4>STap|HKLov>Bj#8<FZe3!uq92V^?h|H^cdau*)`t;M*V5ixfo
z0XZGiiQJZdMCA4Fbv?lQD*A?A{W!IQ1;-#!VUPn%9+<m-35g5J*(>PSn^TTY!~V&n
zh0?rC{v^%;{Ts*4*N2cBfol_U>%`i|Q-d8J!_(DE!<z8WS@z77=4-)Nb`B080tG<Z
z^JnT5B6|%ABL%L7^9?^8JJdA?9p8hWK+Had9YlMFK@YS9!WnWIc;SK4iZ9FexeYpO
z`vh7{O``@KJX6EUgTZk;NI)b%fV>9;9z)R3f;tIGE<8~m7k3c{HN6J{io|`Nn%7rO
z&Bg5ao#LgwbslYK6s}L^socTsK^_eyb#5WwqbDiifbJDeo!IEF&@vrdiL_}M{5>DF
zZcB;(`_wp|%504yd5Hz6Dnq8!0xEla8yCpx^LkA{fd**+X!PM-MhiIF+_`fH%_M`g
z1Y>cHW8Jv>5YGZF7jX04wiKFZbMX|MuO)&@LgWIMQ62;}5T<#EgEVk?tPyVI8RQt=
zuoB?_4(jBP^FWwOuh0SBFVRk5o%Ni$9}VnD5U#OdyTNI0h&R5(H$!7DpxVxWg?z5P
zy~G2Qa*85gOBsWTq+=k&$k49Xi!?6>oj*-9`eqnwMr@eXgCd8bNvhF019BXULFoCP
zS02FMjna7f`jQ0zxjR7vHES9ErB?;r(I`Qk81Tp92&>kkJ)DO}%F|qqtn9+xOu|=g
zfJ`0I-X5w{<?I-|if#HOWLM+?kZGfYzw9(aud1Ku3Y2g>JdE=BcKAnq4Rr{2?>~eL
zf4qMd!wX#Yd;mDTzPWWE$4dt12VR0C5Nol$YKXsWa=Oa`o4LH>EBc}$#5u4A)<k*m
zl?FF?#v2D*fZ_nXg|b$guF=33xbr<oxGW>*pru0u%I#_BH)hmn5GUDlikbRz7$Zl0
zO26W3c|<<SP$6-7K(_6y&Cp|kcLeS890b5@>fMgmy(?Id`10G%M|ng62N|rag$aN>
z2>(!OVb?z%ab5SEY8OSFP&5hgMpud+;i<ylsX%GuPIS~u+`?iRR33RR{RP&;dQE^6
zx3#ibWhopuLrtCOTe%OdMO~0U#wnoQrSzF!U^pYz$gXDg_GPC4$#1a4T1Cy>$OGVP
z<sguTnBx_1a1Y)STF1dW=Zl7wWnH9!1&yaau&+}K@VdX@ahqx!zL+b=I|C`P7l!bu
zrd3xfUUeBc!_4Uz`|)sqmIE(a_*Wz&xlKW#;s_-_5aHPXa}Qz#l*R~pHa>EXx{;jm
z|3PZPs;AwMSSALB$>|c{w4i?S5YeWEi>scZ{gkZ|zgY@JSf$G)!;hXHqC``C;q#a}
z6i>8d)H}{;XP%Cg5pl|&;oCVGv@h@(#At{B3U`oCv;BJ7B@cF8K$2$s%ZoNnZPN>k
z^w`H2xMDQ(w0Q!V8l=+Gm$z2onMqRwpi+^$ms7Pj$oD@m65gpC%y=a937FHw!O6CS
zx95IY`j4Ik#L+;}U(j?rvb@90Q75vbc<szXq~Io8qmql_(gB1t#&m_`wz^BJnWI{s
zjrG*k{lY}3HX;Yp9~}_n;!-TZV4myU8-|E=^a~a|M_k%%%8Ji(14)A`c}6qE*-U^w
zk+?}x=1QmpZ$(^&mEq2G1d)9JH3ReVy!kyvLap}RIx*0O3~G`ejv}P*TJ*J7UUTWi
zdU(W-4h;4>JKjn#M*Zu~@pHdo9p`zYaX?#$PB0<J$&dKFl=*0&D4wM{I*@6`gL0Z~
z79SoR*c%Jg?(U_9r+mQrpKwnQ`Qa0b(E4Ie(^6OWtYlD`_McOyn&F=^N=nN9_`^cW
zNF3=yjq9~JEkDpD`#&3#kT@#FpGsvDE9}nOR`2VtU_p6}Mjhz<h?3S<QtCzx6)N+9
zJw4H$$Fpb7KzoeJ6?3SFSCb}p;9+|Dx%AwYyJ(){rlSNHuMFe>kSmsGWC3%EwzG7m
z$Uztc*(MfxBZ5H|j5;xx_0v>ee;$Ox<<2_?<v5Ty6-L!iYN!Ip#FApVj4ZE$UopcA
zH;^$XSMHBTnip$iq4O?K9aIpd$vz04T5RSAdZ(<)VKzESrI2#a;p7+f|58%7av+83
z%GAfN;;{fY2Xymzi9`CGr@G41o~h1*QLS((L;Yb*eUQy17^T{g$O#$d-6#^`m7WY!
z!ViOnoPV`M@m|J)>hz0!E>O8%b?{!$GA#plQm5rt2|>l{`d%Jtm-tKePrD{I5EiG)
z6t#P{?1vW~z_9~G4_ei$Snz@P+ay^De|X7z+;D1;R*P-t13pPao&brB7mT(**9}mK
zJVur<>Kh1u&O$aM?iEMu&2bfeZP0)G$)*YySbXeiCrZGwDBA`d)E;%fV?dKvDsis%
z69_fn;-FXTY(T)}d()`j?qt4qiZ8?|Z4~s%*=oMC?w3115!4=T^Txm6hg-1c0S=0D
z_ob!FSDJ4@((eRi7RYk7<$_?UJd_8xY=1+W=NEv?lh-K_*;mYO*ZY;PD>^}*rI<Yh
zic><Kg&^2%qEZj=LoAmsKY=QxONA>m8}+bqfqE1A{+FPwZ4%UBvQ`=QOTc@OR%nYj
zS;t2^)5x20<DaRL%ptke8wgKyAUFi~GiT1EK{#^MkP8C0aq0r60eCei+e23r>exsz
zGv>GEjFk+g2Ju|gY5b}c`KR(&X%ggeKs1vDfJ6{dk0W>?V{GdwB=N=E1odsoZqGC?
zkL3Ft%7u1pk@)z5>|X}opbR}(z*LrLW!Ie9BtMEAiuyOxDw8s|7NDR{LVmQsnihU^
zc$U5oEFB0`6v2SjqQ-<<JyQv@w1az3FCz${FwWrsawstEi1WUwnZrh2>0oarIsOiH
zkN?Qd1BoRUG+=?x1@W!3kx|<4LXRKxlz|rp$s6!RM?YRhgey~wTHja~PzxENN!K%L
z7g^GppYh<yo?Sk0Y%#lZxdHb6h>}i5<Np6g;qGA7f^CMn-j4lS0wns&sUbW7p+M4L
zoG=5JbqF;4P?Z8t7rv9yYa$3eV!)J$qJyt~7MjPR3+2GCm0o4m+}SNDyiRN;qGQht
z@7pSA0l7l+B(<VqRaF(}V{PGOXboQko_(r4Q7VEHUI!|7K~KU*VAKDLM@ueQqsA{9
zH=!h+N|3&aitGV-eed4A;S}Gdr<4K?c^<QW>uv(1{{5d#0>%g|dRKyf@?RF+&8e-V
zqaTc2<RcKPJm%d{^TG52WsF|&brQ56dXtAqqZE~i&6?!mcTP6c%_5_rLm0crrmNVg
zsVN?FG+Flc`1a>J^62!MMH0LIcJQhhigEC)JA)>yw6HJ=OfS5srpW)Um&sZ|jeB@?
zBRE$drNX0eZ@1$q@l-f+Tnx&g;^L<RM7_^}2xeN$6m-#VQB`Xlyw(&H%G?fbXFbds
zk)9ND7Sr|n**f)Xx@te3OO<`}|F1?CCoD-W|B}3!J5l=#G#Ob6at<gbfCLS~VcZS4
zcMv8irYnKYu6||>_3NlY5ZW%+Wx-t-ccx%K2oe&K#Ly7MYk5sE?PJL!MA#*~CFYdY
zO?9+7nJE8i;kVEZHXWa!BggNfSO#tIrl<FZ-ZAyo`OT$~z9n>+2$W>7?cnpE3}H3h
zc^@i6Zcgm6f>SCv!UO`UcXy)VzQ{+yy)CYqT~1Dgw*w9D8|cm}(4YzmfSY!UzaW0h
zz$|~Dtf7_y`tr|v{7@8GSxWF5^CpE#;BJ5-xOw8^*$mf1+pN(CT|9X97Z;GYlpqH|
zv`-_l$3m)(&N78C2tt)Ft>?%5EpaNmKm|d{4xWe1>J794<h$nIL51@`7tM0WzIcI8
z>yQJopO6ZRu?NvOIG+R>ug$9en+sCU{=igTVf0&s1;s;hpJ;8Msfussp`s%Cm&>|1
zy|=g$Yc$I;ZLRA{&o;UWoHG(cy<0%!mQz_^YT25!kk=pcb3DteteVu#Sq#ik)`FD1
zvVU$Jd>)iw&`w=<mSc#lY(dlTQPbX!gX?()bJOioKX>3vaX%PU+-M<!LP&zBI%-j*
z1cg&jqQbA}uAu*jh>DWQc_MC<@?CMh!#{1zI_8UDM}euIo@fnF_9~gli!orQTcU44
z8NC>qpz|ICUrS3f>&+QMo1dZb4O$GSk*$H}JSX=7aWmcmsf6-APZ{V;C+3^+K|d$0
zQdwLr6P;Vt9L6>R-7iqk+=ZC{YUxTWXdeSwf_$Q*V^%U(!1yrf3aIaH=R+ZRWTm%4
z!)y?jjz>uybjeTX@t#B&b8Yv`J&+23%)}mYD3DlZ`Bp(2w~Pq%78F8fW;xn38_Dgw
zw>9Ak|4*gj33P~>lVI(%9u#!(S=unszvVu}O9|;_P$EqP#tDYvj023Zogd%@?K@z9
zZ*kD`2QfZG?+u(Sv=nl)vB{xzOn_U=+Pb{&n6WzSytE#cg~fw}B{_Pqv|bh}&Co9q
z5fxR6He7>v8_0aJ0C-E~P(MX=`Bqj|k4lK#mc^bIQF%{9`lFd%WVFE9(!xA^o`aCS
zjqjuT0>P<ocMH@m(2?A%%kEhRfy6827Q#S+qAOkZt<HZiax6;vv2Xv?y&^Rq-|bXR
z)IyID$e_w~>jQ!JZCls?!Ua@D4630oLz!<x3(;fKy2fWYDuJ&gEV0c)f|aq*^4T~C
zn)~LaIZ$GOB5o4&ftBce3g3r5PdoQ1dFb=Bxo7iD*tMyT;F5L70Mh}p4AmkQXexz&
zr==MkWl~5vO-|keZQ=0QA<Gk9JB*BZG0xpt{Im7o>6^d8Ih<k|!Bu!g0{Q~q$_i(I
zZ~%>PO-*y4ABE0MXd%&t4uuylURaU;gN%hTu4yd(ISjNKMh<ju8?=A26$pi;M{akH
zY{DMBMn%q0DJVL68H<G)4YVaxK&QBRW;ddE41T`#BlTBetS1L}hTmk-L~pjLm<YGv
zBZuc7AF~uL{QP`*<B_heu9K4!)G}^aPwRdr{hek9vdl0UDobYVE4;jTkTqzYa^c)l
zJPN!a`E}U_no}&y(P4s6YXfcv)txjwwTwGK<Yv)iF&gJwqf;{jCA$0z%`SHp-d5r5
z<<ZAGr^1fnQHLu*L*=|wL%y<K4WI4vGl7G`TaS6ebhPso&_bJXx*rIRFI4lzfh!o(
zyhP+AE!_-$k4PN`q>KC~u;jJ`X@xmhFgA?{xrpr^w&|<DJ%GYo>Dm*GC7c3>u{WGf
z_+#9GJ*ZYP#=~IIg1N|-+@`JVB?RR2_%+Ze$S&;>8zfHT)4!?$paimx{sB!dswzN&
z7!KJXHzH6JlgYBw@C_RiQdq(#a1pQeg51Us623<aFh_o~Z+pe&{~~IT2Unp^L{Yn7
z>r~d4Gr=<#Q&J_sJ9p~>%E}-Qn9+mP6P8Xz137>7=Vil#+ouTLpdRI;i5Rr}ao#l{
z32#xcc@Iv2zo|DFC>Vt2uG>+!KS_22eq*tbu0mL>4tGiM#yY1Nfy`r=g1eH<APwcP
z>(x(sNG~eF1=7^BrEF3J2cL*QZsp}47(J%KKWw%4za^(}Ru?OYgxTcS+h3Bm@Muax
zd^WSadUCnzZ*IWAg`S>Daw?Pxbl+&u{TsivItb1p;`FZradqLqEwF}X>pu1JxNs(7
z(A=Zq0YTy|%!uB8yq=(AA1wAiR&ZR44NW3yj^Saj@DzF+t1D<f@Y&=;;K&)^9mi0a
z_NU{7#VL>0B}HqTc2l18Vrxc#m(=Zo<DKth3BNalm*<9k%$qk{ZUHiEFOO`8rrGkB
z)qm^D3^j27r|r5&%ht1%JX+|n<$c!C&PN^)RN6?K)n#~mH`Vp|oNRSEmkJNN87`}u
z4P};HP~k!J+?4tUasQuT?&nXk9%Zh-Fk0_6<FUo-p~ld*jIQ^(I)lai-jbD@j17(p
zI~D0^8{sk0k7;`k-D{1D7xPUy@bcENV|QfzdG2e`^Lo~2Zm_l0){c6zHL}t8)}O6O
zXiNBV<Jb$gopXMtWi^g5Jo-uc#(F(*BdvI!A=-zd>PE*PN0-2geZ^)%$#z=SFlUcK
zk8M_3v6z1hFtPwg*0QKi+ddy$<?b)BU4Wi+Xa$p3R#t{KZ`j)m8pfBGm%B|vm{pfM
zAIey(tN=Jeg}GU~RtUSV{s`YfAJ1&<((*%Ws1&)BC4NP$o)4$_m7bOcRpKm2wmigL
zT=vig8(=(x4v&Doo-h{xh~TY)ezOxtUK*<Mlr^CONPsy#H%?+0Sy(btQ-LgNMF;&r
zR02M-kP`-z$f^m_BH>lJ^QH^i`RgN_)%OEZL%rmiLk>ZbU4Es5XcsE}g%El`kTwh*
za==gnT>1jtV-UxonBl)SN-uQ3vDnZ=Nn0B_`Zg%vJcnQj{V1POQtEE_00<O-Srw@E
z8R|^1<(^St9ue9Y1pE0ca_c!3b&$QYW^UZ+PLJ8(>1)x^k0y`am4fjI2-41!&@wjw
zIXmD!3rkB><bqC~KF_XOht8as*;Y3%vxerh&K@cAT<C0LIp-Z58anK>(gf42u<<k7
zpm0u-i8WL0hGt;(hFqduPy{;9F`U|VpVf&dS60Gtl*Kx0t-fGH!3FYBQ|#y_Fpu7K
zklv`pCI~{`-tw|VV-CQcciY8NrqItLjVW6EcA{T{!gaoAqM-B|_OnVxQc_Y}T;;@y
z^$gL0ch65&6*?3-HljnGMOn9%td>cWQ9ed^IH>J_3zQInn4C&3R5fG$tSyC_2n@=G
zR|>T@<BlJ5Hr^i2k4rsZCc@904A{8JT3*QQNAk(43ZrnxT6*-9<-9E1Q!~=2aLi4R
zDIwck=-DqfkDtjIg-=;iVK}ZcoU&xZ?-n03TFt|<!}e%9sm^;&w{L%i`qmJj8fbEb
z?WTOFcr8H64Ak%RBK4gQtpG>5U10WKYC8Qys2OjtFk@E&<#d0LypvmMYk6rvN!CpG
z3R{}pz{+lJAG&99#JLiaH`Wc+l8o~doKlTZBsXUAG=>2fN_JQkgfZ*wU)S$swFvoh
zi;m1j$f`&qM2}ZfLaE`^%e&Wain@lhIqr{-KgAfj7H$m;Xkh+%eJ3?La_A{_U#a~H
z+^EPXI&m0=o4df9q2xA!PCI*hdmvdsn6nRUoakVw^sM4ETqH~Z7=fvwW;<i(kRBHo
z7kYIW^6Z|*!8Lj4+WKYnk1FfwV}}kNNKQYE33zsMZ}rKzzN3hU)u6z>cInyjaVx1>
z{b7TuscSh}T3Rt3bh?F6QB=tibq*0pOVNSxL%dGY{roikwIk1UJ9d{2<5a;&?D(%^
zT`DS?EOw3wUhqvSH*<*0$M}kc!rU0AEsG9LbTn41MFZoy#jVPT@$n(64k-}Jg5ag8
zsR?CVpf7tL=1E0CPaa5&rbh1({bEi}W!syK%_IuO!k$e#+Y1IyWGwJm<Ery{>R&Iv
zKGM|Gyh&x2J3zBE??khCJqzAeCU7|OP=C)%OM`an6Xe&UptHA80z}&aBR(7r4Gn6|
z^)2=Fs$ZTxHZ?^D@i1Mx2HT*mb1x{#%d>+RU|pR}isOtY)y!S-=p@7SVI|_KPy1f(
zSH29IB`PqVvn+mRGG<XU`+gOQdyQpRs5&}RVm4WMouD8(l0QCtnu+p~z?N~?b*=24
z8!!a^x7DTIhLNS8`a`fI;t|)tgrH@Ivd=0$#-{_H#)>>qRhp=xn0%ErV^&ylk0f@x
z;3V(O(O2xN>hOS2=7)iMFlefexiFOU+&TP9?Y(f}0y=mKl5Ed9p7JGM*R4=j@MNZg
zkD${Rorb8r0u(u#4gj0I&Pq%xSn8Zc9sp-rS4<4(pDWmCTe2&?$J1GO_~mV>j!on}
zY(O}R(0YnQ2|QI?vv&8YNy`!wVU88zG%n|T^iH>aKcjZUvqmRcd-<jG=)^?-0T_@M
zz8g-Z*;A%ve&x(>l8F1zpZ<WcgL6yzeniUz49$U$x=VMR_j+if;7yoYFTHj*k@Nr!
zPD$*j-OhFvJQ8=Kv9Ylz?D=<d&AOvsQUprO%TOZ7D=se1$e2a9Z@f=$JyP8Vtbd>S
zXJ+q6OC($@J@u3Ve=0gN_4w$k2@U;I3<C3N;H89%byqnp^z`(w%kbs%Rm<&Ho?@n+
zoh~VvNh(xjAl(KXv6iN$08l#MqqD14X^y{!rn7wLE?#%t087a!6m>Yt?yF>dWpUQk
zHDG9E`R<B2Co~Q~kDMlai59*YW=t>i|INIAN_pV|#4h;w*T4LFL2$F$u4Oqn7SLG2
zGidc*V8hW77D9Zy4m5PqIxXIg3>|WqVZp$-91VTi<8E5Xw0!SYX3wS--52?>`=T|U
zC+xK-!FjpRClKvI<0E87%IfMnl|&>zZL_?@6413Xva|_;4fU40&=U{=0c38VC>MRa
z-t_G9Ygo{S-0;5*bIVaZoW}sfoOPt!Lj)TDZRVhI&?Y|e5=4nJZ8BuSg_Ao%E3Ec+
zzFoqwc}v_`G2>hbFN1z1<dp1r7b1U1qTqHj*VfkDh2t_ZOkh;SXB89ZNJht%%s5s6
zBX#SX_aIDIh8vllFP<`bO7R_r_V5+uwgGEO&Z4GU=mZnvV_f&_F|dQAOT|9a3qte`
zzx)b?vn?Nv=E?1ehp;<t!v0IM!u}Wrt}~v`DxyC2@_bIu&bEcs)y0l;grQWd!`03C
z`pM>^uhCs0C~n?#YC4!cXcgES4}Es<o3kY!$G{kqmAX1R;Xg(vi?vi#e!~8CQnP_d
z8X6fv?66VU36j8HD_amK9k-9AkysXKtT%m#Z!>uEB;hg_Ln2?r`E#vr^;V^3MRXT`
zeMNT?gdu}dXb}`zgdkwTXqEZ=9xY?qrewWAE3R7*Imk(-tTZct@TQjx`veIP=w7%v
zIiu|UctW8O{v#_o<bdwMo@VND@|cHjE@8j7{{F4H!pyo+iiW2r?9PX1qMsZRWV&%M
z7JE<Z*I<9arj`a{5+nBeBIdVhkD8gxf=TgP^r+achqcF?rYYQIs4p)v>KvkZrj+Cl
zfN;UmHLreSTS7+_8%QC*&F)-~m|KV<ld79n4!dP9mX<V^tV`W0tChJxTV(QPvbX~!
zpb$tO3~F+pN88R(u8&l2KD->NL8m6x(hbMl_Fe*x1@NvqRH(gT()UQnJJ#l@ZUbF>
zP%i$rEPzul>Zq10E0^r=eZVo>tlPKMKa4_UfrtV8Hg1?hYM&`xL!GC;MQ1{jLclgB
zmkhn=9Ow;K&ENdiE*n^zf-Rg9<H?<wY`*tO_rMJ!R&MaqeR<C>uA;BY^+P>8l0?g`
zMR$Wd@yh|R@`1y#@=IZ^TW4@9FsZYcrf7G1^UM2aw9bg!gsjGGPh9kyyPk~k(QBgB
z?Q3<xxo7BOM)4`5us+=MHZ_^ix-x$%HT*iHpE`!Ehk~tP3gvCxMj*GC)P}j~aewY-
z7N8D~mD4hY*Xj%5b=88_rq$nI&3cbc2L6rJIc)K*hxvxYQ9iW(hP2Yb`^3Fg(j-f<
zljq{{@(6xUS^U(UH?EU#_AD%%oqV<GC{BVr9rhT>fR!F~`04Ec!r2Ff;p`hz!8^N+
zF>-BC6%wuaz}utxl|LC(H6z6!7K-!l5E`D*-plZUjiHzsiJMd(jg9`jxG7xRIo!A>
zxjeXq@iG`$+@%k_D>byw@X_(&dGv6I<4$ckxw6M&4-d8sW`5&xJAT4zy#h%~oVmq5
zj%YaDxH?DV3Y&$b?suxj8X6L~w5m^)D&qWZr)tB=ymVL8ftKUCMVE<058<7Q|8syw
zSbL=R3z!)3&hygUpgO${(T*r+#)Q%P)NnBNkNt${!f<z=|HxT-(Il;F4|vtjz`=Lt
z(3^U#XFu%FwX!F+qdJ?h5>GDqyZ{RaFZ!b2>QKjOSx#Ts+GRL6yAuB%RjwnG58+Tw
z?Bh)(R$JF%puJNkUz68fbsn6GfTM1lkRsKYHQaDw<)7#)x(kQv3bQ<79U=(tc?!Pb
zyBY=ldF{_4!>|*GNXNI1EJ?}DzlQ~0AS4&nQ^l`*+C86X0rkSHx7lIg6zcN?3zE46
zCx39Qb5^@p<(7qt2y^Hy{wIEH7I}n|3xL%2;!etVQe}HhU)yD@PLaUx@SY2Lm!t6Z
zIjIUytSV-lvH{*UJ)I2GI6=}BVSeiTG~w+2=&fH>lTwV+`D|oz8pHJNY>Z{}e_7BF
zn8-V>&JQo$lWCQIzw=PB-vSc}{~Ifi)iYrIvybNX2XLL*SE$ZlI4UmkZ#ZyNR1yqu
z2vXN~hkp#-8u4chG*A%v@h7k~Y9?bKj&Sc~UU2UnsjjmgteVWQkp>0P(YM*EkG60G
z8(HOsv-^z08&%JjJo69c^PAYy<=bkFGuC(ut2_XY7T&m4e{NrbsfJTj1ZMA8tvfa5
z5uW_n5io=|7XRg@;)7+Lv3_AY<M%W4D-fp$?lc^~x(tqLrw7uy?pN7G6Z_ti#p2by
zX>b@h3SL8(s^z=&P`>uZjm^|#V~#A0rY%0}t9*`Lx;m%SykfjJ@2kr>p}8M*K@tbS
z%+PiD?jJB0_BV#xTIwTvJ8#!>fA&s@(#!lcGfUtDbdPkbR+KiZEj<M@qa+t=37<6)
zg<lGbp<f&MPrgd`9MB|hQ$+?U;_nRSN`NI{JF+}v!X^nH^E>?bS)i8~tmihS4Ns3r
zhf%NarO#u8*QtwzAJ?Xc-~2RdI47Lwp#q-F$VTVKO~P_#2W;fVp^(yBF?v69+QUP|
z%+~I^#*=D>zlX`;FwuYXROfwRJr^4OCr|Yb<zO6w|D&g3MqQEPe|xGQrMrs*Gfgk<
z5H6lW1mB7ko@pyoATwSLFvEj8mgiLd^;9|Nr#5f>Z%@TQO&XJ1xqOsh33XTDYY8;x
z&paG%m<WPR>YlW|Yn|P~w*@1`s3X!ZDs-K7eEu4?!=|W`Io1+IPvcK8(O%RE9xdA6
zb6@D<!UsF9tL)BxV>*SX4PAM8*$@$^!TokD=h523mieD2FET1fB&30eE)&=<&F_Ex
z)b1z{2h#r%2N%=9zCK3k+4q$Hwzn<AF$+GI*PTxMH`w+1!P3Zt^n08shGZx|wx(H-
zmCxRb5YU{_*RdUzv=H1ipGn)>g2#x$7tj-W{wE(*4iWMqanR#UExYD?OBmXoBznD^
z-(}kH?^<Joca$o>^u@L2j5E3mrql_U5_Y!V#(xiU!C^+Z0P#m#9BN}Q+Dnr}n~j65
zV1TgB1yLtt{~vvnGiWq~0~IVEf|F6Q77=02X5}}IVf#1AIiG{W?0Ih-_5<njiq$Zm
zo%UKYw#MIm_<L<Vxb|RDm%`y|=hD9)!F)c*`@5(ooG$V2VRS7HCeZSZW~*cXp8UML
zeAhCI<AnEL2e<>4SA1K@>f+)+$C*=}9^u-jV;sxEwSN)TOa*FrireKayN$5{Dw6-(
zMd1#^y@h_USy;`?koA1^|G22jaPeAj`c)g>EE{eBL{vH#!`rg#K!{Y6+Hi5ds{e9P
z)eMxF82ikz^_ztFy#TJj>N~FO>hKJ!r=F*{&g##))!vY17u2|(AxHlpe)a#NrOo>z
zmqVLQ|Mfs|!)R>(socJkntZR^x`<)S<UZ;@T*csV{2;)&H02t;#4)?{g$FQ(zC%@E
zh&~vIQ!scX&5@zy{{^7j;7*)g8*XcBjcwcI!X_7OCz8H5^Vi}79O3~-ci%tWRA#kw
zE{1O2)JdM?8fl?M!cix8ghW_GgznEC8QC_0f(W+}b^0bkSiFle@Q&i+G5b0Cs|`?@
z0k@;`)XAiXu)sEtALMTxPW?|_sTFKvvIjO;t@vv?g%JI+*mQyjzn%o{cCz9>dZkGU
zIGmif&4*gnHXq(}cu)K|Kt%XFm28+_?t9|td>2%TRsQ0qh}=qMMnZr=qQRfy{9HEU
zT&Hx%U~9YZP+Bk@Rat_&Mf_k%t+H<Kwcr|K<zRQWFj2Y>yeBdq#Srf8rXW0-|77gK
z8mp^P%@eFn(5d4s5Z#Cfb{i*gY*E<tci&p7`Iu#cH~=IAaf5ZIOK5!%=|xX{V%}8U
znKhHN3Ov^4Kfl9Qsz;vVBOihyw{5|VaT2HOOibgMfa(;sfGJC6@`@eQ210&cq?sHU
zpe>(p9MJ;^UK>qSv}@F6P52zGfZO6&S4=b}G<V+MZn`Wp21q1(ClLUW+vw?E?t%?D
zrLN9{D?`Jx`b(bgu@70v5~Em9v@6^W_pGg}I>e*B<g}Ree`9cAFE|+c1m*9F3e5^)
z@lr&$a(5(lkY9lRY}nq|V6&AbOuP^%&J>jFu})NAb#{C6TTSkRY5z5Bd=T53Jnv4a
z$$qTy)LTf(dLQIO$Ty5xEO|$9&Nc4bv*V{}%!y*TD$yJJp?^+Eqxn7aPQpytLQr|#
zn1v?{aU3J(<z>-sh)8N9aUV+lX|KoCm~->wHzB$+OUX1GLcLgQykS<j<r+uJbG#Vc
z>JLJJJAXj1q^rAoeR2o1N{?<ZVA~8hNsq5h7Hoe4x=E_z0wtxR71K4Ejq8qT{dnva
zA1ciw9sV;xeU?iwwq;?%Tinvp5~XCKOoC%$X}xQ{{r=SPS~nfovmL5;+uIHuoYB|m
znG7RBVq=FpJ7Ek?bQYaD87V14qGU$~nr()dj4j#007hU(ZXS{L%bEd34SUABCL%I2
z`dY%LPpzp^M9qIlw`ZS5JoR_M<wda@sfIkJ5Gedm-jtIOvxg{NIis@Yqf{g#BGOZ(
z{NsU-!2Rt8S5*mm9;x>(VaEk4^KP7ciNqe?emuP?FZ~)9twV9-fFLi+-QiQ~ci#-R
z#lkYkNWRmw=LHb53Lxab@AH;E0SSavA{+sSXOki#50ZK3rvbvWcZOur=R?)s!`pC}
zIf<)tCzSs~SC^N^*veb|7M>#{e-{x7Z*`e6ldj1weKdmob(!dS`1Rf8<Ou*2aX5Z}
zrtZOZz+JfbH9(L|tU{a<e7T?M@R0>sy?BuSo%8D5qq{8viQgV$*iLxgg~)He^>-10
zPjhEZSlIbdv3s7(+tWjp{i0RVK_;HROVoi1&VDs(aDOpVBMl%V+mTbJv;i6H{JnQ`
z$iGyp%0|u(M@-cGKki$Y6&yhC*R}d#XV$WdkRFTmdMy4hW+|-wdvi2!bB70LO_$5|
z0#!mhu88^jO2(Q_?JlGF6Yj)$q|PeFcc#>M_!DhyBfZG==<CJvct4FefS|9=0LFD{
z-x9D`8iTM;hq~xzhv5HK2pMzU;c7`DUeDb$)E7(AARv`(7Pwk~WXHzzg^duBIsMPM
zE#A^GfAF{wiGm^zyFd7Em#=F}!{|mwS~s|C#(4=jZsp@cb{BIQsunY~#<#6XQ1@;w
z?H%nBLhw}2XyB?BuVFTXv_C8i(!n|jFjW3d@6<$iOx;6pVMYbcKica)T6iP7to*~r
zpT(BojKa|me2DCyomfXTAr(e!NY=fqh-eCq=D)u_eUSebf7bvSo|UOFn8vwN+RmBJ
zzljsv(L2>+i*8j4&YCDTP`-=}X(5l`ZK2{LAPYp73VwYhGl}n$)n}s=a(sT0TS;}J
z_3z06)Bd8!oHlqgrPI~PfK`$<GiW}T`;m|UBf1Q5_#|%e_I^7U*EtLas}Q6oxpDvg
z`Z^p&ocAa8##&O)^dugio;>>ZFm$%H_OG^VY=;V^|6#+hJvjP$g89?#fxmYSm{vI3
zVs5*$aG6&RrH44L@bC183@_l31(JitCIP$;&pO>f`qzgu`~Ni_-J@`JiGfI=#j*l1
zW(koq3DQ2272hqEPN0NFTzVaeu=ywM(vHBPkr8P>4VV0XITRGm#!cW_)VqyY`$*2x
z$6hE}z1GI5&x9J?J8QTpnH=T?ZSByFy{AInGEt=A4ygbS<lrchgMoWu1N&MbYdN}r
zR&`>@qxoxkz0iZYtZ+vhU|<c`HoBYEIk>hW+j1*@6AUcp0{rs8o8RhX12{a1G8)L#
zWlt{dT-Kq7T5g>Vm{^knc3xfXcOM}&87>~MZT`DZFV4V1wE}#(zooy{vJTTy)iwju
z&y2{d@oGX(DAY`VhWY60C33Jaru)<^IYy!fj0`+k{okdAfzt~>cB1`R<(30J3wBzg
zdQxrhH>lY{+~SmNk^<sbtqX(sTw1=EVGp#|UpIj!(W&hhrg0K6t<y%L)yq3ox#N^C
z-x8<{xAHe|agp@Y<9KFQw7Q|57$S90(BH!-yuBhLWKjnfe|;CLGe-VuocYEHF9O08
z9)_;Xz>l<izg*`TQJA2rKm{yJaOht<*oNBnyOYkt`lywm2s7}&2jg=pMYsRndMLaZ
zGt0TXx8`H=!6*)R_G~la@PC=n_RWL*&m^zWSkC=SQl5WrAhLVl$<M{=pB6nrHo)3l
zLg~FvIg3B+LwaVc64CFtjRFt0jpTl(?5u!ild%$leY#s(@@*;LqA2t(S8+_JnfY6C
zq2@Aq56tE0K)=v>rq=ZkJ<k-SM#U>SLgGF&l;MK??OVj?AGJzwCZ9mq{k%k+vMY2+
za?|}TsF=x&HiYA=a|mpBNRl_!S5&qZzA~3;C%JDTlih`a803!I(Y_9n*@Zii4O%HE
z*M_LUXDU!g+1fP4MiaN8e<a-Z9K*VJnCukjoSvEPSLfUIJg1D@6J`y9AH3t8CLYOu
zlO&UFln}zD($l>1MP`XD{CaEpvBk*+g)12gyPaT-Jwux7v-6sdPchB3p}9NzTe@X&
z>%nXiT2uC>?#ATN@!iKi<J_;aA8wV{)VnLxb>FP8R7;N{P@KwGGxrmN9CyxyF<h?h
zScQPbTI{{_epc6n)UAfW*b2_=S2~&txHLBx55#{jF?V2og)%|L$Nv^Dn@-+LuE5?z
zY;g9T@0Rz|>&m0;Q{M_9|DUAYqwyOUmUAea^Q>`wzF0^5mUzA?(9Ojg@prlXx@tRX
zWMVx;C5D3QDyMPGesEXXhjQyDxu0%2O(~AI{w`AnafVw0O?K+cBgur|F=F2e_6IBH
zhY$yF)6S~4mX5~XJ=&y(w>}te2(1$BD01p-9Rw4o_K=HbvxRp$dE&xJ4kOh|9VVPj
zpA6qf=C<8P2<bA3kV-oLIpb-E5qlH)g$O>^*f2xq(_R{+Kc=E{-1mGCe?f-efjzR~
zn43!<t3B!09%aI-yzi-^r1{N5P3b4+j1-4X1d9YVL{XnYPU2gVoVkj2sjkzOJbF06
z@Ipv%mX1-e6a6D3^<w4agjCsFK+ayZzEK*+t*6sF>IPNKHg<gFG@Iw(yaK7#>t?})
znV;IOc~OdW26}9bNw6>u<{3#{ewxJMm@N9idxR+tHogh?kQZUJ5SjgDeLj$NpWDRV
z`7=kemosj+;2(Y}aK1;2K%Bq&!b7i@aXXf%Z?Jx7{fxfrp=|lX#q%9OG96A2oR<_B
zW0S^a8nUOf=+@7`k?XjCh~h%Z$XoVzHB&uZw6nWNR`n0r>0-+)VqK>mUZXKClNG%@
zv}UXsyFd(Z!J?=R?m4dHTT<_7Uuv(KyKe%j4JU}-v~A4|I7BEjRwVTn6*Kkjc!zQ}
zX{EO&ItEYAi>+9ftst(dH<CRrUypl^V2c0M0w}{nJy&2J*DlJFaxnj}us=|^tNOf1
z!_lT7eB5^chSkxAtkMmyx7zwXuI4bDETL2Amr#4LLC!j*CUMi;@O|=>T49In1t)=D
z+A12G2G*u;vrk!+%wN#xYot4XgM?}7BUfihfl~@?hC%>v!p0|=yrniHHi-ru&Te%3
zLo!9zzD_yaD4Nw`<EmFA^p1k-4SLz{50>&))T%u6OxFw4L%vukZ@ew%s2zNHWYln4
zMls)X@0dlz@J8$6*Smve$;XMdDG&0uf6rZ_UMQ|85R4oIR$;>2IH4m8R!7myLc&5X
zowmqYcG70H%Z#dfRQ1p0%`X;De*PmlcLWh8nR>83zniz>Q1y)4sZel^KW^qgC{rzG
ziA$e-fW(Hb<+}6<&D8#*r@oYb`t>?#T!xs#B%(?q%iK<YeEi7v&OM@42;S{g=;ix(
z`dO9RQp~&9C3?<2`EFgB^w@xSO-(a1Exm2pD8E<@mn*<xOlxep)Y0tYWO)HB#;sYe
zQ}W6E$p*UN=vbSW?3qk|dM80`Gcz-5>-@3RcP`u)#-9%>sL9L9UWAk&@AK!jDK9sd
zw}`*lLto+;lHrbx*-lDI7EPWbCnbHTMpQR?o@6~zE$`7bqTy8&Nlg-i*&tm#hHLq1
zU2?#tb^mM~mtW7iapdF`$nUlb*6$9#=@el}=Z(49`ZJ7-!#B~c-d46_)kSEP7|F!^
zY7@wkDcsUhOV!aHcr5MtCE|jnqm4~@h>=MhZh!w_e89Zvk27-ZUuzbzPwU<0%OHf7
zrI`QV3~3l__O==vsl#<6aWa=H6DT3&-s@`Y|Iyp6SS!AuB%KA^?hn=*D<?8v+{ulT
z1=y~v;NgTBxFx<|zxDmshx$(N$>bmH`t0|k3{N7hC&1GlSy0cu8a^b-EXCpdAmbYr
z02DI=C!FFNmLsg^tMkawQ_?=?fdxqz;gP?_H+H5Ow#KzeDd{a~YPEgS!l`t6k-S(#
z3&V7wDKGpx+^QSmpVdBXwDvOiem2dm{m9R`&2u^<bz~gQD1(mk=Wb*fB#U;KVq16%
z^SGblo+W+ezB^oVeQ<E9)%N-G&+-j7eERzQ^hb{%tB}rG1uHpo26y@9U^LB5iUxuI
zF1DF}W<HofD|m6~RWPMwkgrGV^8!`bum|T6x0qHz$3og|KhK;rWB*s!4<gtq{R<M~
zf#y#y5FxAl2l<O1(4D!TZG4j2>?GAd(_AVzLpo`G#wp)}NHIf=M0unnM?flvul17I
zgem%)iyZL56rbK;Eh5VlmLAPH|Lw~mWYtYU=2pvI_;(AgrW)mwNSy1%%7k<u_m?cO
z+M?g0^FCWxC|Is2cVFP;Th3$xzLO^F6p|z3)qBfG)w+I2g;TTG8kRz>z(v-UV{a`b
z-=i;1{NMo^&7_&n94$+GEs?bjgUfai+50g@2G6Jmvg0uv@1C5beS9u-^{Fq?YAnw`
zSK7wuD=9En$`tCf!>QQnav?x^3awDK#u<Y?=B*2e-?xEeFTP~cU9|nR!H?dqV-3DS
z=0IF9r4RArKpW(Irgv}2TC1(bEm?gEOoxddJ-buV=d36jnoiE>j^+bM9XGuHw@I&)
zwMXn-6F?+Ib}FBl8<_gCN%g^UDQd(@dq3Q8R^{$+Z8X>uvZ@Gw|30avnV%j?aNECM
zsy%s`IP=}j{aol5qRAH_v(c-D6Mik6-B*=4I$$8-q|_n1E214fU+?;Ujk-p%dfC}L
z>9bk1jh5Owp~?kiEEgA7k|E#bO9+_|Ur7*W?dlw(v!2B&yN|9kNIt2rw=JUJgV6`g
zz?TV&Nz4Z>bl)BB<?2v5njDS4fE!9~#ob$L3IDD};%sZ1#0KGs>I7&4`MT}b1?LSX
z!nr$CJbPw7GZ*Typ|>O)f^u&!<)aLLX#JV>Z{4CBu7#J3lY9czTG(BS^f)zMZCUpW
zh_Ca>O|6AjPI6GC`qEF|`H*C{`z1M=hvH3c;W3$|z=b=Z^Lhi6aK$-5M$J@S^g|?X
zjZa#-Wyu`hl+e&<a-`}N><IhlRJg(~HSxw6{&4aic!B-R2D4wM`wAl>n4UcMmrvew
z&Ru@lLq=v7#77~+oND}&^f*uC(+^qEk!QmmzKcyZzHgr1%PHh3ayN<h=bT{M%v6&~
zp3u1pTx&omWrSM$$E;4D-Cx;9g4eq!V*eOU<iEi*F3(19OFn@s3b$OUOmK|JG~N^n
zeyet!#80K$@xvng?@!;h#L5fTn`(o6-!hF1w*3>jHb~CfscL;Hp-l6sb+u11QoQ@5
z(X{eNk><XV{OVx*3#~8d-G1V$X#zV&<$PNqk5wC|FK~0dRraS6{4p{=b)h&>y|<+5
z!va1o*jn`2d$yciwR4U*PVuJk249Tfk3`41A@nVDfM0r37#ZPOy8q~{GP$|0NTFE#
zTxic#pZ)xWUw_z5jcz2*Q1+eiZ0;+`=3WZkbz%tFbS0tLGJd$?N<!P0@cs^E$XXfh
zsIecLwT&MCo_X+FB|oZ|?=0)-;oteFwwrs_r1$O-xeH%BeLCR#_wNRIFU^j?l~<3#
z`>$#tp_ckH^yrNqZs%(irR6=F=Wd)jsy(h+y;sX%hyf;`hca@1#9)n@NTOI<T)C{5
zbo*n1kXW9|FjXJvarn4n6nO8LQp|01|L&E(uZ9yz7>?YXJWJH0z{b8XutqbV<H8Jo
zw~dY~b9n({XU*+RedC#p+CR~pb}4D&wRECiM>N&UJv%Aic!)?g3Qi0QO7K4g*hd<0
zwS!aSOh}h*81_s~Mwp=*Hk~rUPkpq@alJ+~ih9F_DLBEfsCYb=)*wPTRYR(=El9g(
z?x<Gci->croU!t^HBT5s$aE}ovcwr`CCV2Gpb4IvmYCZ;<vE}4@~XJ+ElhKWp^#~j
zdRW$#6zAtMU)+)g;)jS|-*Qb)(bLSISkVb(GKzQ%)TCGyXEnP>xzC@cG9<?L<o)IW
zmnk+RGw!%+!v)5AW>@ODvgKV8m#qeq6@+_UJXkSQ6MlGGbA^$m@(T}Kw{`zCc-LrO
zPuttoIoN;hejGW`SYm8^<=UUo{8pF5;NmqZsTIR(DN_fewt|jNI|u{VJIqDaY8@+Y
zUp%g?_E<WubxxX-XUCdlq0zFFnwbVpD$I=qZ~W#Xw{3(ob^ax#gn(-6)!Jl{Wou25
z*~g}s`VLMhMt}MBPp(->Q|oROS>7JnjM+(^;^j#q)yal}i6U2IdbG_=)Z}|wx5oFL
zh)?Ce-qOrqp*U$2p=5p~aDNFdf~-1%sdR}{WLGCD=T{VJOk}W{2^L!D_D&tpo>Bgm
zM3yg!`OIv1qe6TQ)3LjMO1Z4tE-Jm1oS4$-X{EXL|7q>a!=YT`K29a4sA!?h(xioi
zQ_50|G9;o!2%$w~NR}*-rOi^5w9tuAmdJL<l4WM<w2(Q-kSVe+StH9>$GqRC&UMat
z|9P)>d#<Y_W9E78=e~dU@4lbk_w$`Ab-=>gNuT_F;3A!#byv3hYT@?H52D^}CKhEc
zU0j4Vx!upG;Yl9&(*N;ZaOrRSJgSZ10I8LI_;QALm-k}>@~gHy-pyezr;X|rr5x$}
z-e=Rz&SCY|8vVv)3x;3#v%2?B^6o_}P4dtr-jdG^&-fY!A3gb_phVT_OR6mUS68?2
z44z9@Tt-lzlfZe!Qb%{jq{r+Q#_Gd8JuMEUy#3VgVRM(}HU%1_?3f!NFR##7l}>bD
zE#k_rCbfrY3pSA5KFpK!WbU>~F&K7jlGJDo2;UGo>{XxUYHiefziLA$$GeceY~4`T
z;47O&<GhJ!1BFCCwd9eW>0>PNZnndV^Uar3-i5w+oE-P1cSWm*;h-18`0yb6*R_7p
zZ>Yp(3E%B48JG8rZ47r3@ANNH^7oX_Ba4mMTNJSCw>%ohQ}M<HK~0MS92T69W4asN
z^50#g*R*V+kba_lFmd$ees-<n!aWF#Z~E2ADqpAG2DIJve%&73Iy~c7`qH|7HSq~<
zSNOtYZVp}0^{Ts->Yz`T9ORuJ+$m6ah%MG`WqCVt(vlb>T2i+K_%#<_hvOr`yu<;m
z!e6z=?lgmZk>1wA-@afL#36dHHtTCy*&GuS6SxVn0WhG*kjV@>-I@Xe0cVgW1g@48
zh!lqOo_2lS_ADdgTNQ5?TuEYM6-nWG2`SyG#zieh=nn+x^v3Sp*(TCNH6@Al>C^I?
zy4Y~G{OJgJU}HOK|K!NzCe*O%85-W4==d|TOU#N5elLp1PGmi(&b3W#Z9S2A@A&fE
z+*~8}o%ElH@P4UXqsJ8@Q{&>aDF0L|-3TezAUc_{k8!=>pYmhE_1g=cO;X-nk0Gsb
zYA75<37jGO?`yS0Z6>I4Q1z`-Q5l1fUr|<8G<=PA;1?7^7iJKsPC&B_eH-@u2(f%x
zA)yNgAqo)>2lzHuGt~#M#Ey=Rh*OTAeqS@{zO`V*k0#u_Mg1kg;-nbC^ldw{e;p@P
zHY4hN`Hvskg;(*<z;=^#BG$I?AuS<0d&|Fn(|cvX&pKB#4pHaYJ#AfPbyVeURBj$h
zsBE}BX8n=8?_vHq?4C9H2v#L+Sp%v!WIjEgJ;cA>u$n$@d{SCuaEe<Yc9*7XI6UEl
zN>t)5FqLDmdtYICht@Z64^I#-TF;YtH6y&uux-aWD#!DN#`}eLZqE}8`cwMSjC}lT
zpPiE8;hd6Qj=2BV=X>2<&;Yd-8<*C>{bi~{Up6vTWoF4XLx*_>^^69l_m4Vmo;Uw}
zi@4h6aVsNHMrlfKt`ko)bdi!f_VH)!oO~XtdFOda^6M7os1D9@_pif=D0cs+NoRUb
zf;ND1RRHtxf;;fgs9~44pH6?|QAY(Yz@AZz%15{-a);bpiW)d-r_s-RUv)h>rj$&z
z@WoCa9&2bxAB&AeFZ(o+qS8gd`^hZso^D<1M*63VQ6Cyb-o5W;r6@0d*3J$BUZPXo
z2no4=DDZeqZLPYMK_04N3knK8e*75Py1t~t_4ApEJ@HlV-s$eKKcK%%Tzp41^`*-S
zjag0u67VTT<tZaaok#2cd+=NmLp>z3tc=ngAEs&l+y`ZaKm2BXT+WZXpHte!=U;1n
z=6WeqVivK=^&dX`4yVPilJl<w#1rV})~gU<iZ4zFJrl-~kX~iv<#7N9R}$7nckgn7
z4djKLny?7~Q6VA%+5mH4=^-*ID*F3(1Ih+hU4X^zAcSA(nVGP>MfE3y1-IUw&<+$X
zK;522Q@Sst&sj(M0#}_Q%sM=Y_AdEYoqw!ly$AJ`Vc(%QhJCCnQJ)fn@#ogR{@{E5
z3YP!(5F1r1?K4+iH3LGoe15<!(9TgMh+79-_;`fnx%cJejtyx~p8TN{3NCyMDwi9q
z(2XU`rV6@P25k3OC_1iFuC%2mc+spAm%iUh#j~}AMp;Mx)T9dRMGwACAxVbqu+k&$
z31?1ci}QoFnRdH};1$PLZDf@06`C_kUzV9o3J*r4*c8dcvnp?*j<H&~TvJt9_;t!v
z?ZkWBNZ&ZAvJue>Zns3o(_a4Ka|0EFLMR@5^LO!J3JMN&LSN3VDC=@)CW0m3CDe=%
zU$w;W6o&Z6Z2>$kCk|VFl&TPiId+AKum8uFyzFq}*SnrXMJe@{si-7#8yhF`0sYZ0
zfs;F#*+lVi$lF$;{wVNO$pd9(Hs#tPB`T_iIZsP6&&zcM$=ci5iKrhh8|O1VZXVGl
z7&JZhm~6~3ll}0R;4TW1$1exN$M7rqfqUWV5o35C5T+>dB;u#2gN{}Z<kl?m8@!6|
z%_xN10gH9$DRURu9b$!dC$bO-i=kdHVqfv=fh%(X^n*I>*-V8g#MN#$H7&+#*%6`@
z{)g>CZU69FU|=8uZ-^e1+6|&7aUzG&Nl%?h3=V}^#)}u9x?Ml)2i~(G<_f_IA)*-m
zgHW|aoC2l~T}s_bXfma<vonl5bA*CXj6mH_oD)jBh*Odla-Oe2{Num8SCQ2M(&u4N
z3_5`mtUF=%3wa3PwFvxeMz0Y&fPM_ZxJ{;Mgkv-!l#7dtx%p>Q^p-K1micgfeGk6g
z$Ce%tqnkm1wTxkzg<s3@RGM5<NN9^VHFHabVtEs>sM;6h1a++-M4s9}Euwl^c%^bO
zH?_NAv})^RH!NgM{^tpo_mi&_21fy(A9L=mr>WW=7Xdx^2JQN5B}xL!Es~Nt@u|=y
zH$o{3kESM8AHKKpQkj&ybGLD!E}635xqz-}bkaXKW)LR6xYC6-`U4qZG4m4e#6^Oz
zvP2x%3KFB`xq`%mgmz3!MCX9653a*ivD57vjLd@M|NBavMMsFU!DO57wIq{WKlFfp
zqdg)2YSp6fYp`BVc-=wx3K8b*TPt^-hn*;AF(*hb^T5Ut@G|PE3p_W{iZs2~-_|`k
zJ8U7~6uyvjA|Mm0)xCSK2Me#*(oVPyx^vU>8`Zr>lB)$sAvDlxuxn6UJPizof}Gr9
zQ7blKwgYSkGusqC%Pf}WF$wr<WJ)C7YY3#@NJz{gPAHn=3yN8H;SAs|?VX)7^$f11
zfdNh@5|||}#tf|`aFuD=5tp>>Kkw2Aa`V~$%>J3wf6e}{3Rt&C0sN39!9?tM)TX8i
zjj=BgW;!xrAt=Fy3G=uQ-Kg>A->spGGzjVx-ZxK%1Mj-=Cn-`oG}VAF^JlW7A|u&D
z?AYzBqTcv3v$}OI?GYqY(<Jbny)i#7o-&bEyH0tMkYEdAL&u0B@X@jqL&d`()Uk~C
z*y8AJ5uZs?>+U}`qMg7btdl{LdW6z#aVxij%-y)a_yM1<J3<(4X}y)zKR@H`w!tpk
zdO^w?&GM9O7x!Wy&ffMo^93OcI!bqG(-mC-&;hqb(<Ub_?il>P2yGSavVp$7F4PH_
zhG4dro2Fx<<I1Xk)rB?$XZw?K`Kog-9*^Q9yH5S92@%0ateJsq+*s1$`2nkw=lRwF
z$Q22(J_*2yT-qDWV-27fK*$3Qo3(W#xDSMo2ku3ypxJqrnvoq<WfmSi;(>9-YQ<Dr
z2L%u;Cdfc7)-pd!J&8>8Wbdh*!@`4&g(yxTa2Jjdg0uKVuj|GythuHsxALX}1~2S}
zlc;|1Ro3jdXWZAp;deN9?wg@XgEEyR6UqGe<;ygBC1mSEim=3i==^Nu&3%7gVF=8M
zc2^S?O}XY3Ga@wkpFP(a3k>Py!nGei2E1-T?#H;Z<>Us|DL{q+GPg4^xnKF!@`ZB?
zqOXPZLvS=!oQXaNn2jsT3Ub;%0)(LafauHe;6WFt$Ax^Cz1MdVjF%$0L(@dqrz0<A
z&xoz*TU5pTt9!gP?EAn!RgrIKXy_PGZ!lj-NCFK?*xL{I!H%B$%@cv8Y5bka%*a!6
zuF=4cgar6T8h71!kLQGe-~R4YJtOZO%ef*c7oq>vv4xvnzs(8%VC#rRD=Cd7dWyQu
zRhO?^nI#IBK<e}2ygXAP;1VB0h-&;Mt7A>g%m8}k4hE5Tb|>5|q<=$a(n&~4dY3Jq
zxYT!9_x+>9Md-js>vOT!dFZ>RcDQ?9;jZ!4^Y-)|*(Xg~etC}|^hf_roLHTm>RS2N
zc0Z2gO@Bp729**G@k)K)z~cjUU}2-vWyz<O8{qOthfGaRS6Um`=qzS665X!~9p$)r
zt$_H^II82KKR-!58T(h4h<2|vCuKvZU+XDGwX;KP4=HIYXucv8y=j5@0BpZA)6$~R
z3D6a6E4hs^O#_+9$wz>YhvWH&&P3Q?7*HTXIYUm&B+a7^5aX;10z=X8=arg8Tm#&k
zseg@KteTWSq^ql|gO$`vNGs6^{}HXgFfdv{#JmqWlx5*?33*T=P0~++mGn`{HwMpd
zN^yj=_&SB~2_#r7%Z2m4B>qOfA(KEOLKq=|g4nye&<~=LJpcP>P@BIR^zhWyY(7yn
zP%02AEo^M0ewl;z1#n9u<Si)Wgmt|$<A6-LQYdu(pJi#<(6(6T*o8L?L7s8}f=nHm
z<-(KP87venB9;lyC1$Rz6{_*4mSJwW%^~I%RwI;?5YGEUQU^NEz>c#oXGAn?G4El4
zN#q*vlBg@b_p=YJ<%%=L1FE-4FM-%aNU>N6yO9w2LlqMc5I|5004lo^71irTEj@v5
zLikdvQ$>X5f*zwUTBY2dQ*A^*GoXDzM^{}f3E+|1*XUtF7F9m#I$|QTre7k|FE}9`
zW^kCYL1LJpK7CYsd^Up58X&a7bmPsey-rO@-udf%bbls$rL&IC{{Kkj{_BKLdM{*U
zFJ2sxh2B*BIz`>77_NVb_YU>6hA4h}G_?Bo@<8rMaCd1~l7wPebFUCutVarX_Dje>
z`1J17_s0N97OcE|I~`6uk5v3{hx!Gf$m{O@gM-)zkTR4xjHfF1Xyd_nVyufTUacY6
zF*%WwlasrO6D69h6xirw4b>)_Epobmz?fVYd6ZSXs?6Glf(&jgGFxdcK!GH;*aim^
zuI#H9N~L~=m=@@H8Ip0*KA5I;(<YHnO`>uJuYH(gb+JE}#Hy(o0g7EB)!ZqDjF$Xh
zhH_3m+22&0Vz_a4pLg~14R{n`KCUu2H7fjC39|)js{Uq4+v5YJ*=dbU$~f88#L4Qf
zH+<-ZZhWH@;#=*<2d3s?VDn%yyq1LS;er)iZzL+u2C#lPU!nFC;<)9)e*F$MQ!`@>
z2G+(W9eW4hVK7D?qXDxI^dgJJ5`qkEzyGAMyqE)Z5YFE=vZJXs+_J_DU+r<I=oY0n
z+45$%<jJZq<Z~#OgKE;tYa+~VBpzt19B${AbNPpQez5pJm@tf1t@=8g8*O;|S_ZF}
z)6fh&V>K?mVZbjRX9Gj;ufs5h^n>p=#D)EXTv#ZpLO6{qdYn50c*acM8L|VGjaesy
zxN4B>z6g;D?*~PV-j~gU31-*W6ah+ZyQu;&*@o^|JU>pjEWoE_|E+k~XllnPuqpa&
zkd*D|HzS1HwM-(L$1Iu4J8?*#3+~uCo+e|W+{w+#smbrl3m2S3ZvAvGW5!`>E}`0D
zMx`CedQvVWdKPj@_pMRPj>X|n#7z(lWWWrP^iIATh8>(waO{|B%J)jKBs-lumk|t&
zsx30`ezWi1G$cPx&&!b6#8Vx43jSrXPjimeIfDe&=11Xp3_xoh%k4WOac&IDRLc!y
zP3(ssB<}4q&~{D>7CU4VBym&Cj=BT{U)ULgfEW>gp)M;gpTm9#q8GfU361*zZ)RX(
zVuJHW({!gwdQ&?j>-mw3Rw{QPB_j|WYr}TBb|*&G3BmeLJgmfWWs^&Z1PKn7tfV_I
z>8|V7kgOzn0bGFk@VmUmG5Xi<7y3HjaO2a<@wf!nYX6MhNuYcExV<)JhY!PcH12L#
z>k)o>5jj&L0Vb#GNZ8f!pBau-fo~!tg#en=_Y;#>As6cWz7C`&gl3xX8bS(v5#A@H
z*5L^Y!SM<*>0%L)?1w}BP9AT*3fv5qRpOQzzN+8fEZH&f9i-wgJf`7eH6x-e3D=VT
zP*YRm?FG@Dgfv0Qs%83*#7kW#3}Hcw?d<^j4C2V>V057fmr-Zcvtd?|k&%RdY<hhY
z%sD-NpC%eCB0wUknnH46?cMO1SubqE03?m``!tf4XgwK=h<g6U4p(sf5;HO~HpRZq
zn&p%4XWh$`-uxq7)z;STH90-oNd&98Us!u7yAD@GQyDo9Y=w*0sSiC&6%`dBqxLso
z;&B7KeDXqLJvCm2l<J6E)nLP=#lsTJJ-(6k8J7joV|_Oj9oU$IMyq8K#-VyLBpVuU
z8Z=6%A7x17neSnif%hB%ogW{L<;jy;oW~h6=%h{;$HW$Y<On)XKV4wdvr&EIz=G-k
zJEK!;AA4lYyOxlw_p7kn^C;6rrqW9Vth!0Tb5124tPKgBBewTc^#M_V73La;7s}dy
zwkiugXRE)JD<yX30Wb8{%?h<`K?7f@MYXO2+<L{=PMNuWmfK7ZA1?MD)u&)W1hcV2
z%?}F=`_og9iO&ptu+QDDq7qVyea}P2#u`QPVoca?P}*@npYD(4Ozw$fnpjx$#K=iq
z(rR;GwZ&Es`)d34g`R0EbnmU%Woa3ut_l;D*DWI9U=Lh<)hwA(-7P04hn<mGv=&&D
zVZYCFOJ*8SuM#`GhGaf0fhl5~G@5SCE1=R%o^+XnSl{=lw$BA+^qX!}d6Hb^GBL5q
z7<(+#p2=BbyUV`7d-Pc_5K8b1ytts&PML66R8ZhV+kd;SZV70FpC8?609Ru|@1ZA0
zy*&X3n?Zk9-mIH!M|dzmSw3XH8tJ_H_=M*{Z<EuNcFGDzet+pO2tt;xlm!F#u47g9
z)W`SxZ?w$j8;rV*?t0~=opJ?%lRdk=9#9;Ull%9Fm7iV`4k`qA`Gtb{d3iU2Umfep
zO^ymqPBST&D=JBadI94ZMlgBU$>ez^ccwxXDzailcYSJg_hGUA9kjIVPZsIMeJ}iN
z-}<Ik^!q)B?XdP9>0cbqU+g>nwihkqzk>XZ?^1IqtSZ9McMA*EKe|GgUKvYf@-ky1
z8A~c~HbJAiOroZM?+ykFs12t?I-$H$@chM|Cxlww;Zo0#VY8r@)jTEZ{a7RMMu*2n
zJ>oJw$aAHDwjbZ{<N8j1Z+-i$C+gxdngv;E#fsm;!&}r2`Q9$JQHbnCSk}<cc(x_l
zG*x2SaPqTA#yRW$Ph;&Ft(wR9-}ZhU3aMuo6&6klo4N=juq#JlN7Hm`jwGMRY6OZX
z$X<hw!f_c%2;K`RJ}LXyt67<@c9^u5g=6&l8Km1QKg2YRWco(%!yqZ8QECnV8l&6d
zXf!KI&F;zO`5PPgqiq6`dZs75v1&PX4iAf_)$}mY?uxgs0{EWW`Zff74H%O@8c&NA
z7aeWMA-b!{!t!o6^+!gL%*2_)mf61f!JF_b7@4$*P1Kn7sRHCu^z(fr7Sd}Li{yKj
zCz+ttw&0$3HxZS3C!!ucoa<o$l=#y%L<coAHI0xHH8VDjudUppqcf3IOIS4zar!N{
z=?rvWkp{-t%A2R5ihic!JKB#i4+axwWuI$-+@?)PQs5QU!E3D9tLZm=p{GW+sS$TO
zxqlLFeFNkQK7o}^afsrSc5KFY6g8x#(r8?O4`%P!jGI*b<0k1otxoCx=267Ot`;fV
zxn*}#jr!KDgry$oPGqF#aHnBqModghad9#5Q`Z6l(#@aiZUZrFoq~c91q-MTRaGr&
z^yFl{&FN;TV;X9Yj4H|CKtDeH2hJJurB!7H=AphOFdyNu4RqPurG@0n?^Y$ls6Dl^
z1`iNEo@a1zQS;}|FOIpofo#Rc*StviHq8U^Z{3CsPl5I*Dgxh^&8nY-!dpma)8@^&
zN93h6eI9@aiB&3M%1zyq4i3GDCs9#RI<}zOfxBVcHtNj<l>jv5X{5w(cAB1^#@_%)
z)z;R=?C><A$cu%X;iP)p0q-g+FR#{iOZVuc=vrMn<%g}UtpNDBdw9T5I^C+swER8d
z*%Sov4Gj&%ytw6etF9TMb@4((q6kHsZ_cWx|9lW08yoxJ!6z2$M6LzSP!%TGfq`>H
zufwOBXOJ1Tad{Ro01q()+;Ick@n-gUM=5UV4Y>Zvm0taI+YsHRUN4oDq{Caf={vCj
zmn*^rv8=AXUSp4U`eCxq`0qu%*pZ?ob$Qe1IP2aU_U+ha@YC?|Trq(a$HMj2CLAZ|
z$?1^?<}rdl^?Ouuo;{Omx-x-@Q-1SilM+lk17o5oy_}UpczsPX!Ec+KjeXi(Zx(y=
zX({PYD7I(Lq_zPNpSzB322vt?o51wh7FC{=mG$edzwX_;7lz1w*w@4G@+tYb&9_EH
z#S`dPi*yfc6k(TslcFLxA#^soVe6JHFdWbglT80;zhLfbY(spp!U|<!P&}9|wp(-n
ze%!N%hczNGvyXIP?QZHUx$Ac8f*NE7J-xkKw{9Kv_62wuD|}GrVJ~^+%+p|D3H6%c
z(V*Bp$mnN|94AjD2ZK^HGn)XInwzU(J0DXks`>f&jvaV_p(dU4i0t2Zc6(HLCpG{t
zxTB{M97tRXqpcR(&Pw5k$s%=so_>ypDypk7e66iaQ%y{kxD61s7>`5NVJg2?43wjF
zp_?qk>mU<+{`?s?Aw4}EaTFV@9j@5#Q&v{K7BpN7SG5nivtY|pIyKU5Y++G{FzH*g
zZ~a}^M(*(FWB8k4L-hp_ZpIaz)gK`-$0s^EI!Z`LV93L%G(>DOGU(pX;fr`-J26&q
zh|K!5@W)=P)-S~ThC_PJ-#;SW49740GC#N0ukkIuK>P>aZoYXx_t@85LI}liNkj{$
fhU0(vHqHK32Lrc<2(4U4yvbf2BTB~ZV^{tI>&|Uo

diff --git a/Function-operators.md b/Function-operators.md
index 2e4d964e..1b54d124 100644
--- a/Function-operators.md
+++ b/Function-operators.md
@@ -94,7 +94,7 @@ possibly
 #>         })
 #>     }
 #> }
-#> <bytecode: 0x55d90290c0b8>
+#> <bytecode: 0x55d5181c6328>
 #> <environment: namespace:purrr>
 ```
 
@@ -120,7 +120,7 @@ safely
 #>     check_bool(quiet)
 #>     function(...) capture_error(.f(...), otherwise, quiet)
 #> }
-#> <bytecode: 0x55d902aa9948>
+#> <bytecode: 0x55d518363bb8>
 #> <environment: namespace:purrr>
 
 purrr:::capture_error
@@ -132,7 +132,7 @@ purrr:::capture_error
 #>         list(result = otherwise, error = e)
 #>     })
 #> }
-#> <bytecode: 0x55d902b04180>
+#> <bytecode: 0x55d5183be3f0>
 #> <environment: namespace:purrr>
 ```
 
@@ -281,10 +281,10 @@ withr::with_tempfile("logfile", code = {
 
   cat(readLines(logfile), sep = "\n")
 })
-#> Function created at: 2023-12-31 00:43:08.117083
-#> Function called at:  2023-12-31 00:43:13.123988
-#> Function called at:  2023-12-31 00:43:18.129344
-#> Function called at:  2023-12-31 00:43:26.137767
+#> Function created at: 2024-01-07 00:44:00.426587
+#> Function called at:  2024-01-07 00:44:05.433612
+#> Function called at:  2024-01-07 00:44:10.439013
+#> Function called at:  2024-01-07 00:44:18.447574
 ```
 
 ---
@@ -333,7 +333,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Functionals.md b/Functionals.md
index 724c49ab..263cc0f3 100644
--- a/Functionals.md
+++ b/Functionals.md
@@ -36,7 +36,7 @@ map(x, 1)
 as_mapper(1)
 #> function (x, ...) 
 #> pluck_raw(x, list(1), .default = NULL)
-#> <environment: 0x555ecae48430>
+#> <environment: 0x55d958cd1430>
 
 map(x, list(2, 1))
 #> [[1]]
@@ -47,7 +47,7 @@ map(x, list(2, 1))
 as_mapper(list(2, 1))
 #> function (x, ...) 
 #> pluck_raw(x, list(2, 1), .default = NULL)
-#> <environment: 0x555ecaf2f158>
+#> <environment: 0x55d958db8158>
 
 # mapping by name -----------------------
 
@@ -65,7 +65,7 @@ map(y, "m")
 as_mapper("m")
 #> function (x, ...) 
 #> pluck_raw(x, list("m"), .default = NULL)
-#> <environment: 0x555ecb07d330>
+#> <environment: 0x55d958f06330>
 
 # mixing position and name
 map(y, list(2, "m"))
@@ -77,7 +77,7 @@ map(y, list(2, "m"))
 as_mapper(list(2, "m"))
 #> function (x, ...) 
 #> pluck_raw(x, list(2, "m"), .default = NULL)
-#> <environment: 0x555ecb163178>
+#> <environment: 0x55d958fec178>
 
 # compact functions ----------------------------
 
@@ -139,7 +139,7 @@ map(1:3, runif(2))
 as_mapper(runif(2))
 #> function (x, ...) 
 #> pluck_raw(x, list(0.597890264587477, 0.587997315218672), .default = NULL)
-#> <environment: 0x555ecc0f46f8>
+#> <environment: 0x55d959f7d6f8>
 ```
 
 ---
@@ -880,7 +880,7 @@ library(rlang)
 
 e <- env("x" = 1, "y" = 2)
 rlang::env_print(e)
-#> <environment: 0x555ec937cf20>
+#> <environment: 0x55d9572060a8>
 #> Parent: <environment: global>
 #> Bindings:
 #> • x: <dbl>
@@ -1015,7 +1015,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Functions.md b/Functions.md
index 9ba57553..6556d48b 100644
--- a/Functions.md
+++ b/Functions.md
@@ -20,7 +20,7 @@ library(tidyverse, warn.conflicts = FALSE)
 match.fun("mean")
 #> function (x, ...) 
 #> UseMethod("mean")
-#> <bytecode: 0x5610be5e25a8>
+#> <bytecode: 0x55734d1505a8>
 #> <environment: namespace:base>
 ```
 
@@ -347,7 +347,7 @@ names(primitives)
 mean
 #> function (x, ...) 
 #> UseMethod("mean")
-#> <bytecode: 0x5610be5e25a8>
+#> <bytecode: 0x55734d1505a8>
 #> <environment: namespace:base>
 
 # other package function
@@ -356,7 +356,7 @@ purrr::map
 #> {
 #>     map_("list", .x, .f, ..., .progress = .progress)
 #> }
-#> <bytecode: 0x5610c05c5a28>
+#> <bytecode: 0x55734f135980>
 #> <environment: namespace:purrr>
 ```
 
@@ -559,7 +559,7 @@ f2 <- function(x = z) {
 }
 
 f2()
-#> [1] "0x5610bed32bc0" "0x5610bed32bc0"
+#> [1] "0x55734d8a0d10" "0x55734d8a0d10"
 #> [1] 100
 ```
 
@@ -654,7 +654,7 @@ show_time <- function(x = stop("Error!")) {
 }
 
 show_time()
-#> [1] "2023-12-31 00:42:55 UTC"
+#> [1] "2024-01-07 00:43:46 UTC"
 ```
 
 **A5.** Let's take this step-by-step.
@@ -874,7 +874,7 @@ withr::with_dir
 #>     on.exit(setwd(old))
 #>     force(code)
 #> }
-#> <bytecode: 0x5610c7d2faa0>
+#> <bytecode: 0x55735689cc20>
 #> <environment: namespace:withr>
 ```
 
@@ -958,7 +958,7 @@ capture.output
 #>         invisible(NULL)
 #>     else rval
 #> }
-#> <bytecode: 0x5610c81bfaa0>
+#> <bytecode: 0x557356d2cc20>
 #> <environment: namespace:utils>
 ```
 
@@ -1298,7 +1298,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Names-values.md b/Names-values.md
index 472e65dd..a9337778 100644
--- a/Names-values.md
+++ b/Names-values.md
@@ -29,7 +29,7 @@ d <- 1:10
 ```r
 obj_addrs <- obj_addrs(list(a, b, c))
 unique(obj_addrs)
-#> [1] "0x556dced402c8"
+#> [1] "0x56426e6d12c8"
 ```
 
 Except `d`, which is a different object, even if it has the same value as `a`, `b`, and `c`:
@@ -37,7 +37,7 @@ Except `d`, which is a different object, even if it has the same value as `a`, `
 
 ```r
 obj_addr(d)
-#> [1] "0x556dce9f9fc8"
+#> [1] "0x56426e38afc8"
 ```
 
 ---
@@ -66,7 +66,7 @@ obj_addrs <- obj_addrs(list(
 ))
 
 unique(obj_addrs)
-#> [1] "0x556dcc722500"
+#> [1] "0x56426c0b3500"
 ```
 
 ---
@@ -124,7 +124,7 @@ And as the docs mention (emphasis mine):
 x <- 1:10
 
 tracemem(x)
-#> [1] "<0x556dd0825628>"
+#> [1] "<0x5642701b6628>"
 
 x <- x + 1
 
@@ -136,10 +136,10 @@ But since the object created in memory by `1:10` is not assigned a name, it can'
 
 ```r
 obj_addr(1:10)
-#> [1] "0x556dd02fa0f8"
+#> [1] "0x56426fc8b0f8"
 
 tracemem(1:10)
-#> [1] "<0x556dd03667a0>"
+#> [1] "<0x56426fcf77a0>"
 ```
 
 ---
@@ -163,11 +163,11 @@ x <- c(1L, 2L, 3L)
 typeof(x)
 #> [1] "integer"
 tracemem(x)
-#> [1] "<0x556dd0baaad8>"
+#> [1] "<0x56427053bad8>"
 
 x[[3]] <- 4
-#> tracemem[0x556dd0baaad8 -> 0x556dd0cd6558]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local 
-#> tracemem[0x556dd0cd6558 -> 0x556dd0ce41d8]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local
+#> tracemem[0x56427053bad8 -> 0x564270667558]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local 
+#> tracemem[0x564270667558 -> 0x564270675188]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local
 untracemem(x)
 
 typeof(x)
@@ -182,10 +182,10 @@ x <- c(1L, 2L, 3L)
 typeof(x)
 #> [1] "integer"
 tracemem(x)
-#> [1] "<0x556dd11a3198>"
+#> [1] "<0x564270b34198>"
 
 x[[3]] <- 4L
-#> tracemem[0x556dd11a3198 -> 0x556dd12cee18]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local
+#> tracemem[0x564270b34198 -> 0x564270c5fe18]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local
 untracemem(x)
 
 typeof(x)
@@ -216,20 +216,20 @@ b <- list(a, a)
 c <- list(b, a, 1:10)
 
 ref(a)
-#> [1:0x556dd198acc8] <int>
+#> [1:0x56427131bcc8] <int>
 
 ref(b)
-#> █ [1:0x556dd1eb64c8] <list> 
-#> ├─[2:0x556dd198acc8] <int> 
-#> └─[2:0x556dd198acc8]
+#> █ [1:0x5642718493e8] <list> 
+#> ├─[2:0x56427131bcc8] <int> 
+#> └─[2:0x56427131bcc8]
 
 ref(c)
-#> █ [1:0x556dd1f3ee48] <list> 
-#> ├─█ [2:0x556dd1eb64c8] <list> 
-#> │ ├─[3:0x556dd198acc8] <int> 
-#> │ └─[3:0x556dd198acc8] 
-#> ├─[3:0x556dd198acc8] 
-#> └─[4:0x556dd1f41178] <int>
+#> █ [1:0x5642718cfda8] <list> 
+#> ├─█ [2:0x5642718493e8] <list> 
+#> │ ├─[3:0x56427131bcc8] <int> 
+#> │ └─[3:0x56427131bcc8] 
+#> ├─[3:0x56427131bcc8] 
+#> └─[4:0x5642718d2178] <int>
 ```
 
 Here is what we learn:
@@ -259,7 +259,7 @@ x
 #> [[1]]
 #>  [1]  1  2  3  4  5  6  7  8  9 10
 obj_addr(x)
-#> [1] "0x556dcfcf8e40"
+#> [1] "0x56426f689e78"
 
 x[[2]] <- x
 x
@@ -270,13 +270,13 @@ x
 #> [[2]][[1]]
 #>  [1]  1  2  3  4  5  6  7  8  9 10
 obj_addr(x)
-#> [1] "0x556dd04f3148"
+#> [1] "0x56426fe84148"
 
 ref(x)
-#> █ [1:0x556dd04f3148] <list> 
-#> ├─[2:0x556dcf554b38] <int> 
-#> └─█ [3:0x556dcfcf8e40] <list> 
-#>   └─[2:0x556dcf554b38]
+#> █ [1:0x56426fe84148] <list> 
+#> ├─[2:0x56426eee5b38] <int> 
+#> └─█ [3:0x56426f689e78] <list> 
+#>   └─[2:0x56426eee5b38]
 ```
 
 I don't have access to OmniGraffle software, so I am including here the figure from the [official solution manual](https://advanced-r-solutions.rbind.io/names-and-values.html#copy-on-modify):
@@ -425,16 +425,16 @@ x[[1]] <- x
 x <- list()
 
 obj_addr(x)
-#> [1] "0x556dd095be30"
+#> [1] "0x5642702ece68"
 
 tracemem(x)
-#> [1] "<0x556dd095be30>"
+#> [1] "<0x5642702ece68>"
 
 x[[1]] <- x
-#> tracemem[0x556dd095be30 -> 0x556dd02523e8]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local
+#> tracemem[0x5642702ece68 -> 0x56426fbe3420]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local
 
 obj_addr(x[[1]])
-#> [1] "0x556dd095be30"
+#> [1] "0x5642702ece68"
 ```
 
 ---
@@ -552,7 +552,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Names-values_files/figure-html/Names-values-31-1.png b/Names-values_files/figure-html/Names-values-31-1.png
index 9030d80512412889c99506c4e7ccd4c9497dfcfb..8fe9b73c3b569742a8b29b11a77d7ade07469854 100644
GIT binary patch
literal 62664
zcmce;1yq%7yDmBr6_pf_mXMMVP`Z(BL|Rg$yStQbX#@o62I-iT(%nqDL%O^6Gx@%M
zt?%Dw?{n7J<E%Xg4iM)1zE9k7-Pe6RGw_|P80r&(ClCk(RYF|&Jp}SV00Kd5dUOx`
z#`V5%82I+sN?gqr0(tHa|A!F1vab(;kU}Jc-zqvKZOypaVV}C+-uCxw^=z3`6j1ow
zul-z&_Zjk$>?f+j^QrTSP%Q5Kl|-)IDu*HLl$LgL4Vy66$fK|PMsf`1&$+kHxo*Cv
zOpT~q6`De|v(MRN1HOOhZN((TcsqSxT-<BGJTu0@z|nB^w0%5Z#AyGXF5}SQzJK_c
z(Xz8C#rBRV|KW`rMZ5KVFi;4j!`!A4`R*r3N(clU{*wSPX%iy+hp$MF_Tb+k%Lw;)
z?tXCg`v3PWdwNi>Z<&pabJj`H$?{60I!rHpm{7=x&<qbo&ZXp|7Z(@by~GY~E!C9r
z*gJRgXw-i66wM#Cxx<0;{4MNQfb{EmyU*3LNB8YCaRpJ(&_2Y*Y}*`o&+Wok%&n}L
zrhJz}BO@`@VxPCK@9y?HymBVTLz%4Pee))ATkyb}zMGPgQZPW39_NYg8*Kq&W8+o1
zCpu4FBmUoBhX1bhma9Y-?}qL+Ty|^U9E{{D)E2?U#`ul~gzW9@vBF-XEM>-NO;(t~
z_V)JDM1qxzwXWyA(P@JsW4<83w>$nr4e|V|?u)^2k8;~JSu$Z#i~!*muBSUQRhDzw
zSFPI9_At%B=8KDqO0x-@gJlUaVJ18tyFUY|LjE7r)f=z3Es^ehE>OB#&L-^6WA#5(
z7BjWBYkj4qrP;|Y8{qLS+Z8Z2i|IcceR7@`!wn4$Eg$)B_QU0pdDOJEZ~Jer`WYL}
z7U6%}*x0D7tXxa+<O^=SGfyTZ8?L6rtn-xH^OT^ptKPY*-o@5nN{L8{^<>wbiJu?6
zjT1=R_CC9+y1HKe@<u{NMs7U(z8f7I`-F^^mX^(Ow%}zKSRL1e-@a0}{b5UZei4I~
z58*TSGW|MSHPCoE>$+fdwwh>}%<X7#aMl(|xLbcZt5#-Eb88pnrl6p3Fmtf7veF&<
z%GA_!3M?e=>Gaag9GF>dC?PuzHulw)AwROO;Aa7y^`T71Id?~Jjiu%8SfP42G4JHm
z6u0GUUBBneq2sKx;emuu>Oi7pLyhHJLz!X!g<C^VP>}0R&04IwMFhoqt@b1M$eKdp
zh7ml#76=A__d5S}lDK;<rclc!VF<FYvB`22l$4Yd7WQ77{SJH1jYh(!@$TI_6&2X<
z@bL8X^w7}3pY-sCgD$!(<FT<ZFvcVv=Y#d^lrekvW<S&&mE+S>%gJ(ME30?I<(#&w
zKFuGidT$o|Nj&y`A!T0?BEbit^Cbyw&#*7)sbJ4H<~3kgQ&pDyE=M_OFOhK>uCC{9
zua`N`laq<bd|VIR-Z*AJq1?`Udkf8kY-Uc2fz)=I>?ofOS9`a&xBGcdH8eG692+lZ
zbeQ5**>+VLFO*(1HJ&X7dwY9V&AMzZh4X8@dq?VX)@$i;cr+~6cy-VX_N6$>7)us7
zvd!7vf>dLAYcQ@W4}Se>;BaSYL7oYf-1o-mUgS4dQC+5i+r-!+3ov?_9T5ekQR^f#
zSR{$wd3rLVk`fll(t(~@EJa{rSALU-R0@SeU7NV}VxP<n8-m97q_CIG&CP-V<YHlU
zCzGaG#<kwS_OH*uhJ{+XEroIKSc!>>iTPs06|rfns=oBWBP5g}rl+O7SY>a_-K@p<
zy?N}hW1C`XX{my(Kb$4=OXuCA)R$d|=sZ7x4}d+UYS`$AAQud{+-;nb*%b`n-mNRE
zso}(aa(sNOq^Ky(alF6OPU5;%0%mAFRas-$$s;q&;GT{Uw{Ut{rYU%{f9}SY!Z@Y0
z93@PLF|@{4N$x)3GTW&gNnt%$gsrkA>C8jSK;V`Cbr@k5b(^tnPte`{I_0V}Wn^p&
z4F#p{`edr{td;R*pXAoMGm=tCNeLbfseGzg*s4ErM_g%k$1$F|)CYR@CG)C3l2=lS
z;mPDT>?0!c(N_z6%HCanxdW}Js0i$AqfT+NdD5r`w+;vcc{XC5Yuw<~`Qh!V<xN{?
zZ%j;#_U#2sOiax4a%v84eHK1w_wNY_2}OwO{f>j$?qV6NQcKTNpMJ!7_Ds`dR_}Di
zaSq+{ths8=!?iu^d1z=TSQ9#xB6>zfIVERf<Mr#)InVWK^C^6-Ypqy)$Hkda`qS@d
zZZ9Mv?KHzh(<{vlp|Mx<`%cqg-m3JmSADn)&F1Vn+#yG~c^iCB<Fh2@-X`!6ho`;7
zn(b)Q|5Llu@NnLe**zhPGpyM(gr%nVwGT$6xmfyO5AIi<@CT$MD|TLBDL_2ff$qvm
zyFhRD&~NvuK#&7_G~*c?9gTvD>f5~VSbHZ78y8pGl@eI{+0AKVqy6R>2wY(Qz?aOU
z5~IO1F=A)E<BiccVr@;$>+id6YeV}tH#bt7bw!Fw_K0ruINf`>hjD2yYm3BQJk<)z
z=v+IjL6~Kc(K^nOfY~c4fpzRmGk$Xa-j40B@o~Ietjx~cQ>n-nZ_f)BnJ9`*Ya)gZ
z7hit;FrU&WVd*U0dg#<$T52=+XiuP_dmW;!tc-)^PrXI97KTj5piw=~5k?`gb9LU-
z)a29r5DBTZoxM{V3ZnWe^-7=S1w<q?A}+g}nws;LK&(V|%h^(WDIuZe=p;Dcfp{s}
zxRA{O_d_P@`_9e&@#?oPvS2_^(34ZXs)PgrvX-0cOKPIA$w^wDoWjETwl|CWXEnBK
zF0QV=7{0!~o3P@!YU?F<e4ny_$~@Zd<KER+x+S^0wOsTt!VPoEc?ia0T;tV>kMc6y
zhy(*DrDCgZHL5JWW26CZsp++Jaypy;<(-f<<8iqSQ*Hb62ZT7j3*e;WHA)hZ8Ch9W
zWWqeejcvF5;J#VB&Kwy;CZUb{2>~JQ2ZIo%Zy=fl<GlJ%X{IJ7);@PVGgqVD)v8;f
zqU~-EcCEI^b6rZ?23@^Gw{<8}3JD3x^B^jvIy557_+}o7KcH2&@zk}^aXazz1Mgp8
zZf&|HhAHlBB{VY=WrlN3OQ>YR!0T9VZb4FILgfsbhUHi89A7WKy1l)`y0z_h^1R8a
zxOBff>NHb%NUFVJhhB%V2+w@C8^t#pz|L<$R+f~KLVEa+zI(|2dcU1SPV2r02q<!V
z=gTMCQ|cc+n3$MwadD*$8i|Osf)F}NL$g%UPJB6g+E4|o4jwP->ZVHzW@ok9t(A*3
z&L{dk93*AIfV}Dq)g-lxZL$=0b`I>4qZuZsBp!WC*}w?4=x{TJWPR2^-veRc_ALDN
zY|^xtm7U#RFtw|z3+y;Q01{Zrm2i__?&mq|qAM;hFDK&nsLROEyS_L^=RNugeRfr$
zUhi_OAjf-rv03bBXLodb{0s+&LA&9q2)a`Xf+$tNSAY+^zEQqeUEg_M3+0*7Jg6al
zy)KJiYSv3rmA{5N@+!^*ckuJ$`3><cooI>c>udam`MkV`-fzi#KH{0}M^cH%2L%TU
zrD9BVWWrsgo_4`1-h6o|W7CBcqvPt6;0@eWW~D$zF(Z$oYMd)JPJmJ)56QoO|DwBe
zi38kmqKMM6da5HdBt#GTRXXZ7dGZd!xpel>Oy*HCRupeMH}NP#imB45YRssze#O#p
zXQ>9b*az%=>bsXqO2I=zLp>?jP1s$>_)fsaNH5LQiT^0n_B+*buM`a*>@Dw4(d?-o
z(iV1^<ma^|t1npPQ+d&dc)`;t6<O=2gLB&;?%vrMO8h96a%3_h&yRI^Z3&G7c?G40
z{ZAvK7q}35i0;F0Jx?BK++_%j*<p~Ai<`FpAUQZ-5!X{mo4@pIRJr$_p)%8w{=N9r
zJLDceIAmg^q6DQM9MD>Fs{dAu1^MguA0r7SW0=3cH!>PZk0hKdta<ju&jB(a3Lj6?
zQJjGg^f`TLzR6+?D@!?I$J@IHOUn(7=h#%LgI(&g>soVRm>aO@)y6AM-o-v+)}n%B
z1jB0sul1!M!pVFSqVM2_=b!I><rX_VFbhUnAp~t4NKk((#O+mg_b7qn2;J8;?1KZd
zH27^9{(}Q(ItgU(+Jh9Gg1ww>%|$rK&%so-MoG<FWwgZ<hoj+3NZexx>?r~<SP4!^
zG}KPbY+hbXNyoV9xwy1c&9d&U!n_10nL<saZKshQ*kj1Iw_g1(X_nY6C{tb^8ulyf
z_(oC$(#2AF^cD#&!sy`@LQBOW2>e$GnyesdX5GNHmi!!6?EHtEB?$<>AHA$fohix-
zR8gKfq>Un+v_1E4XQuKprQnRJj|kjsu_hue0B*#989K>pn56LhBp2;+`Ei7<zeBEG
zWt~oHxlTmg4`tevxCH_PQe2FAW?E3(oP`9Xow@Om3De^+%P=>CVU^P3Vq*)xcxpPd
z%B3MV%px=FgOOznB8gdC6bVREu+8VspRc^-kkKlJE$u%dYinyWo9er(g6u`)!T(A%
z#b@^}uXirDzIitEY4fG@TN9DD64=CH5l8ep1otV}z>IU(r1Sh7vfYvuc;|AHEg4lQ
zo*Xo!YDQ8t&N2G@q^HCDcI6(lnfLQ2(;@W(Cf{mbzI5r~p1oj@ab_TvYGb23HGSei
zY1m+->GGvU2^11G@MN_FuvKl5YrR%8)1y7(_e!Ivds)T@s~-}uMI!XSHFh06!^Rem
zw2QH}%NnwusoZ8MR;@+dj#1hVGI1p1YL7b7Wxe%eD5ZxY@LdP9Rlu_pIot^5FMgV@
z{B~ZE&YqrGS?8e?p=u=2bMrrWfLvHwV|MZol8Z4952^YYs){?j#YyW&Ab$1RN$hu8
z8bXcyPlifSa34dQgYIn5^hKy&j4ki_?f`MV8tHiIb4U@<ofWC{m>W4>^aT%7BfIv?
zxrC$}JBR?QgEYk~eZ&S<W?Y?<dYyqC_=t=_)30wIw=Lgf3+}{z;63pl{_+gjAd!_F
zW?+Tc*;0JNG_6Zf#B8$z4LYgM>AS2VAWXJ@4b-A8^;*$?$kk$~cNP<MdzNrxYf>cL
zn$7XPm-Jl;*kgygJuvI|ap)6+iBWNpHFh}0Z$gwOQ#j=0-wJUd1~1`dcvDZ*$Uu=p
zGG?x$Q8t)|4HA4FI-+0G<ZITV2qyI>7`~Y+DYw(*ZEu_d_Z5)6b17-{_f9RtGn6Uo
zinY-9Y~^%8N*`8tj&O=|?cEySUL}rvcb^aVmPY)eWd43sylsQ_u39TTV4?zHOTP$!
zTlNDNg)j>xKR8p!mUZ7$pZn0@b{_jmoyx6?DFhUA#;_tyfSKXR6hwV*biSOIlFA!^
zX{as&GuqvVE|3EgQ`6~xVo)ayX;SRUhLD7c-Y$^bkl_StUgU!M4}cC(k-UHO6y+iM
zb1rx>3Ca{p02@CGkS@gl%pqNGM=<$3+$65QBG04%Y(?qh{ty_}uo&NiGtX<7C$Ia(
zQAWzeez>RG!Ez_SF_vJYnVFfqyu6;5Ri0C0V<%UK{rvkaSolR)VC76W!lpgnuJbLe
z?*Q}3f(`J^EuaZQ;<FB@ns#TPr)&JT&P;eHzJgeMqqDOfd-K13|Nad?8~2WzD@eDi
zy@`!C$Hk2R3WF*;ZVi+~3sOYwuVscej}3BMKuBa!iyeY(vX#4X$w@|ZWQg4@;yaJ~
zXA27phRM!ZIXQn){fS*pCJd?FPQappI*yf<a_*J3mR5t=L@5A|nox%s+kRfb0FcRd
zbF-5ttF1r1efz85<D?AmBaiOKGHAL0#LUjhdVF#MO6=o5gX#74^`OSOyjtV;T>I`w
zl|qUAr2c$0@$zJw+x4WNe=R>Jr+O{f)uLtvCvZ&)0<kiu7>@3D$Od6usoxRqAqL4;
zkM?el<Zf9wII5iwSB+R))(z#7ILLgA%*>LQX2-|J$HrDY*ggV)?vGBw@bYC&e*WhA
zy0L|Yg|V^o(V&>N^UAYf7El}`KY1dX%ySOvx6RsykIfHH^Dizh+07@xopX#QlQ?aE
zFZ`S`xZP_;=RceGhOH%nGJF^xbsOSE7lVXw1BZV_noz#OZ~b6tYNEEWj*gBVkpWOd
z9RekvnVISI1_PM=`LP*Gjwf}q_9GrZB^H*Jo}hdNyg|d|%r1Zn-K8E!=_K3JHHZI4
zRi>0?ol{AB)SziH$9MYfygUQIQFYJMkv8+^*2RwRyp_lH=p9<tVo_51-l<i(^d3Ys
zlAOGRGrTzJ?otYgD<v)8P5VJOFo>$<ieb*q>{ECTBFzG0IbGgX9s$&)fwxM~iAV4e
zjf!@;L}ZI;$Pwkv^s*&Z(`aY!+81kJzy18TZu<HwDL3mWSzHEoc8ZUz*G&!qjus3k
zDG9C=eD&W_d`it!nkTx(_6)OZ53v(i;d?alPf;$Dr1Y;HzUH=jV9=WovyFKWemNd?
zq`&D*?I{t-5`+5+rvlo~k<GOQ&xv7*+~rcA)!uKvaNd9K{I)RKA9WifT$=FqW)VcI
zXMo31k6qazJdN`_kF!`#j)s;@XO{dzoysz%LI+rp$9cOJU}<Esyu7^CzVC^6cDVIr
z^mMNm4Qc?XQdL>W_Il!MIZ94zg&o*U*6Ze<P}!+$uB74TC!JU?_?(5HEZNdpKi?%Z
zzSAlfrlsw3ih+1tMS<#KR7IQDVQW0D;V6^PqI%(t|ATv<Q&UqrC{4?2|E(z1EJx|d
z+#CW`sXPDG0{R@NG@f%i0KR1m1}oO6DXpvHQIY^$P&m&)N9IzjzUGHDd&=(8%ULlV
zV#m4lChACv&duY5EaNGMO4Cks;_&u*bS8pOs>Va#?swQC!{sRaq|2-_!_%j7=Tp)t
zd3ix;fQ|wzr5|k5UK>%PcW$-g=ZLeUT{7nP><nApD7JEB87y16q=W})2A^|sE9z(W
z#d_-8SqAK9tdk^EXnto-!|((jO;u7$ztMll-i2_^;D~q?h()dCd3_9@Yq0j!;z~F@
z4Glg4!4v;xthSkm#YGw?DOW%P&E4Ld0;+CuZ0zs4^)VR;=|>&p;RRat7eh;Qw6r&%
zKDmO|tsbX-w;}RMtWW2e-8QZ=Uk1T8Sy8;Jc!rh_&oL2_bG}k(!dPVp2?^7Y_+;Q&
z-owl+<TP+-`Sg-fV(}Nhdu@?a<T8is2i)GumCt;R{theBhH#wHWZ9cvKHWRJsvOXQ
zwjE2E)f(Qmv_D?iTdbC#r{;iIr>~|~V|3q%bZ*7MYj=IU#QYM3JvedTBUe#b2_j{^
zU;v;QP4*XC|If<X<u^3MQT-_j)AmbHD8;7(V_R-jJ%B{%ObfbbT~CtgOpTBr%LUV@
zdweN?$#cefh}vv*YGeDUhI?H_L7~iXclP@7bk1SQyb}pn=@T*&Q`6#-NGG0I7>v5x
zMomra<Hv=IV^9QZ-x_+J7<%5GdN#UWo&Y+~AE1^dJIc^d+B>WdK#*IJSi;f6ui=hH
zRjZ64D*cX6R3AQrSZOX>cv*t-&%`8w>FgI0$rUI`RRJFhvNERk=2)TYW?@xAVxr8j
zbb&oP50Bg4{9Bn}i^l5{(*iwEjtz>DC@CESH;|!UW{`NVpfC&y%B83xX#l_uls|Bn
z=jUmj?4xZfv4-u`AK?M-Qr>xZc2uZZq6;WZu`r_ZPC1WK<6%&PNe^HBLifN3ZsogP
zm(#v9(D|ifwKH9_*7w~Nq&9Xy@-kR>o@`C9&pHV`lGoPee=r9g((z&{7U^PZNo;HQ
zxkLN@@6=xs;Wk&kW79zr1d|Amj(~NHd0&h~HglJQPHMnA&kha2M>#n7G*lFrVZ+5}
zv25q<?SHM>eIpUnx;!f^<eIIA7gZD{F2oHXb(D^R@Ppk0nRqouMGJHRI6bZat=b7T
zhQ#F$nPcVEWFfG6kak_Jk73+gTmVu$K}L>a)b@aBI?h<PKYy~@fC%|Y4Js)BF<*!X
zL0Pk>^9m|K0mX_;G&fQg3&8B8#Q&mrjoT^rm!JuHuc1No#CKejK?dOVZKcQl&G0%^
z6NdO%V3h>xK!p^;N`(`jh)9pHhUNU)14Id4HG?;tEEJC4Pmb(pW;AU)wLvE7WuQ~R
zo%=zb8@0;#hlIG(?YlGWU;U2X&!mF?H{A}ID0s!{74B9i1huJc9=88Lkmy#mEk!sf
zB`K+=wrb7_5$JmX@VHl<qrA}NH`&glD~lh&1d~$TZ!{<%5Lf^lUmrxW5GeN@Ye)#W
z7A_s?18@h)LV@GS<xo4qNxJ;SXQLJtjy!c>N2RfNG7%tNL<Un`zK31kWNV<^!_JaH
zAu=u_PC}Poe3d6O*jy1I(o%42{Z;+{0q#%lrvhxkGz`bo0w<crLaqe{pLP5JVn-2z
z*RK=S@<Oh2z7pZqEWJfPz*`|?@S>O079dc58j%-K#4BgWOG>PjP$6yrFd%MsZLcD7
z8k`tt|H~Q8<Yai|j~y=OGJZE6ho3K-Meaeq;==Eu6BNo`p0e@IrC+9Yf)IYW1`QmP
z$KO+#9?;TUdRd2t0GWTX*YX$wDS86`yB-dF88W1TL|2jl_>I$t>QySKJul9w%wHP@
zu(e4M4$r;1BmR?fh$&6glrI>1Q<ciArLcMChR=uv>gsU#I#Ut+J36Pldjv{<1+cI*
zKZlCL=m4;g*mu8$SIJ)Z8)<$gMY5e28Ga`T4^9en+PVQ8CxJhc$vb0sV8=-j>Q<zr
z^hrrsJ!Uj44uE}#%{};)#=X|dH^`Ew_;QZfZduIrekbQ)X?oRQXZHf(OCil2n|X9B
zS-)1WQ}laD3=SB**5L(zlO5@|EVoKmK_Mu4i%2R@Zt6^4^6pRIYY=B&@OOcRwAAMv
z-11ZVR#nJe`EU}1)N{O&XW$fU9L(66gJoS__VZkVO6~~50sRNe10*9Oq|%qG9MnBV
zrY%OSA&8>otTuX%OL_XWvD>wfWlH!CaLNR=FCVU9J0{darQQsg8=YQWM`m$)d*d_U
z#V@2)4nMtqSyCn1bN!UuWq;<iYjZlVFBN#073iy}PQ5PIQ;Q;s2Af#c^wHmeR7#Zo
z^2bLarXIwn>8!d*0c{1H7it<YwKx^q(y>T7T<`__iuFI!D@_6Vw|y+?<f5i6W~?Ef
z{G7iDYwhf_oYkT7)&E-4Tw=_jqr(zx(Jcv4LjL<6loF`hzzb+b(Y7CWm}+IN%o6ff
z2@VXP>1lM7PR{V11WcIViJ>yDpj`U2>>S3pwA~u$tMv&~0GtD#_qV8jur<p1F}$v5
z+~tV)Y<g;9MYJWwul8d07zn*K{U9mwwMBq==xUcE3rcgU8GggN16@kUqTxZv7E>C1
zOfruJngwJ&0B*cLm3X1HxGN8{=iZ$Z6Kl5V#~*Z-+~IoleH&;0jH11xlo>jbo4h;1
zP`L(s;+cve_;m3Du^*r*C?t5!vgINvP8&7Kk2*_vc#Fm-jLbYKN-gco=8Q6Cm}I{V
zDvHrYmbc_H9=dG0U4-6zNxZ(%m8<p49cs611DBwJenn4|({{W>*phfp8u03Q6#u`8
z6-EDHTZ17#hg4%5k(hFwRKTxjK*fP&eSOPoP)<o?f&aKFB!n&8r(Z;Y$|0O$=H0xy
zf@(F@yf)U$x{`t2wO1YxH9@cb9Kg1tXj>^MdH?!>@tY~ciX*90VU6)$G$L{joN)(%
zmQ0v~hz*5&YGoPe--5BFE&&lyIhAH4?<R=(P+6&52cuKm54Ln~fJ2>L0<;4nb$Lv5
zbVunHhpeZkCxCo}zjOvFw@=-id_gR_(*i_9M3~_oboR9_7WysBKAU@kvo3cnz8H}q
zQad(({io>cVG4{}1dFMZJ}x@?5t(j@>PwdM^cH)i4<8cFqwd8EP74BOQ;Pf-g_h}r
z80_ajN*9UcYD@73L+;zSlEjDj=y8?HIxf#_hz%y(tLaGp0V!;X#4^THXA%@Fl$58G
zSoaeigcXwl`j9zng`<61e+^|R2j4UlPbhb7{JOMG6&AZun>cQH<jji6?d2u_{B|Cz
zBbBUXvmEdq>pPK`m+!X*iYD||l$B)avc`(?9C!-WA8BY?w{zZD>sp4!BcYP(W}8XU
zMz5@B)&I(7cdAF_+r`vW5Z{6vV8BNwwX6lm!T6;lamg*D7k&{g_z_+&7>C=jE>UY4
zu*MOV`QaunC~aiKKTPw@n58@}R=-+YNA%m*admRPZ(Jov&-8PPC^0YJ7t87(F<rxR
z1asU!Qq!k0_IVflify)Lvrs&-9wX?-9rh5T4d`Mp_R2#rD||l?5aQK9_$V$2Qc;A~
z%<S*kJ=VY_kG<v|w!;mwgnPAdHedDhlJQ<BBer%Q36<A3e<VH-^$h=C6Izk~$8J71
zrNH{ZwZx^5-yTDlpMx1Qp-PnGOuK4v4zI8C(4?$KE|y53r>8!zo)wH8-8m@{>rrmz
zFo~zx4Q%8nVSjdJP{u?7(ST>Hc&eiER9@vW0bu@2qjyGR-NF<e|N7Lsn<y!19CqPG
z*;9hpr2iDGf%6u~0u9Jf#7jmAnt*?`s!s}Sn<aVd$<j{zy03i+6lfI2B}J@G{E}OA
z^)8?34NDav9gcMY2$1DxQ_6iQ*C0Xy5=`19dI@09(eWL$Z9pw4bbbHih;`{vmd?)Q
z92|u4n_MC|67!O_7}jpF1+hW}X~N5mi35ZGXa73K>L%AK8LmcZiLotChhw+=Qd%vw
zr5->83F}Ns%5L?3iuf#Xso;Ylm=jsfL{q|kR;M+~hv=0HPwLAonNTUyS;{by5;OHj
z5kwD&T0iyY|Ag~D1o>%Q`z~cl6inY$!)$v(h(krEP|H|#J@5ii+8E=;aWCuWSbJr*
z*e~h`0s-qs2rP({S}oS)DqyTseP>6KxFILCzP3%?=>=ku(zJCL6uFfU9>1A;UURRk
zT6m}xoo;~eEenftIhKowbM;{c@{!Ij2s&+x_UI7OD)EQ_mmeZir>W6IkL&pKXI^zB
zB>J(;NIriB$@nCw%a8FIg&N!kK}0KS6P^Y(&G(!w2acE79Ask6F5eI~BS5m?uKAU2
zITfKoKitJC1)ESCs~Zicm0;`-p~Wfov&UpD^EzW2hzH6&*0L0oCu0F0vt+l*uxl9X
zb&()?@I?RB9ng=HrRuz643P$~!UN~QHv9b*I$!!tK5d+C#Dl&Za}xISZNGJ@6OaKD
z{wGiInidADtQW5Dj8<jb%N>JUzvLZ+sb0r2X;-ExA5-7u#e-G3E#!ATXt3pHJst8P
z7t6J#eR%$G%=ws*W%QafD)SzshyXs02cS#fV%1n_DQA-oO?UltoS$ZD`4Nj_9xGJ!
zWVdI(#iNIbv5dnfaU2V|HO|@T!6S%spuZ(VK(y2QdSe!S6wx)R%C;(4Qa9A}6KZIf
zSg)?R00Xj`GIR>qsygaY&POeS?ds*0hPE}j<Cov)upw46a6iXi;OJV9nw;fL4*T?C
zy-zE@l*?t<@9a0_m-_V2!Ki9u5&4o^rpRuTPk<R@kpVN9zlfxl;+)R%tX9||w6t@?
zZq3UNFH*ORiPbM*Bh5|<ZdFHZme~@MKD1+r-4GXv#MzvDFQdLOO5iqY&!_#`b_>FU
z2!Avz7=)P9sy~-}*byZyKNRM}3pJjbYph{EELA1^iX4e`*&x0wI4x<PAr^_V{=xe+
zo%19{t6m_MS&bF~?+PG56y(-<vv({{%AbEN*cWv)m37`~tJuF#LPIwCm<8P9)u#v#
zx|x>Q`9l|DW&|(%@VIvnuiPaLFj+9=uH3Nww-TkRFi}F-;GZ!3mz*W+VlNOu-2O)j
zV1#%D!-Ii>K#=In^O=>;H!VNXoeFAKR%NOt%cRP{P+!L`5!uFlXAVg>C}7n^$xQxq
z+I^s_sh+nPI5We7$qkt=h5NMW3v~aH>t2&ht<C_7htEAfmVcBfw=Ud!rdPr??-I=2
zw;zt>YTCkMg!hzt<^8^IY29=_``OXM;wk`w(oEqSU_}YPQGP}ee}2W*36TX><#)XD
zt~C)tSZiD%#+I({IvntUVDdTOBL@wZt3X8D`cmNaL(RCyTVMe2<fB1WHRujQ%UtU@
z`$5r)%Mi#`RE9&}qov<8lK6G}R1C!R7GlNd%-UG-Evb60&!A}VJs@pL;j-T-teWL=
z+M6F37}%O9Gc_?GcAEdJ-yJK%(euBdv6*fWAqrw|WGc)X$LgdA&?p~1mkBOZeXi~r
z%u5&FI_Z^*Sy?diDPk<Zx+SENaMhzZAgh6{L0sN~v;-0Ih7yK&1;jWvfav~JN>3)N
zrZxe(?Vwl8ceU5df053AvBZC~)bi2Y-Ce!XjMMob1LFlA3h^ENEmT-oICpbWg@|y1
z|Io{n5*;<~6?D}5GDDUAwGmRVE~+W`A}VGNu0VhpwUl}qCKNthvf6E$9UMFve9e9Q
zYlT?O+tMxd%;rS+K1A=GEdQpDU}|=X$7E>%=(%s3YF~bkkO<UQvjn6-<5hPfT<k)p
zT%cNE663lY?s>HoMJ)$dE>A%B)4qDOy}o`v)p$Mimlo?LkUF_z#}Q;08PFG0cgSy_
zq&BwGZg}5egJX(V)b~@LgiXfLv<hS~gLq@A6VVjQF#p6$2t*CezQJi&76MJcR%FQ}
zPA4ur6FL8l+rY%cv^S=1c@3CrM?|-O({5ECZa;hW3~;+Q7qG?~AjI%E@77LcHth4c
zEnDKqGIsg)=AM@#y%ev9@U>O8O{J^bJ9*vUI8oTNo{F+z@tgS|U;6x+?N8zHg?`o}
zObWADj4eDe>g%)5b`RM=b{!5wf^^_>KA#UK;b-tU8_DKlO^tB}&1}p1<Je0zvKAx~
zH&P!X6BEe>^@h4Sw~11Hb8~a3!*(rT(tR-C>M+piuCQKe`|{;W%Jr5Z=ni8%qSvfF
z0%Nr>G1=JOo~gCp984E;b#-0qO+1B*iGbumn$P$pw##8J`|0+S<H?p<O23M{yfqL<
zP1|(8!X#{h#A{`A#JlLER9c3!yotoKv$HJlP_o;8g%-NgfwZQ(57t*VUaTkn@|QlE
zY){GD2R1_?Ce=lGz_K$d(?THYK=(ls-Yz92m7kYK6_TGv>tjW@<csltLp4_1eu@V9
zMvZ=w@5--;z7@475xOHnS<+Tfq7)jIX8=OR!F0<`U&(GbUj4pP9{Dqg4+W;Ev*9~I
z+Uujpd^`7PXlP>fCnV}S0QQ@R0d`Y-dr?gNe6HTL3PX1=O=O4P)y2gJ1GIRs&}_wO
zq0lHKbP~S0kIjD6$xebW`(`Gf^_!lZy_vS{9~~WaKZVYrKJ<>&zy2e}KM9<$*yA`~
z*voUj6&Gkbz)p{~lX%ws_#yCUKXLAA6=>r;&R62V-fGDVXB|vE^g^*ES~-$1-zAo7
ze;t9XS<tmHo{*?3wP9o@h6X~lD|tjib<Bo%IK?YT$*uf4EfOe*0S5}QNM{m&m%O=b
zyoF*cHeBx5{<q7=8r&~|xYEgZAIMm2`y-vTisO=R=^_hL6t(E+>FsWJzGeYl-LU41
z)3nKq>9s?mt;f|<xtBph13A+kaV(CD-s%i585sH;9y_lNu~#1|hU1$*_E#7E{fayt
z>L+#)$tD~x5R*aqpn=ub*Yw^E>*wFBp9SY)>K;={43*ay;V(6m3SKBGt-w&4lsd8@
zmI<q-#q}p`gqGu#*DH*k&MRFpJ5KaCSb}G64UgNQAE#ItO<dI*+?;K~?6rASl$3gb
zsOa?{2A8;M5$#7n_2!EK<g6CJ>C(t<JE>EWN_SBqZ@M3ct8zPzbKg$kKbJv~^hKyy
zr_#UP#<bKSo1$+wCJDZ>FxwjVquSl^2dTLku_W5D)wHyHoIQCQ_V9_l1tB1%#GJ9_
z=R9w(fxfh%xVRrE_uzrioBzzaCHvmzrh$QX6qw)kKtJZv=iB+?Vee9UWHO-7j6`D8
zjN1WwdAR`8Y9~{cyW=J4K!4fL-~pxv`T@ClqxSWsr3XXSQ<YJ3xE&CQ$m!idYoe)!
z9tKXlH7%B@N@H(ng+*9Ma4g?xn*JHW0=c2<)(rPneea9hLCpjOszP<fVs%EvQ5-?N
z%%9zn`V;O}t1x=^9g*(7;)xiLZ3hDe-(b#ib4|W^Gs(rpmQZ^jD5AE)sEW16esO(#
z^X1ut2OkbJ_rq-M=j<1GFWhR3YDo9bQsB*nOVijwT!SZ2T8J|P`n@Y41TO36yS%yF
zMJBTVj!^8jLZ`2<--#6cpYpapVbYzwvHt~eTcg|g0bDm!ZCBUW=xJrOH@PVR8XG%>
z^I67PE`dxmclEZIsHkr4Zolvw&`z4oe8a|OfG1kn-HBdX1hwQdU<l?4%U+7FbPTu2
z`{ed6vf#V|4s3m;d?*5mJfa@~T@P<<>bn4|m^)d|2fq}@Aza6#rJXSgzd?f6=)x%E
z$0<T^@ZjPtXSwG>=r&7e=g#x9JuN-;k5OK69VU=~@4Rr}bRD#L$AnjI?n+@i{6xd=
zZ?e>hhZrsmhn#LBo-@D@-GFvi`+8IxXk=j{BRn@}OF&2+N4(URf{ubBjpvx+o}JfP
z4)z`>f`L%5=`*4~+H(^N3qI?`7SJZHGMgA07+|8zhInnh4HDsPG#yTG>_l`Ad@Ckj
zQ0rnl^AKt$i(}dxf@zzCB@tN9svDn+bxAK$Bo#A`0Ey29ZL;vtQ1@Nexy{|(+oNp$
zYanqGk}2_)myzk6t#>`$o%5VBuc8e9&t5~wf3ep<$}X|+w8Fj^<d^6Xz2!6HbX3!{
zP*dMx(%_1ZRCpEJO9tgLOfxMF{JO#l)F(gE=tnDrb+tD}0UCoy19|BU$jA1MjzHi%
z3|EG0to{VDM_@tfy}iAli-+^}IWaMx)83CPW1#4~WQ~0dXmp^`%geji$S<aH2#KP|
z&CN9i&FTFS?%sO>l`q%&O<qBH{ilTAhi5ABF-}%9G?@ARK*EHHjaQ=MXF)5}^FI=?
zrz>x?{Kc@@Yh6E-vFGp5%LSx!I#BfO)*XgnEFx>43Fe9LP1`E~L{nn7J_OXfSIgA=
z6XnKHX)pgn!IyG#lsz*xZ3<Ke@O}koTL1~4!%|!5&f<w-f3o{hD7y~^P@uxjZjrob
zme^`{J+#+%2uPwcK!E`)a_o!mL_W&!Fr5D*lG|!WIee3EQv*PA$hBxnRZr9~qabK?
zysf#MHdY(^#%$<=ai>&RM*uB)xW;=r95lm^Knv4iuAv?zvaGDE^mKB9P-R6$x8r>6
z?~%I1KnA=E_8vrX(AYZd>h6Bk4d8@5ks$~)aHx^J3@V{l6vK*BC#J8UnFH^12L{W}
zN}mm8sc%J^S#pH0ztB_LwjgnZQ<Ld{n$*XO!n!0#DIRGq1bR!2DA2~jWzZ1!bPNT(
ze%`|#=Hj{o&|&We`c}{i7y~`30U`97vK5#G(5?bO-^gZhBLJC$Fqn#}YBg|nAn4l!
zs%RZ{G{^)TrXro>U8+;qx*<cjRk2M~u}qEZ>Kf?B{ET%vO2agibhXvz=@w`(Azn0;
zC5aYPz^e?;zED?2Fr#<__1>}tIqqLgkwEk1-wRA|ao>z6SaYoC9fk`1DwhKHNYPVx
zi7O|PS@WVKbb5ZC6ufDDqcVaUH@J$RO)QXCH%YF3B>06yu)TS1)~jO^z&CNF6m#^i
zjblx9z#j;Wt^37hqe4*<szHzQiI5YbLgr`)ilnGA`s1^wh0fX2!@PViyZ0wsS7ivd
zx;D0y{OY2U#!p3F;4ae4k?e~Pz0fq>^I&Oj7|O?$E<Oze+iVS(<qpF4deXc<<p3|h
z@+JuYHeTg93{BTyW_Y#mRSa-+-KrV<$Inr*FkQX-H)F=9Q6XLhfK?VC{)&V&Qtvu~
zc$snLWazJ{NMx#kt3LWC#{|bzXh#Y#B@rD08=MYGkvLS8C!>7DvVa31H3UT@X*OsD
zP41lXK5rL(UsVUX4C=NDW)d?jrgrVoKZ(}j@y$c=8aD|!bcI0*{|Bq*qU*>W+JXiV
zC<2Wh0fxV-V)or=#-p%sTFJU1zFLQR)@*v$OtQ6>=WEr_cXc1jMSHyco|Y&ugsHyl
z?ymoj=t^0e^%VrG4lMW@yaHT6ZL!)^l~|B4rZ?agD>vHygl;AYfT%(y)<U!<{2tDa
zgp5q#-^X_T#+N*9Ibi>-2@J#oK9C~NX34+gKf0@lL!n!X6D5R+@L~r&e4H8P^KZfE
zVM+r7UR`drQIRnQJXjJtaXx}fto}_2;Ih_AMY1-s!J~*(;5^#eH^Y(2DHK2x5xoYa
zNX;BmTJCNT<NIhryH~H<8%FY>jju_z5iJSu=?H;qsK6&fK!bBk#FbOA1$a&{nKV`V
z<JH<@2b$0w=8iksf5vwa&=c_=Cv(vmPMrwcmH@wD>j(r1%!qR5kguo<5obluAM7?d
z%BMpcnI?ZyjaB#UpO>l<y!kX<imt7sYk73O>J;7c44f^qc?h_aH<Fy5%BB<B#{7C}
z%8PwY`DPYo_6lyTv|<Y%ohP6ClM>0QqrO>enN^%mEQBvV1b%h1HP7?*@_N^3h38A&
zk=WYR6qD6`AAT~q#w@sS-Z{wAW**(PX><l#D;r(y&BGX`b~GSmeCvhewJG&ity`-8
z1QlNE9!`(5lpcuYJ2`;-G-87V<!%S5;Hd3?%^X6$hJ)(fsveNT6%``p<`E(L`c2<|
z3)m^8*WIw~CcS|+YG=%%pwuvOmh6TKUo0x;k@H5DzoaQCK5%j0Y6GJvg7d*y$v}Wx
zxuqOlf$Zw`@lC#)R=KsN+qooJH)v&)@Mu&lx6Jb3;QxG2^|v<w#=P6X3j(^Wz#Vip
zRPR{6I;kB_X1+gcwR3LyN^mmEij}^d(?ArQ64>^eAb8CmBK%OHzr?guF+Y_0gwS0J
z?h24BZ1@1Y^axDU^G_CJjrBNQ|M}xm&_Gsuf&)ahz<}KWuf4;7ZHZU-w5*q@eJeT<
zdahqH`#00km(MtC9|~PWfHWNgat+Rzpyjxp<{+I&?5pYyFwlmv76liY*eS9GS1e%3
zi2ucRD}90U-GGG!feN)2aM+;11HU>y!_LtdI(KSEB&8M5msb#0SVL2pA^S7ONj8YH
zFEFEJmf&^INT7S{3s?l$e}O)rCv1v|xTClJ!;#JQhsGz>3!Gyn=uo5nKXN*$BRrsJ
z{!QjNhGP>;pl~&O8P6k3XD_n9R|beJ%-@yn9ZY{&eKmf?d+HOL@a$h}=6maRxj6r@
zlpE+LJ2ZjwoaupH??wYUuRogjz7v>z@^fej&Dg*nCnLN5tf$>K-#L@V!|@He?AIbO
zq0o13-~89!g6=Q7)51MSE@zm{MMkA#e`e^;U?N)!F2mC;tXNdCZWG7LadHCy16X^a
znYVDMPgwc1bHzF}Qo&;t;IhxCue$R6yOZUy@CLc=fIN17(X2+*EW@jAU2|py|BnsM
zJHL}7B31{)9qD5?nA8G9YAd_eG*?Dd&g=5)X4RhLKfVVVp<lqQa3e&UH*Rpu=ZI-A
z!Ad;1`LmIwy2RDK2EWz25$hXUj;o>hfs?)QjNz>yikYvV@8WbUICcwma2Re%d})7C
zUF9U0u`z(^>iU!qYf~w(!6s&6-l)cRw>LUl`J8%_LE?W$quUk!x*XKp2V^_~V9Ovq
zD0RdSn3g#I4Gb?L#|aa{)R1uuic7z})aAZ?&2zBWG>}{(e&AG-=b^84ew|rns+&c9
zoOb}0WEc;DxT%LyBz@zTyU8JgGe<O@o#LwoH<9vn184?Ia7C}cw%@sBY6LEl|HGvv
zg`HB2s%IA@439lZbhT1pbUt`QA!@k&oj|nFg8?>`If97by}J6y7)~b>2#D~I<|lfJ
zLRhlJF4SC%iYgaxlalY!po;C5L7|$SC-^Al0XPED6e0!`*_=zE34zR#kr^+|lB01>
z1!;nGr{X5036>5vVbXlCbhh319**xQuHCk4o|Uk&s~zXRIZ@e27L&&|d^m^N;7NZ_
zj3`u$?k5`Mfdf{1gYit#c;C?Q0gptg=Ed34^zHNR=cV*(t=jr%i)oplBMGE=aoYXy
z5Sy<oQQF)F$x9rVQQ`v1s<G~Wyeh6Wp@$=YMC`Ob?#kwSHT);}t<e72?aub4Ht*Vp
z=Jd{`VE0+Zxc>9Wx&sr<er0p8nl&Myqe@Orj?b5#_VU4l2k?Va1{5>2A>Ci~Cw5Q8
zUPL(YByWE>@CN6`ijc`rngcRte{ic)41ZNWrG?Jnrltb+b@L=o^m0}GyN<ka9uQ7%
z9{V*f;2)FLf%!4IBdT8_y*N2@EbKm*`K6{!^Xg`#ph<sr{^;C|I9%)zK+s(?uZ@AE
zdTBo-EX?Bo^kLx@vc8%F5j7B*17sur?w!sQTk-253W@h0Kc-^*E`^^B?7_u&bpMU#
zdx%1S`X6Rmf>49A)>=@PA;s!{%lh_LBlOC{lZBx77kxq^TxsRTL)@RM&FpL>U_P<L
zSuH1dAwT&mO|>~ewV9xO^O<14*0Zf+7n$Lt#|RDV#B*GDF4jzTv@U~lBbb`5IYwi*
z^fa{c*kr;-rToZ1yaLz!s<kl4fLX{#Ctc+CSbH?iU3-B7xEN+S#il|p6R7^D%>M4Y
z5>SH2`~3^>@zxB^{AcVlcM=^B8|UM?_$~0~Hq7Uf#IC}h$RniBzx}1YG^+f`&ISi2
zN!@{=4*Tn=buF-MH-DGZ-aX4nYr$a*Wt-7eQ8Xwn|F&XAe`6_>nu3aA-^#EdWW49x
zsLHbGdEfAvt7<J!am)t+&RiPlAAwNY`1fFa$%WFO{pR<k0{SsFTh*|#Pj*|rp9dI4
zfA43|oO->O!%aFCx3-GuPi%#|-T~<K33w<lB&nRb|KYR~ko;gbz`zhVQ_TjCfg|vH
z*a@dIqN6o(aoCZ0$UC}t_g`9B@BK<sdr<xg+(3i*9*@`6@;^Ep#-%W@6tm@I0MO`k
zw{0?RaNM$DdVC~3np+rLt~mP=ba1XSZ%QfBU2paX1R*AUkdhBV^jc$@{g2T3zg{u`
z=3b;^W23iXbDL9Nt-ySh!$ee@kif%O$B5MUt&>_n*1b8BXT2A?)&t#yFTo%bLLum+
zmg4Cj?LC7Vc7>ke&vC<O_QX`!TM1yCpv`2ONQTnxY)EkhHl%7HxDto))afRN<BYoX
z*8>7x_$hN3(3lGlzw1jG{ovp_i5{47u9aWFQ|V%U2WnEcYC*S|5yfv7j{uE&yN4Rt
zT0Ql4A#&dEo-eJ!4*YujVi$yEruKb1qT!Prd0G1qbjiPPT^NWwc$HgKg3bJdcw9#D
zolDR1)_$eUs|0zEO4opprS}9pnjtbxzJGxYz<Qp6l)s~&U2pHn))3eNA<zaYMn?^&
zl&LW-J)&v64G%JOXM2~kw*$;TzyO%RG7cw8MrCDAp(`ahAYre32mM%CY7Lvd>R+b}
zblk%DoyJ3`!WoAu_!$z8+gn94i7C}902=GD0#;bFm@azSZ2rQy3r3rQYp#Jy@~O5J
zSm>P=K)X`#<#O9*heONB;2`cp!P{qOV%z(#7FZ-Mngte`DSyq;9WgHeSn?H~PA3f1
zHulLsbX(LwO#vzL&pd#?;Kp+^sq@G8*TA|Fy|Y`c%B?Qb9)EDKWV{Wqf(}%0xf=O5
zzsa#Y*4;=yA;3Kh>(VKUC0aOq5dlUpDJ5NVt4l%T*p@wwtbe|v_L&jqfd%6cfLm}g
z1X-07`5%4Cl;a9-krF>v@D&D2Y@N)iPgcD;P911O@$1vzSpn@TJ`EzbA4aj-v1LjC
zN<$dD!5z~7<<Qq@qNo32Pwx-z*9K1ZdHRCffo6#u{y{=vrwscX8HB!9G#?g5%T$4m
z3c?4PodTk&vCvmZHmQz()BA}Qm2`QfDicSJ(;*F9ELR6j?;7<f!=Va0x9KncM5(2K
zg(F1=fJQ_6U){-nr_uWgbb$5FO1Gb{h@SyJ_VgF~p0UqKyn_6zDFqw%)u%3ygGr0w
zcVIlwX{<dJI&RXUA(lg#Y^5wAgtZY6JP%I*v#%oeJ8u*sXGBFqZE4S*)+e)|JL1>C
zypsO$F6OWG=LhSr&}$t%OkGj$3h=q``{D{vsMQ+_xh-J6b*bY6x})BKJyW4(*R~g(
zOSVe=xWQk-qwkw4&awbY>i7h|o|^cyykdjIgY;BWJcKhycXU(?R#+w3U8*{HG)C}R
z3lHV`a$wJBXvpg>H#${QI63|pXilU83wRE6WI9hcQokPm@^p3BeMi1z+X!;E!zTqh
z`=}+l4qfwM)sE5|a+ZUBL?M3^X6HI(1rqv%<CR4n$~zw%egt4dMI`$3fSVMU2z_k)
zaz$5Hx~WBFY@ro4Wnzqs_D=L=BdYI<6Q`&zR1%(eFWsss&Rl&xsHVXLvPeK6GfmsR
zes|OB={~Z_AOXm{G#Pd^){2QDQ<sg<_9`?#EdrG9A#IWEaepR-M9La8-m@e+0^J!T
zixFI$zrryY=6I~hJ?+`qtt`BI<t_f1A1WdvRY_hM9@fBwanVS=pgL6RRZDS=a5rLR
zP)s!`m6-1qd!|<I*xzQd%q%<4ORbK_Glgu8vGyi~w%(e*JSzK~TBOSJFPEWi6=@~T
zf)Cl?33w_Um0Il0A^S+W|BUA(p#^NWCZ%^F>+1>!w`jW--lJ>aGI9wv=r0f4f2oCI
zBLM)ZMi2CQ5H(OuS?N<W@732RaSqe?U_|P9Y~0FQK)sLnl)!<_I3$1V=R~y_Rgvw0
zoAeL%M#t%;{kmc{3hG9$El>ql-y1R-Kec3}!|g*R^GU;V^dwVO+3j{v8WsSk)ATOb
zHa4fr4mrJ(=OaOJE6gxDzwMuOM-r7^y+yq3IgxQFOPoc1MM0bPL-D}_VQ>Wfeg++a
z9GdM9Riosj02}`~lHZ?Shy-Cr5I>NAdOwWey)YMopvw%n&fdVo6U9&bv4lg^1jQ~)
zo4KsN@31p@`C!&Cc}<>1gbw%>1j2&|dg+QWE$<3ut3`ERQoYbB`gtU!o5+F%&2iZ+
z?Y*f;)su-x4XEc`W*~F<EyXLm%RH`<{d*V{ezkxw$V=};hLY%yv+P4MLjs%_ExUhw
zG_%SX7a9Ge95xlB{qv7#8(&n-;<3&*vw&iYjA8~+GTeMN_|<#T;Oae{Awt;56dRKR
z_fu=GdP?K_Yn#*#^2ejb!S>$2st4bhoNa!+946S6_o$uy5K1HjPbCm5M);+WQrpIS
z$4ty_%%`|`ifmRK#iK!*iq{IiZKgsf>4H%j)y<z3`*-O}xXg4n)QY3PQwc;M6GVa^
zZ-VB+Xil#Lc;tl5^g=Rkb+iIed-B%GTWf2_t2^JlZ?j&(eTJEpU2N%TIU4*S6#K8|
zzIy<kYem9S%u%lGY?NZPYWCD{o7*dHF-}woUzKfCwYK2rSzWKzz6*1vZv5=p%KVQk
zj|OnYgQtIP_in)<&lW7LA=S8H3fJXA9;+Evt4>!Erl(rtTlKUY$!PJtb9j;pp_F4!
z7Bqe_)t4uLX)_R5`iaxD7i+_3o<n@~;4#EeHwkU@M5eU7`*&`=yP<mA&SI=0<1*KG
z!l2On;86Chupro{m>@Ejl&1|Y{NEG{CgIoqK+xGcGSFOfPX2{Q<-{TN?b8b8jtjxl
z2;mo-?BbIPG$p17MnrP1D1MY`HR{MuzCZIwH_04of!xZIedcvLeh~|@5cuevhxgpR
zze(&xw^&wwpR1Vp{%J%rI03YcZiX&Q5g|TeQZZ_xPq%e2su~?D4lC5Lu?+~&d&6MO
zbowqIsX>Avjt=|Di_!@*+3d1P(<7p94()!w)AY}M*FC?l+*DV==ny=F?p@VUMqbvd
zP({(ibd{FAxxy@Zk@=n`sxzQNx5VmMETWI(OYA4PiY6thEesj6dj|)Y4Ti;+w~a-O
zinP!A57`p8f5`y(0UjeG;|pE2F_VfhcZAv5S^GH}NcB_V)M5qY{EDz!)wu2zK`%o!
z9M5K@2t`_@{86^V-UTb3YJ0>NH1vC*`lNY{`IJ#b6*Fo)-axV9*31X8YsDdaZkrJ{
z!RZm#S)N+GY#8;eaF20Pj8<agylyeuQ#g|XOpo?_D_S!1eQvob{i8i4il|N>$eehi
zM*h=4;TI9&d)<v2BPOiPOg#NdLit?2g1@YZ_)2Ia#VG{6a5sg3ZYpON*r`{5c*6JG
z*}AY2)qsH0G1Z5thTdOV8M>A<N+w{0l4H*ZpCoxc{#rPiC*M#@Q7`-I7ZE;ui*jB8
zI~5q(R2iLOdxLi8S^9gAWOspQ5x2A*@<%5q<Kmwt_B`|NK7McfMV5O^j)d_q|A358
z-FNb8lRat2KSrN}`xYUQ2|i26>F-1PDvk2VhS#x0G;G*Pcs6kuyoLM3E7_Zcqa7vm
zl<?<{JOkH?%2+Sv*zAV4916FHvX)Swt;XO<XT6V|GNGT)dC94K68AWW%w|HT-l9B(
zL6YI`Xkb)6jWhpUb`A7&OUue^ZQ!pF6MONym$bCBqGB}E1p!jE;Jk9f8}tdum)X#)
zu0qaJYjV7CgxkENh=I+5EjpgAz?MNjKD(}7^n6`wz(fxZj+`Kni0X^>i%h9F2hbgW
ze-Oy!X<TgVY#C?@%Rs?Nk94tc(1S6p_<-wrAHqJ*F&aMC>B&!-#5q}B=HQx4V9ml>
zy5;a`Jx;`rzrb_iF^vWTNzSQrs|s#v(^oiC0tqPaLr7jb|IY2$m%@Jw+Rn90l)i!s
zyStWU6DSY8e=1pifWIcehY+~CxwVDhM@2>sl^ZLlAR~Dt=n?^M4lOlR2?)Sz8!HK}
zw^b<##tdazZWz=lx2l3=D=<kj78*pS6a}|I_dW7KVuL`4v#1bmrYsx3gU^6j>5dER
z=<d!3XAS>4ci9X+M1cSgFK5um$AGB)87{7&fdMhM!xs3+5^&g#&v9oOK17{gUp^xu
z^(C@1^6*T5=YIlGMEF#aQ&kXWppWrXf?{k<%Zg`GbSq8x>zbX1rkRF<S)GBSV%@Ap
zE?(X#{!aF3oMda>T_%`M&cmHAD=Pz<glg+0(Lmt?(60|A;<Rb<0&aJz_4C7vh_g%h
ziyyR(KDCEYrhyLxDNrk0%`a{MCnA^NCj>!H4&JYzz+}H7g1mbWGB{shT$qvbHNYu%
zWX#ZuoK0&hZ3DY~7@^GBK4Sm!eUx`2?{bk|{Mzf~yV#ms=z_q<$2Y56A|224abqNx
z6%T$m<Q$yp1Rr<<-*@;C%ww2#qsR5dbd7DkD+weALE>#PEX!Wn`0>+y+L+8+DKla6
z4?1tA7N(7uis;`=+i}ooS?0b+8#5NW*l^eX+n&Oe>me!e@f^0RJ@N5)xUSnk76LvN
z2%K&y)~erFEQI`^rBK74Rd1k@!hgCyOCU95#m1xxhJUa7@Rm8Qkm1b^Ju4?2N{wk|
z<PBE}@ZK?#n1GeLjaAe<zIW9XLpL+-C@A=aaQ7kPlpN|J5mOtOB>Q6V>Rwl*-yt7Y
zNpPOMvBM>mP-pF8q;$KsK~0r%zUcYJ!*%yMCeJ&g--u~zSYsCOrsj72aNS(hVlCX_
z=cD!zb!k*f7gs5Gi>EPvq`bDHdQNYZP6H;9d2KhT$A2wEn`zW~H<0QdIEZfGl+(@s
zpzAH5qH3de@xj0XaX>;4&;bN#5djG?h7cq~N>Wm!OG-)+7(hToLQ-i-5s+?a1Zj|N
z>F&DwjNk9xwf^`1XSvk%eRAfU_q^|pXFvPd2YbYX51<?hl3wBBB{jV6suv|lSV`uu
z#xP^}ihP)?PG`yEG!@MXO>^ALSe`!7XVL9j^&@Hv)|v$uR9PlS?s^Wn;$4@i7}U{1
z{xGz;=Dp}hx@p*aA#;AP*dF?F0fwo1(HbLCzB5G=ojEF+cW&gmgI1wM{@Aw8X!D9Z
zn@QFQW1M<?pTnWILdPs-93NF&`83a@Jt(QI`R+a^9@;^n%qP51N1*!z+d2@Z<+@`q
zDg_NeI5b0JhY<Q;dpr<|t#@qyzMmT;sPM>6H=gF(@f3Y~^HrqX5A)V19yD%u%(Bh7
zsrad-h{?SMn6PptB@#O`PA0cj3AY&V5rp2<EVV2st6XR=zW)V>!(HHn-OI<u=7*!;
z)NnZ5-+@NMFdV2im6-H!QWhU~Kc#!wZX-0yBGKL}`{k_h+)&!p#4XZLbxr-ht{dWk
zY5buS)QuglQZzOL?;dHZ^YvdO3mS<Rb3)#tJ7H}X+Tm)c1Qt6#TweY#_*zvB^jTkM
z^W!QL(5X6siGa;L(~)ey@Fx!%j<rgi3VRt)inR(;d6Om&l?lPS>n`lX#dnSC_Dt|4
zb98sr)XI2$;0a(WPkGPX>hOwWWyA4tS`a1||7_Kh#_pnShr;O6!`->bnL)!GS^-OU
z-0s$r_K@jF#cPBwWj*QZH%M@oiIGWsgWrU)#uNLqX`<WJBT7acr}>eF>i-Ueo`+}M
zx!aAJ-bB6wHBi}wij9ep#jdQZOsWLXi<;;3;ZjU`GAAiLljl0x+coCTqZTDT#7^fq
zay(I!!j?#xU0Nokp}TJPm~+%Y@EM<*9*fMaP}QA&^{VSsNny%28|8Y$#{CKGk;@wi
zYY;Cxm6oNaHyh`LY6JgA8^^-Pm;ja0Ny@d=Rh(p-<yO|Q3t{V1#2ekj?|#7-h?_j5
zZXrD3qpA37^8Xb@xjL156-3=ZE`N6tF8?dyR#Ri6pz}8DTn|q+DC{!Cfd^4BwcFpe
z!uSRR*kuc$3L@tl#k%q%FC_=#1diOl->Y_vbt*K72fJQWzwPq1;jvD=W(~JU9unnc
zQd^*#@XK={ERCQHtO@Ke)gs#&c+b7CWx;vH2w1(rk&==&_D6}lvV#^rEU37U6!y}h
z2JX3Tew_7Q!j0o_#KieTmuu6y;E0<&JwJWT!^$9gWlGz7Ks`+<-udP6^*hu)Y3hc^
zqo6wNGYPk$uN2No>oacQtpS^}o>k`gWRG(Uf4xO-$Hm2gU0WI|?KjN10qNwbKL1A|
zl$*k3k(7a>9LsrA3KJQs`lbm8J~`|Hh7EYkSU7k3^bNtT?MX8E2*!pTtY;`n{r6nU
zESZE3(Xv}`{?A_);07@$$t4KM{3yAUSwpajYHGXC{&|LD*kx<5j_y@+%_3@q0>b1H
zY1dhm(dHdXlcqJ7l*4ACa^sw=n;+BJ_$~_c`Sy5T-NFYxySDfzavL>vXO2Sz>egpE
zgzOiEU(PB*2%(^$;75^fpLv#q#G)hV7vgl1<>-6%j-fERs>+3WY5QZhM>~g-qBHu}
z6zPB5&0pCkNlfYzt39}OQ+u>Xi&f_Li@M#J@&7!HCB6CSW$Z;#QXa$R@Zv1cbb79@
zUyyhXb%FbQu~WbJ43N^WM0(J3Jd_#SD<;w_mbO5nHxr@z;IZdb`!j9-@)<2^|6~91
z3RH--98`*mD#K-FW#$qbZ)35fJ{voC0wjc;41ImZ?Wm;rsin6*HA6mAZ{V>})aF{m
z;^-2Qi=?TjE10bl4#5q!V0oY&)#kE_W7E|KUImE~?&6ZVyea&dK^G~}Xs$5zJ*L}{
zyp+W$dxHNbUfNfC&(r*5NGpEdmdohY_ub`YL}&@+K5hBHUAw??6cplQu02tr1~9kt
z9WTFn=?mH?hL9&_aLyf(QX)rk>zRJvLTx*_m{*Rndd=LZ_0HYWR<^RD(<i-<6dZ*z
zuIfUSgU|o?@nZhRW8lqX$}@+4yw2{!;d*A82#Kfn_W{0@WtNihYrk2>$^IQ5@<@hj
zQU78Q)ui<2C;WFE4S2V)LiU^xntbTddbI!&guXc3dr8Jc79x~+kY>^6UGsC*f<Cx&
zr&(!u-}UU`G&{BTwG1z<)zz#Sg!)pdr+#E`e(CY@j_Y@)oI-$w;5nIXZdJf@zQ(~8
zxpT1o#J=-5%I&JCqsGP>=a?gAtP0b7Tnm#I)+=^U(|T>Kcb&Z>mTa(#G$}UFE+Muo
zI!tY#3b_!9no8w(Nt+R4+&vG%!gwmw_PLrikH!Phb9F6K92qXBJ~Zdp1*ksN508kR
zo^Zq%e^EXXsTZx3USO!v6Tn|#aoU&)_wHSZ^Un7gT}R6&s1(6~CksQ@MGqn&O-#t!
z*eRQMxc0;$uF&zM&DvSd4)e02UnF%6>B?zctmZoSYY>N|IX;7+3!4jJnEl1$DA<Qh
zsbzy2F+&RrLh7%W92A{ka}S#vh9qyH30#Nsom7`}<nZwVikCZ6{0;c!nx8*`_i78;
zio&g;+r%jO>$S#8Zfha??T#3XSGGuhM|Yx-V|(a~n5^AMUZlEzBKGHnQrd|l$6yxE
zpB}7uc~YbI8nidoR#t$dC<nZpCkhH>ApZxRUra(m<@<$$w`a&S=Fy)C^qQ}2B;~c2
zXkaj3J9qELF)Zq2`|bLf*wsn*+-#q>!zge!Cy!oz=tQrNH}CF(BVH0(VVs%~!NCu(
z%%C<2c4^WdAl^j@T^3~v4Tc*R$&j3lEI8`Mub-T=Tmd0Z#fR$Ib)RBZLmsM#d%axU
zOlMlbZ%t}$mE-%q<ZKXo>cQUp9sbbfduL?#jZhxUSZN7|hgWvx7hTE?N^MFS=I7IY
zZpYx_$X7So&>SDnX8Ykq)Qqs;4nQ;ogDKQF1V*sU@6Y!@8ze+nlmrEq$x5cw?6_V&
z89KQ`{ezcH*`Y4w%gApd`^=7>Rjib94qwZC&yu`<D8`8wa{U}+1-r4tAqbcB=o61#
zEfMU--+${{k1}`fgs6t%3MxK-e_idEp*$aFo@2A0kRIjlMLL?B6*!es&mF?dk&1th
zza<MUj(~f5bd3zXu{hW-vw`B}{J#K?ucXB({Y310)=u5`Y(7~EwrC&eNOE^a;xBQ*
zQ<`9C#+HYudd=C&D!cgXF(Y;dEi0q))!%nUoBJ>=f1ab39;$yG{wCx#R9|44Ll*Q-
z=gwJTry4mfyT|yQ=pQD%)&;o^5aI7)p_$$0i3rO7c->dm{Y>FfAvF=hhg1`fXAXLJ
zq_c8q?mP8JGHQ!?GL+(7Ag-zF<TM4BR(hLJcc$hZG`eTA&xJ;ysQGGV<ShoTh}xL6
zioM!v>DD?k@i0x)RYiB!ZNcz$FjJg-;e3D6W0Bk;UCNWlZT#*n6dbb$rq=Z*NhL>z
zdmYJg+NDk}Vf(!C@L+o(YSPUPIbTW-Rhn4oq{)jZ84Skf=bzHB@oZ*f>x5pUC=ZLS
zL!;ZOStpDB;<yV>iGY|6a$%#wU3<saJ1&iZTv{b8>{8liPB<7fQ$6|W)S>pml@7h4
z@O7c)Fd&q~N3+vRHYPSSIzO)dN|arP16SbL4u|%5{02n%e@Y|Jif=1OJnv2@r=SWc
z562T-Mg)gryq1mMzkE0A{7XOO6?d{n1*~=)3GqsLEsWvsLD&l-h1}duj+ZE2)AspC
z8&fae)!fIH<k#{TQPEt)*geiS5Q?g{(BC)DLkE4-G4wp)zgG9*CH{wFPCij5Fpge#
zig4QFe>W(LjgqyvB>bjh54qpz^dDdNdA?tk#jJjiEW@2%nDk&a`S+dYj|tyh_z*1=
zKeWM-&5nix5_+%Nfr~_~%_XoQDk{ccOaOB>w4od8Et;h4qlxG+LCE)DBQ$w;XF3*=
z8fc&xt@a*tcz}|N7qC(+wb{#9r|bI3qyJGNajQ$rYm}eIbDb_r!F8DT><r#WSLi8m
zRcwJk1VlxkS7Vc{Qao4&T8yAx;sW6v=g`@T2P-Ak!>x@Xq8D1pf^PB+@ffjWDf()5
zbFjCzHgbeVh0KA?olRXi%zUVraG%po6MtKl+Fj?CV_i1;Anybn)M3!|1SQzekPxN(
za*)p~hYiD_mg@l00r7Eh*JPtq2Zd0b$1f-Ej;hyWdsKajml2Y#x}r6E{ibrGz+y<`
z({romJUevCiZ<Ft?<km3;?L4ZTWT0fZ6N({mUcOCaUpo$2Q3}Y54mw88*)-e*p2$W
zGugz_CN~?dWtAT~fh=lgr{d@*IOoNXqMsqVCi$&~msj9j^_M?okL?!DqS4j8f_B;V
zi>&a5Azo!L*N^$G>Vv!L#8!V4{TcP2X>vc6w)y^<s%kIrS4rrE1>6qzGt~1lK?4`U
zS8PJh9KGX2P({UEXmZuo)-E2@Qbj)DK|xlF1GhsBJoCNZzI{h8dYIRv1sXusYk|i`
z2&nwTN)1YR**tKPnCwl5%C;i^Ua>*DN@JSP7{`vqE0~kj?59aCrEH;#k+LH|Q6H%9
zbo2wN0}_+NC5|>ARf<G-kg*UGR)bVY`CeX+aUOV^ORx?K3Lox&NqLcnqVvB&UBZ_U
ze3ASDnk(ebE|KNWT0hvAdyl4W@aIUW2H^Jf+m#-V(w21ochFHvlkEuu_4W1t9xv_l
z=VuVO!iM?P($dn<(D3UQpnF&GpBp!Bfac*NAgmhj$bCW)$<r{N+|i2cxxplk8B=1F
zJ__r_v1I!dnQ3OPWTb1HY9)J2=IQKWi@nKICWW_8DRKj3smG5U`+)xpSo6C(Nqh*I
zDK!vz+K310K6vH4t^`wmPL7v9>zVmb2Gd;3x`GQi?I>4>uR1w?WfUpJYX~%${!ePZ
zUKW~wjv7({<DR{?FUn)XM<eyN;{^S}^Pn${d+Gd8%7jl0XfsqWea}hm5_OioEP*wZ
z3^(}At+l`ed*Px`UmO2Tl1PrK83DJ0`QOj}bD2(1K>a**kvV{M>rfKGylu&($1V)%
zbsLqK+Y^-rS=AUaV?5vPxP7U7;QaD`R`eEh*vm5rXAR`5TkU^?5F{lePvw|n^Uu|B
zgU}#;o^ww_aq9{*zA4|5OGG9M@pp$xxdag4tlrbE7PHlJZl|Ab@-3JS#xMtdQjJkl
z*xick<BMUAV<~iYc?1f2QTPLras*Iks<)MZVt;v%X0a6ZC-;v4&H3nJUwGrvXZ?no
zS|G1U@eebWQN(_3eiDQCl_-=75Tz8o6Prz%dOoBRA%=T*z%M|8p0YSeJxW-VNytJe
z=$?Q}o5Y?YHbrghdMn-|KqnpYby5ntif%dL3L28wl1JtX{Plggp0Ja3#|kFa(?zc3
z{EtJd!oz!a9{`a^p6SEBaMmxAPwfh@PmG(~!+Mt`VV^J!j#Ak%ti^SSV->tU(UOMG
zFI}f;!*3M#;!b6Ed%<X)OJ%g~g!i4+yqx9;YYmw&TJd>AkZ}@%vE?Wac`A|cd_oA~
zlTgV}tk-6z@lQ@fq#o?1+*l^fASi|bVhH4Z(`U~<E3$?(O=5pSC+DS`nE*CrYN2Lh
zAkl;%>b}$T{_*DOItN*&;l~OQ4kDV8OINL*8V5@z{0p7D8bt^WL6tcmFfdl+)h1W@
zzJk1bg;f=G@6WY;Xs1)pHM#*&N51U|`MQbca$oZ;^Z|3yeNf%xRaUk47d&DplV^=y
zdk*_$IqZs5O=i;Lz*MFG*)MrD!~}-`JVx-DMbzkCjRV;sMn*<OXE=CgZ@7;A^vT>h
zDIW&VQEHhSQvd_zd5&t6bi6~J+I$=iry3I*5zT+lx-b^g<+Nl(P0JbPr$K>VgwAUa
zq<NmBeA~7G$~cRCIn#img1P2$Q(_HCN=)2X=num~FJQI+Zb@Od%sjl2VSUtE^=WNX
zg5%(W!zT$weej}Et!m1u;6<0aGW^c=NDcDQLZuEq!og{7>#cwnO*=-*<^vD`1oF*A
zLIMH;(5?Y_(s6;sqr;^mfF(Y2eoNOoV4bn3!y`taLkE(qMG`&J|HB0!&u6(8Q2ohr
zVJa~38b%j4bbd9Kk`ANyP5UzPMR%*G0ba5FdjSpl>TZ5MK84yep7&RPMo-!XASLnz
zKW>~aw(d|{Y<`G@+AkM$=7#))8Lr0W_r21V7S+op4f>!1d!&@?;+6jbEqMuz!5Kh*
zOt77x_|zC#Q(H^g25<p#HRT%yh8wjIU0=q>H93LtU3y@anuZi?ti_eDH=%=K6?c_v
z9fY_7!x1f4_`Iy{LS@ObG#lo$8Ksvzb#f9!--x;K?u?}zY5cuXEB}N798U7fw=McH
zoG}t%q+jPN3l5nZft2Q}LS|q9?ql(8ASM1%vy>|IUo6_<PNEn3a*|gKL2~G7oD^No
zsS9DbS6WR5&t5HEGtP6KKau$~y;ls4UcnwUusJZ?l#pBk)Xk392aAWmP)2Vd#s)28
z#yD_~f*_RVY6eP2`K|{}nAySNd;p_53195K+$bV{_w^&rA$o|<I*RcRL+y~6@KJu`
zQ+%!zwth!jLhP@w-fa>5y`|^@^Z~9Fi^X1OgKiJ#Y*=V+ZuZSu6-H#72#ym9Tn+$m
z5A>Ys0E!rpwVQe4)FD`F$bjLz@>_V_prutC*|)w?^A#ri`P9-gttY1`v2R%!!@co|
z!Eebz>Jbp~1KD_J4TZ`E$ZV6M(P(mVrV>c1lc9n2=g*&UaTqThQ1zXeo13$)L@hEx
z<{4cSq!W$(=_wYjP5sNn<tL{vj^&s6Qgo<mZu5QYS+A!T(YvzpC_F_Ibysf+K|G-C
z4Uun#cDu8lZ*b;ACD%1XowqH`%{kE!(JX?jKftW^rCca}1ikrlAhIHiz8#s*PQ&Bd
z323kw*0w^id(GG>0s_G@o|)Qk9Al}3a|V8?P#Ei01=V}^?wzAxcYtn+G6;(xI?^t*
zX5#112SgG=8dHG4?%zMw9eb*^$|`kmxJfSdLQ7O&?hoPNFh=w3J54YD|5GT`N}&zM
z+%=hq{evYp`*C6d<Kxxe8<&36b463iLe|<JvnWotOiPCN(vjej`!P6N%(8+TIXT^x
z>i>e-yAwkwl93&%G^NrFlNaKGNQ6|h`AxL%G4hV)FN<DhdmjG<E3hyZzr)V+Xp%si
z4gYBF-j;Ii{GT38x5h$IK(Nj#B>ysxK2F`~L<&Kzse*+X+pi`j%m=9qE=IvII*MWz
z4CZ)KfNZOvJ|~%M-OijS(V1GOHT<RIgJ_N+uOL}A+L-54YCAACe`0P>cMhiqQ2|U5
z#+L5$uqE6Ozt+`@R89=z-+#$sdAxQ6c>fxRmiCI_^8xod$N-@cje&`&h9eQxN%ulF
z)M?ErsreB$VUv|+nUs>^s@-EOwmkHyc+ztUfv1>uNk(agYA2)8n*V%;a8Me7E7ZA=
zVl1E}d7$li9fP?)c1hat_{ZR(f~28CV!a$RdOALbh^}&H*H7~zy3L*`X@x&09b$to
z)_suZNGMAE7e%5-01_^@96g3kOh|Cvok?CyIaTp%V%kl`C02@E8msY4zOQ1mM41h_
zt$*H{jKbeRHBp0-7JaQQ?gM^xEVVNU3FF%&u^xoP))F^R=1JQN1=g(y#G=cf$9`~X
zO)8{UEGT$7*<d<@h39ngAn#C9d=;;Q?LVi@I%#YO*a#!WuVySj7d7BLO@3<Yy4#_a
z<WdWT6=Wd7S4$&ro9jCnIy%u$XK7Q?2CV-pxRXp6n^JK<L!r$E3fgFEPz7gN)9hT*
zgmdKtY=k%8GU;64B0D13F!j!nQ`=osxQTvvX)6BrNZ@^s#NK&)`hiNV=t7B_{mBgv
z4+pynvg_}s9rHU+j9~icpM#q`Q@7conbBH$myY|VON;Sbh~0QP!~$BR4UP5RtPXoQ
z|J_JxW7my`^z?}1wFOVb#5CE&RXz{%By4SMna;Jzg>zTd&=JtNkAL*uI_Ldkv2@<-
zkB@r5v;QlrqBwGN#V=}##K5xr3OGY3m*W0TG(fX=OnM1O3w84C&_8GN(#7a{JXi=6
zy>179VwxY(lHv4i#~9q0Qe)oZ(V(-1|LwWEBICyH+MtjQMr1xz0X5)mK<;+Fyg$~-
z>t~Es%j*8C+m6!bEI6L_D2*Q&=!XX??oaRFckAaN*sWY@xj5W}^^WM#PAgzvp5ERn
ztNX%t&4Q1t#C3TuK>c{uv|VFB7c!*0ck@fqkE&+4wFjAXDy_52pZq(8PK{kCx0Nw5
zOazZXBg8cWk4fCmsVZAdxLX>z?LM{g9=3^YLG$y-JxxcDCi00RS$TKd2L6Zdx=bs$
zbFs1A{PYOhS!{2?-*sHxW;G|-g1FrK?QOjXAJqoOG&2XMDVbsq+(n|CE5iDpha`T2
zxGQ`<6Ifoa5Q>)d(s794;^eebxre&-^Sf{S5!v|~TG!#{Jb{Iul&Y+;6U@PxzGs(4
z5-HNC@IPq8tr3y6I6mI-<-sY88tB<a&Jv_t-C1Jx9JS{ZR`@x(^|GcGQfSZhA?(%R
zKF0qtS=2z){~Osz^*11wAUW&ZEvt#W{+fALPOI`~^JPdqpoo^b6e-64gcW=JU$jC^
zr$U)s4AqQ}*PjeQ+bg6}P;@MUjagY)8Il`&%OuoK<k*0W>lFV(pCDZCx6{h+M(u@K
zj~H$~&p~WpatEjHX(ioCNBomm(yJVI{{P?@mO@?zA;=bpar&`%j>YbKeEZG7k5Aur
zSRL;6eW|Qp#Q)@o76AJG;^U!$DC0+VyC<Hg-N%|bFwzVr1iUK$f8-S}HSjSpikA>v
zFGvCAcM{0oB$`vC6JosFbFjR(!ZTM-bIB8IiqltHF~wd$rT_Pjkiu7?tQzEw>|Sk)
zuqJU9%W_;l*!yKl1IgJ(jj}t}Fci|=_&s?y<DR>>(>tALUp==LSGOWr@8PijL2(|_
zaCJqb;O>6P{fD=q+-refHhW{IK^qC`b>nq=>%)s?IvduWd|y*0zx1|)(j_-m;%zFE
zYVN;7Y^(ysS*W*rdO(Z%*REM#l~k77w+q!42o}BJE62S(_0*FyUv_+bZJ_<IbvL%~
zkpS&<))6-&)#yK4GJ6*;(3`F2>2b$p_cTG-z7+Y@3XkG?H=&krDIc7_d6l?$&%MBz
z^L4-HE`^!eVLtFQ&%dwGcN-S``!A|;Up_6;`2@fK7H#LUCt-qrSMqu0aO>}!nWeEq
zqzxbHf5)iJe*pqDd|oCy5tKnYyuk9>8rk_r*McM2j-?#iX&8U>=@SkpEvDEElJX^%
z>$I^!W}^<6vq4SW;+~?Z%3XHJ_@76S1M&h@`(dk^{{DWK8kFLham#_2gQ1m~KZ}`T
zZ`L3Q*bT>77MA=F`e&#M9>`E+&p~xvLqlVs^Z4?!)!~3M)l1;{OtFd!FMqD?uSe}q
zde&6U<ftu8qIJ~1CTu$40b^7_Yy`V*zQc;{35N^k#w}S~MKG891$)NE)(it)F^SC8
zwW|kX#kXgtyiPVzr3?HESH~(acJi(Da+gbQ?GNSNrlfc*U8>(V&!tgjR|L@#QJea`
z!=)1^I~DNsDHKWxWbQW3jDHIwDepHiTiUWw?*Hbe(viW3QQ&E=!u`fZ{|Md12XLqq
z%n64SImnt3IM|xc&)`9|NxWZa;?%^ubcIj`jOf?(w7<)CTC;!t{#)(OY8OVn3_UaD
z+sVqOh|TY58Co*&^<UDT6m$Tq+nM?jtgiJww5%haKzGYy8(w4i@d#uGC4#>9+*?CY
z%n^|)-27Sc0t-`V>gY>THCBrY{}%Rhtug5ca5vHwz8ZHA0~ltqDQ+}D6{+Mp#)5=9
zgQZ)={?ku*zC$bu7?kTSd7!ZVbq*`MIytJK3)!^M8L8%=djW7ro?;*lCOHAj{2ll&
zQ0PPOQJ_aGnk`)gflR<QQbrG#RO{e7RWj7_jJr}JHUJK|6)*5uy?P07e?_OMwzKD2
z<cNutC8_*q_VCY?QZZ_|KHD`1^Ia%q{d|1~`n4`zx^(H&r%ybe(qWi|ZcqEOQM6Oa
zDg13y!EG1Vuk7#^yZqf~<}Ho}^St)LLGr^;o&Oes47mFW+Ao$ktc(e9(up|P*7%ZR
zd7ak(f*Q-vL8u^W=J|!;a&5`+YM``;x;g`_g?yj3^GXejJfgaA;eBi@mmLSOvdpo4
zwycu&(pUERVaF^*SG5DVmCvDXtN+ZtSGicUK=7oGjX&odm>mdw!DGt~X)PArMxJMY
zgA@5I2KZ6?i=SRb5^x4b-7Vl|R8^KZt{ppmd~j&UW_xF6a$@2_P#{Td!@%cj1R&h7
z3Abeb>N;xcr)q}paHIF|;E(P7o|@W*zCybI<~SofnshF)N|XN|0P?nD=HmI(@Y=Ng
zAscQYvdi;7&;svqny8B7^==}~su|b`BXk{M(I~yB+<I)u8#n!_(=!KWxHdAA+9a2d
zJ0%Wm(ARycJkKf|Rz1qX#y0M#HGe-f<4}1K_VBXNyRK9L$0!_5=kj9sf%)|InM-|0
zQsg%0;=-#Qi6yL0$Q;HK#|BJ*f<aWxO_T%X?G+y~@rlaI5hTvvc8K{=+#u>mRbs=y
zc2F`kg{o6XgZ}H+uR#?>9Z~`NSP?G(FyZ6_&*<;cQdXC-dcL`c>;5)U;msd;_wF5_
zpWEBuG2*+7$Dv6{L4i}VNHO~t01ubE906qYd@_2<84`>|`&Owx(^FGtlR++~Y|kHZ
zeM+rTC}{17WhggsH^T;abgcteL;D8@kYOpYc7gVIR8*8RsJ*=Eww!AI1mj=w00=bB
z69i#%Na$g<Rc`;nBr(9St=!G9JtVP6-%t74bSpC^r;j}fl&VV%W5d}(_=Qg|kF-tH
zkqlD12Kqlg?x9hlo*gVx6qMa2kBJIc>*4YvQn~RiO$o6PcRBw4$xHio98v6DblrfB
z0R*TkacBndft7Zk9chF52xLm<lHtn$>UEh6Q2_}VWd#LmiE&Wjuc@giU2747wD>wJ
zE6=A#z*2fz;$Y?MybqC>#)dH>nG63OpbxDA%Suze$VW-=AnY0i%Rv%!=(zN0?#N$q
zq6h2W4Vs|Y+o#?<AIhVP5LLeO^AiGY0<AwN+5^Qc5~5bXNuFkO@9*y)>}@3{BnSfy
zi@cmV<u$dlZdYo_z5DK`NDQyD#U3nLI&SU}7mwGN%y>5&U>Wto3w0{#Jqe{R3?X()
zI7E?kp(l$T4N~kK#(6_4vpw2wGFYN&A{xH*v5hp9EM(XicKs)rTyw;&IViM8@V<aq
zB`j!2*p~vtR#(^7ieJu!1_j;2KJX&i8S~@fGwodl^^G}O!I<e}nMgiZh18D`@CgrB
zRa8unXVG%m=pqMIBaOmU5Ia_4xp>}NuRY-*WSnGbvaxrfr*~k5&L#`Dt_dT2uf4&-
zqJ=bF;D13X<{KDzG@sKNaQXIky}I+GHb44I?b+>qfIWj0{5v}hDd{R`_?Qk9Sb<#x
zn#Hv^*sXVt_SdQ=_Cc{dYI^6=O~VV1@7{HnpMV)kY+t|3#;X2ewMhBqf7yX^a)kk@
zoIsH$n}CzJ{7B94(x;fGBl<p)3_<MvL~~y@?Bqg_^C@paNKBpUyki~QkP90DT)(38
z!_&*rf-il2e7-si>SkYfY;C;?Gw=*}E}{LYZj1p-LRL~z0MsV{)PW5TbughXE5Mio
zJj#Fv5nX!A0@>6llEb~>qaj##xNl_Xfv4pk5RjQ58EODHyC&xB>C=)jG91HaEY`ti
zi2F-PO0uxBM%;068ZL85R0s?&{1h6x*?hAVMlBiVsf{+C<0|2{9I1euUZdEKgM%Xj
z3c^_ZJT=cpboBJ)aF$Y1SiV@n&E~%7Kq!0gy_s;(J5wHaaeS_gRV8P$s7U2MRt{t&
zkmY*0Fx%o^K(P2r;hN~q^w=70jNPQ{a7g&BcJcLut$)?Je<p0>K1{~ja1t5ZxOf0H
z<D>0y5=_x_fr+AGCkVT6CfMJDN(SgUd-4M7?$PKNNN>ZaxpSBZ0V`ccCrZFloI$D@
ztDmh36BX1PgG{>93z4j~6wnI$dA(sMK|rKT)&_rq9)fmaY~kj^#CEW1A<v&xf-QOL
zvC1)bXK()IXy03;E-^Wy3Cl^py}L`=79;Y?SpKfRXf~_zkUK;~%=pi6sSSgB|MfuH
zvRnIfTc@Y1=C=SHZJ7C_3=p@sQlq9!JpJSAgUU?E4(PmAs4<jpp^cG}k;y2CDdFDd
z3~M??Lg&0y49Ev#SkqELFAQ=_NMOY50%PfIR}+vluc!ayd(oayY|##d8-P<Uw81Py
zHZBFD&pmU&!NE(ei28Sz5wB9AT=WcpN;{2ilZ~-FEZTn#fZvIU14CB;+s&Rgv_U`C
z?QjQ%HGuzo>~YL^-$yBXm5|70q>}6gg`bA1vlb@XK6P}tICy%vp)oopnwg#NkIeXi
zW=9J5?IeHv6EH*Wj){W-8el@gv@afWhO`4L@RDm&u-qNTR`oL;r4)Wl%*@`nw@Qmu
z)zu6b%wnHby)TGxWR%?38R6&Tl+92cK1dahN0*&0aN01Q-1>SGx*F}ZKT0k(dsR)Z
z0X=Vc?HU*cN6>;_(z^GJ^}jn!M<1lf2eJel23qy{)L(+trj+*(3P&>jDPN2Sg4UOF
z3TkwOADSUa1}_n0*%xc|KnR~zV+Saoo<{cNFPj#_$+a_`-{daaT@^PpH00UJcG<JT
zz(C2=lKZ1~w&rs8nOe^*Ly?HYblg_ZQ)~*FL<hBy7oVy)^zCY^dz}uu!Rixc9FCDx
zy)J^NJnY-Rxfe6V(!a2b2H}C#Z$!@$hw2t1bPjhX$wY-xn1zVV1&cp=>4na=W?I)Z
zM#h0Pj%Js`Wj41=V5q&rS=-p8DWK}Cs?_E}U?eP-vAMSV)m4{lMFr#A#e<ytol*v7
zzeF=CnUzBtY52gI3-~fXW>%L!w!gUYsPp)*CNAOUP~8o~erIJhO^+S=#f0}vw4^a{
zt@ZT+uKNP5YoPdZB@Squ!)pYlm?B`mie!axSzi<O8BBOchM5Jp>F5f1Pj#_66w29V
zKH`ASD2}bt8+|$Tn&iA<OH&glPQAlv49p+I2(OW)f$Cvp4!2>)!sym{xvr7({0jkz
zo;Ao4aNIz*n%r8<xPWY!SyiaOd8U_vnhe;fv(oV@-<P(4ZMNum00|HlEc@GAT>u?E
z#DZ}9miqoXgQuYew*%5jgs%p~00}WhZ=S+$65tKn*qwfJtJYm8wErD5^L(?Ef1(0{
zhFIs-2BFQ(G)$Xl_AYcM3j2-yy;nPP1ylFn)@9ctmaK}}H7xIdN>BIPkai<}RYb->
z#Vt*R3-jl@f#0`0RZun>l#UPEgM#dm>t7KL4jBgvs0(@;wEp>FCXFpDECgvA`{9Jo
zpDREr<y!E;TI^AJiLj2Dut()LL^efR^SeGw@$~s^0ektzD2>1GyxkuEu=6<FXYAbe
z$~e$wfeIcl4n_&4sFgc8I(BWhU|e-KD31v{T&Oku5DV7BJgaSak6yW-;=viT<3q-!
z6J3lTS_qry6uiZ9P`-i4!p8{a=~N{xZrK9X50R8_f6L_56tGMOMlCP2A>wu#k0d28
zbf?h+trlWX)}c!C{J={$lX7!E>V&IkIVP_D5r;^FXzFNA`-qfqr-}BqwSIT{Q|v_=
z8X+(XsYfMBXeufe_N9C%5c_JYE(?o7F=+I<tx5@q0Xdb%;I2Iie;*yKxsnJ5`dnc5
z)_mVufW6Xw?~+8iTslETsM^TAixTI#Wtq{p>0CdqWjx$>6=l|TsCseV-NG{Q3EmZc
zrUNq^2#pbard7K(z>|@~8s|B|oComMQjZ?J^YOV2P7lH!o|`v){(4esIZS<e{??GF
zx;a*~9LmUIVq%E69w6EfXHs3d<hZjsiOiV==oQNgARXMhl2+uSYmqmf%N`vbtU<hC
zw;*UYud4HQ2Yv|TopSn2vFou%`^l?5>$k21NfrU873gVX%(Av~-3y>E3H;4FP%KYX
zO0!$|!xSV*x4rB(3851*mI6RFP@znPWCpS*`2E(WN8tSizZ;xGzXZ^AET5)|#E6^f
zhxs5i4b9y2G$l2)Q>&2QhR-dO`5qQ#&1Gs7T9@}ROvbn!I@)iXCMMG3YiNyP1dw@w
zwzzZTJQ$O`M~0ZQ?;jq~jZ6<AEy&Pl1-y0<r*$|sr_fN}xdQ``LU>`CV-q}4Akge5
z>VtfHB;HYh4xYARNqe?lU99Ww3-I`mdO>Uetl*bB(xg86Rd3JGyKd_!HA+8v6a~Zo
zpnt#&x*h=1v>dHk4TEP4<<cGkqC)CGauE^+LGUp(JxwnF5y9)@<k|??7$%*Lj~xs<
zugN4TfZXN8{QSJo9bceqngXqWJl4agBZ(NjA}so>0)^Ryg9Rb6LpibnUG~75g!4ds
zLl;XP``h7JmAAOGV}bKF2-Ku)7Ok+iy1fr(zASg<Y(?n5tfm>h)Dxw>vn+%bIau#-
z+h3^z7Y0LLf<Aqkfs+hx2p%;6!llJpsBav>z(t67*WeGBOzAPb11=K@bP-)Z$PhF(
zR>9Q*M!0nH9L4-lsXEm{AP(mM5>Ai?lhV;?Mbqa}V;C5?AixI7(ZW#a9-KSFoKHAg
z!NE5t&uKZYi1i#7M$3eAkHSQWU66d*TpEszjHIEarlzFqfp8s0idxt=>;8pU7YtMw
z`_m7>!DJVZkNi%w*v`*cq1qa(3mjm`gGR4<ry5NVlzsd5EkKl@P$RQr23K#K7x5^m
zXue1dh(ufMM?A934*_pDbL;f7iJyNkj}0!RP{Hc6<GW^04rqw_nw6E+W^~c;Vde3!
zUB(s`78uAPVsa|KdOu+eQOo$T=*HZst(KwO>;0pkpnx@a&K*}$T1tLa3^-lAo}DzV
zuCAm$Z>M*vs_ti-eC4+lq^+|81uI=j1yM;xQc+P+jJ3dJ@%Ze;a}LiwuUaC4S0Hrh
z^B1s$DxN8O)m)=CuxYZ`Yrzk<#}l#}>g!q9*_|O>Y_u=&Mu_X%BzLy&N-iBIB%<Os
z*RZwCJ;4ww?~)IFi;n4pGlPw5Hf!bw*&D8wpV|IVeT0L*IctIiUivij8i&Duer6ss
zhLXl1gs<3-;o&gN$`ELA{^%#V9}f<v=eMXi?i@Y6_c*Eg9v0{zrOw}0kdpn~2R2|Y
zVY^;v6QTDwWyGj)c3HR;xL+YSm<2tZ<n;N50^JUM^>@gvHw32R(-!;BJ+;%hheqGP
z9ub|fc5^`v-WAgLHkKesgx?Q?MKG(In;`58-Hime+QhoSUxqn@=E=LF(S+EaAE_pv
z%c0Q^6Mxu68(Vr8I(jCuOJJd;g1J<2kpquaVZk<0LjeFF|8vz#P^vpTywzt)^;yUA
z#R;u1@<zPWK;nh5XJB*+)2<mQe)J}J&%J+WL{qk#Go1Bj9uRzN!h7zHH^&FHLG=rT
zazJ_M05RanPLeqk7QsGTi4#(mI0e`=@iN3vPhr3t%<%<uB(v?=;tfR-J>L2w0U2zG
zqOr+vA));?cUDE^8g|_`tIX$^pBdsHd0D`!0p8PCUvC8Y6&Oo4i-9a{QB%6}#F1%!
zy}u&)mmm#7OtqvWy#B||&L!%NZ^XwomtW2Cm2@W;Ssdjxf3H@R!U}#}|8+(95dvQQ
z{+*(j3b{T`^Q3#SCDgZu%iSP7gA5oBtwox02FUt?6M|%{YT!A27BZx}mw}h+UE38=
zdoSqzX&%bF=GR&t8_YHT`5+LbEh^0N4-+Gf00Nv`?48Lv_9nW2HYy}n!rj}t@xdP8
z|DxTYJ;rH!n-<cK>KfUx7g@vn*S2i_y;&gA{Om#nIxflIBz>>~kqmQM>2`_7v}kh!
zWkB+&lPOGFu4Di3>d%#M&(JSclz86A6pAxsb?Bh+6=q}Z{jyZRQ1i)v9BA;N?2mfg
z@v9P<UH`)cK;96lQ;2inI0^C{LGuGkwkF%9dM}5~#|n<ZRIXfZB|7u-FP>ilnE2`C
zE5|0kh<XqA3pbdgitOJJE$1F?AZ^Q`&U(%}1nLI(>s723?4EW3b{5PXa)aiBf5yGb
zIPZs18-|bX2c&<H<b7xu=tg(ie;C8`?c_jr#^k15SIj}of~D(Y)kj#;xUKtd2L5d!
zl)Hqr9I&y?`XQZz)-~-FU1jBN2p^DmWy$Sm52`wHP#PU*r|h)%_1HU%IyhupQd(G9
z3$$6|02YYH#vP#eZ}s4TEYwKQz=aV&Bs!2mK?2=Iyg;AB`S=7i=)-iuV%r%7XvG0A
zf|QI5d?qCLf1T{*vmW}Mo1{FtqO)BJ`@FS(ewa$yN3wx?iST4WIlkorg_8SMgRBC7
zC4&X41vE}K$bHz%*`Gtc$NrQDvOP#lz#062o*Q=d<{G*$T2J<u!bG2!=ciRJ?={a0
zhfez#IlF7Wy}x+4i(g9sGr+8rA01?M%qX&4gh&tC_sYhV^3+5i0)bPA2<budE;M;P
z{r2p&pUC|Krt9@#+QJf5%X?(Bse7A8pPDwqo1r=sl5}Fb@ZZg*7{uh9%lok_bm=&*
zrAz5Aq+QQ3)DV*m&Xj5Zdl7y~V_}Jc{RZ!qCJE#Q0yccvtY5``J9_+up=_DuzMZ9e
z;6=P@2bi((G2^pv8n46!(9&~%DugpL>Ho#*!^MdPu4rbUiKf8AU5j7It&zgEo!wz6
zZ8_NZhc7$usT>DLM9D(AruO%h>*t54UGwwJ>e5P6H?(M4TN~Ie1`Ly&QIMqi-|wyu
z%mz@Gv9>@!Y>YfZ46aaf{`~B9=#blAJ&73$k*Xu>(sFivcy9%%rTuaGy0>ja*NaFy
zKqYMyN`a!#L4wWIt?`xO*$1{Ja0FzrV3sVig;J6~g4OR|Gk%#jWb&-<h=Srw-O<@I
zKS%!C{<2UK0tgzi?LVeJ-oAYch>luDw{I5w1}Qo)e0mIX8YWkvg@CfyV^{J+EOYU#
z{Y>jzGws2Aw+||+^ZGu2#pummWzBlJ;qS+tu)51^0v-VwIuEH2%xFzWNI)QND=RBt
z0Ag)HArjy}8Y03klK@)k>FFU4%P?7~Z4M3v`$|ix^hC^}WjYJb1>iO(ojlbkg?9-k
z)W+*Hf*2gKht2voK^$7E#G&mO_rEBNft)7Sw9mRF>b8#$!gtyT;N_CAo5Y}3q(Vw%
zif7AQlx54JVJCi!y`%iq@_%g6##@+efJUZ?fi=?xyz9#YZzNEhY$V4R!*nw6yb$dQ
zSWifPY8IirSNf!EyJWoKd>5^&=!<M~{!r4Od@RrH-&xAD|4dd2^FU}?qQVSdf$;cT
z#dBW2QIa+}1}a==be2LNeb(XfEAz5%D<h)E4&E%&;?p%0uX!e??kRbnv(-z{4k4*(
zOM^<fg3nkP;w}`<-*OvP`})rWeLlSF?)?XIy*@Zm^BP9sRUf_EceN!0M^`(o*OtOe
zEEUpkj^y!s@N;(BkX}QT-6EFPw3j<iO>;$x9Pj~^G?7m2_II{A+k@Rj#94~UbEQxw
zp-kt)_$sk_{dF-;{CDczof4f0T<3eUtL{O4@-0f+HPQFN2o1>#tpLjkPYLBYf5U*t
zNEMdLM{JCewcO^JME`aoT~%;{ygH*}VHE)ll|lwTCo4<iOF^HM;CvO5jQl+|{1T<O
zz`(YnI__qO>6xQ|^aGF}tbZ84DxhA~+~A#+%BKTG4`I#g0c?Z3?Ia<g-QV9j``asE
zd+jYIGe-n6A^5Z?S@BXm2}K7=qUzHpzp?K@Pu8y(5%qN4flMcc<&!7#lbH5)2LE(4
z>GKp8VvMK+WIx180Gneh?@wxZRz-uWmi!~AQ<dL8X7W`Kc>;IATB`wgbLl;E5G|{D
z<VXIR0>-q(aG9n6aQ^^Hh7T}X58%y0xXYkxnE4lmqKdxSG&MF(?e&D--NbfqqQvxf
zgyA{(_>||Z2R_5S9J=WqI;Zv(M;MPvpyU{@bApTjPz2KW_wO)gRX~3zul&+lc#+e&
z9{dtH1OSNS1X~$lEF369zdqaoVDUXhyl`+tZKmGU=)(1p>;uve+Vge5KoGflb9#DO
z4tR6G_j~<1hUX$bw4*@q2{5td{QZIIuF^ap+sI)d+X8l(o!urR%Mc-R16hgk)mx`8
z&c}pls5Cb0HKD_ug2qSks?HGc*_4n*{HvYCLM{hPx9L!I7Xaq~+Vc1D=_|B(4q=Gp
z<Fu|H7Q?PkjsWZu(u67KFxlN)XzH{jxSrh6uvVhMX`k2v!_UE$D}R><isIw1inlnw
z6H<9y60$Rul4^v+6mmJu=Lg-SD$^BmfLsI2OkxT`uE^iOJQeiPc2a)W7iF$ErNmHi
z#Ku3|00=`%E5G<?`+xcl_Kt@;)0ez<0CaB5`o=ytK3>h;{Ql|#)*sO5%3sZmK#M*`
z(5f|QGXg!}`T+M{A|mwC$#sj?H|C}~gweNHaq(2eRG82TOOpdjx8DV>4NL&K4Jc#^
zF+*&~yx3dRGUDR32pp&dZ=x9*yMfJ(zB&sulB>+j$+{k(Y9F8zJ}@u<87ow9*)IA5
zIU4vG2Qa90(o8u+t%my%K)_3XdHMMP*??O1hYzHqzSX1K`)eRVQd%dbIq7!!$a%U}
z>3#7v2bmQwuZYAKhC|<fN$YOp%zuCJR5^0tM~GKQTBy33$&Zg;)A9!Gx^4}ASC--E
z<d%7I4{h=E+MCbE8V$^chv~JcB<`Ltd`9RfLSHWGb<_LybIU(<OCr-zt&aI_M~47!
z+Szur?6k&;iZU@J!i$A5Payul#|M~htcd-hMq$sRhYv47nGcQxEQc2{4G6;vNB&aI
zY$YF76hjX4b6)DaylJn@7~fu6VF!6E)CM#~t1HKCpBgazP1CGg+&d?xJsnFKl;m_k
zU-e+l7-@pMOAJ?fgoTUE8}&gndhgzA93wNc5VaW}1i<~552A^wdH;@j(^H^>8+aiu
z0Ho#jilVG6T-^oe-*7Jwy&|)Y^M{;F00p-z)|~qWh8vEE{am+MbNG!A+18#Odyx92
zcUgZdDT#|mY7P7YmemsUE+2z}AaqBPSny7ZgT-!p3;C0-E^v4(%61GP(os}YG&G!o
zJQNx60Grwfsb#bq@n&k5muc+eKwYHQ^oP)50Rv0QdiOMN{Q%b-0)jpH^YGA6J{XkI
z+1UwAS1)JUC*hESAECKzonUhZY3*|-IF11Lz<T`r`7^A7@ZNk5Qn+l{>2iD00st<O
zH$#0X7cx{|bK?8{`M&{`56mC5Q!lJdl@sVS<y#0uoKwVkZ$c?IL20aF=@1>Ya`a(p
zeRF#Il6{QNY2sUd=Pa#ZotKYE4c(rs|Kda=>;Fz6UcGzKKF?+zN)*xYP$uBO#C6O7
z-~(7jI}|$5fG-77c(9pwxz59MNN7`lQ5N*DC7{vD+I1Fm&Jl)zxVSjHBC!bw3TgRF
zRiPvSJqCHS+W<}Y<B;hKO7p^8_U(sIK8WFc@d0}3p#FxyF523|03?ZQ_UawMEDAd3
z<;r*Oh=Lm?{_X>Ovc5RD*C8hg(9Xin+Vn6G9$|nn7sOq#rz`<UxqtsY;rYu2u%=mM
zyR(_ai`Fqea02u}T0jEAt%C<2TL;(y$!>0$bZY?r?$0Zr6iUL~!FGZ^YAy_*oxs%M
zqauRmrV({5(aHj<=5qKAJ*Z1SLsB|Yp;(s*=>R<82pEx&k&&0Hzfd*4Jd&ZD7Ee+7
zpe9sG1YnI@wC9O~6ddjqT54hEM}y+BlrL&Wt9*<9CW+O3=jEewRb?JLw!4>XLnCkc
zLCKI;&Y)#*#t(urDB?gO*V^3tNKOtBy~i0vc9D^i0`ya6&_uyeVRFU|_VX>l@eGzm
zs6?hG+M_RVKdS&6gAC3bsUU!VlK=Czpkf_xwV>t;h8AhG&q#z)5rR88IqgEl1p2p-
z);x*n_wV0>XEixD1I1#9gX<d_9!g0?0MQ)M3+)9nq<YkqsuY0AGV8JUlRL9KT1}gU
z)Sunm-63hvS`;y!(R>vM%|BX@&_XDehe$6N<@YDc#V+<nS}x@EaH{40g~0?6FeOrq
zn5-u?7OvFNh)qOVl)bPeD1!hJg0TSl=f5JDFfAM*BzoLPEo`jvj7G?1^7;8LA(X|B
zw5(8$@^UJ2oz%Jrgm!-SB@wOQI!yFNCIjT5%3+fN{SW$tV3CAx-+t;Bt>&Kf0z2Q8
z*7LGQb9w;6K?aO+n$q55A(CMz<jj^&0nQ+nFi>Od8wXQ|-9VC}3Id6!oJCl!%l#v_
z0L9c$P-use)`DHJY%&c5%0SIOgyK)cFGI5sDC-^Or>25s*jQO<b7kV<;$mQkhg!;R
zBgOI<)X&Gqqt*Z9z8H51v@1p+CBf*$l{$L+#SQ?tK&ZXnJRjzTut@n2Dc=6ma~Aw|
z`Fw2p>l7>Zz<_|5+crt#5*!sm`6x<~3J9f8`|UZ?0iQT9?Oi?7EhC@VHX)&_P`XPT
zm^-NE*7Rl5#Q1_coArrpp!?E_y6(dl&;X(X04upzTR>%6YD832I!hxsI^*?5@+Ib;
z8P4lUNlGO`umf&se@jR}s;4l=doQ!{sfI=r?y`teXJPh0&nZkhfOueFz)``W5QkQ3
zaq-Q$o~7cYvi<_Ac)J+2eo;3NO&4?(OJk&@)7p_yfJ$=Km<Z?-WEtiZ!pqgs3A<db
zba4_exzxc(R*A@7@}@yYQI)J9&|R+^;33hFF;;wYr`7P)nr-H$gLuwq+O#Ej#r_`D
z)P{4fa+04v)>Nys%TSH5LzRIz2-JTdM*tu6tFf^WH1Hc5W+0i-*Vp&g>MLB;NQ{e*
z2epNlkR`G10*v?y5(%*3B*M!TaOmglUctYtsj<vp?E|*@>_Bh&Q>HX?&{~IL0tDmI
z0D|G5Ey5XmI3OQI3qSh*6{iEhY7kLzSF~w2)lQNYJSa@vo4hRQs*$O9BO}Qo^>T2-
zMJg%>5ZF;k`;82&Un<+nR8M-7CO$*qB$n;1eY>Ky^|no|F<H=6nPcYXiBjeYksv(p
z0^A!ty&K?4VB1YjjpP}dA#nxH6f`*O_NLSH0NKTS2T)B(iL2*1^!x(NWga#vHUUCN
z!DvnhCdOz@`}+D6u4X?sFi>*7$1x`^7IJvB22ClCFB*WLyj}5u_R($_?i~$(izB&a
z+9&Peb3=Jq!bolF?~@LGr68>=`zR@+0!T3HOzJs;z4)J`D-ZVF#ORWoRd#VnTa0><
z&DkQthH(=6M5olHi4Yy#pT7th0-x6p(cFoXCtuX1<|+N1n3yO!w|wTz8S!{42wtE9
z^e~Z@ouX;BD@{xvI&U#>FgoJI&ZTul#KdT2K8JW1C*a%gVsmj&J+mJop*)iz^*rOm
zJgkI-HVo5#`t&JG4k-oYp{-Wwr>=LPvdl5`47|#Q(*abf8V41bF#r%O0km!fw33Lj
zuh!|xCR$oUN|qc5_VkjtP_Jd#xTxc)ZoVH*WVZ3t$a90qi+xJyOFg_P!C7+AatLoQ
zg@dC4aDZvB{6G%s>W<CB%&NJ=;E8$I%Kdc^#Da^Hz)loS$MgeDspdR1BHjVz0wbf8
z@hS_mwnGcIJ@vvE2q`Fo)XfWP3ay}CGWlT*z*kNTRAc$6IAaQdE^r|Th%Yap>IU=8
z*;PKuo?BMRiqj5g(4U3x0sjX8d2~NCB!qFOZ-TAy^729)O}tBvWsQ>467;5=YHTwI
zTdc^I(9k(@R<mmGwSgA+G#0mj^pXJtavNbXFPgdL=5_>mWxIL)ux*EonjVBGrl#{?
zs-cA<U)`b<3ew647>M$@*xA!(ej^FekJsgp*g~AT10#9ZDg06aJXv_PJ?j0uIs7UH
zj(1GF*8q^Tba_4jnzITGF<=AH4Xv%^u+1Ul2mg=;mgjc4gSq*lS^i*B^i{Zc6nI#0
z{v6~|AtnT_U;?dYHkKm5-2lRxoSba(YQ|(1nit7}fE)n3`jOD}Wk{>woE~q{>-sAb
zVgs%Wjht<csVODco3YwMq7Z;Xyr5M~cQ?7C6ud}R2`3$K#G&rVVm9aUjmEDXf7}M-
z-g;~4+2+MyU`6DUfr`YDKM+?Vy?`n)gw}gCy{^6K4JQ|f^ty~e7^n`Y99r>T4h{}_
zGSrc52%<(X6t#7Arw9pmx3{6_WCmg$5bMGESZTC#VxWoz9&&YM#SyYfk2i0ih57<D
z^)AfJ!HS8wz^Vll5A%X!fUwR4VhBB5U1(+oZ^(&(*YbT_`m78h(_+7rDZ-iWrgam7
zKq&0^st=@%Cq<e^+_o_U6b={<(5D48s(yHJ0K?YR)d`hJq_jn{GBYuWMe5*}L`DX{
z#MGi{aP#2bpoMd9TJ-2PE9)#wnX$LRI^5P&PTH(co;<r>yGF?uTYd$j<8(ka`eozB
zO<0<WTYyG_BvnS`v%k=$6TbqgJl&hF+AFAi#Cq+owbBdOrTbdjhLO?Hl|k%q1Y^Sx
z`W}c<fkMd0($X-%Admn-Dw7HRcXAh!n=eq!f*n;0aj7k&;@jKXV1dim+oc1FY|lSQ
zdW4YGAoAFUdJnYrAi*4zax&K0ydKAkBV9<a4I#A4*q#UB7D^~6Lvb3S#-8;-A4y$Z
zU9c&Tl);933HUTz8-kR5$jCa)W3px<S*Q1w%K52?S`5}`SQ07dqazfQ5HD~9t{Pqd
zT-l{YlSw5TT$$6hhtN%Ge;{OI6o3eueeLbuSL2a4Js}@RyLz_vgWW=#6u;!A@od+(
z?)ge%CpJvtT<QPhwlFJ`{RxYAbabTftLd<LK?U#a*s){S34rl|EC!l6)gxfw5*`Nc
z9u2?sLXN~A{yBtSq3TKz0#qMvAd$7kY^-Cc-@(|UXxZdc<AoMWK3TXCyW&!)Vt0_e
zpt2Q=w)@zln>{f(*>!Y+(_X^fWi>`lx{EUM7Aces`o8H@1D~yt<A-(&BobS&?bubn
zjldu2+oq2nKL#;22!qQGcTMbDJDC8Afd9rI{_c0<(k-J6D?+WuA0b7O$F8of=H{F#
zP@8FqR+56APP(>v>lsb#VMth3;+EZ=q?N0c8~mcMO;C9sVq*b*VfCoA1gEI-r?Rzm
z2?%a*9}ggz5oygZ2J#n>mYksZL_`G5pOTER3y`^}qocEN6C)hAPUwJEpD^IhzL+*U
z*O%CJLCg0nfA?&rBy#PIYN+4+nu7C2K~4^Uh(_^`QUxUXruK)ElUCQK+4q4-+Pq-Y
ztEk}0=fkci?nlt2F2;D}%4Y<3gzMkGer|cQN#)xsuIc$RIbv9+ojqa_-Dy&|_(+!*
z2|eA#i%^Z%=ca-alJ@r!8Cfvsf&mQ;UY%2<Ru{zdv2fr&ew>)`DUbk3>$EOGNL<k~
zkd28Dog!(EZRw`#<;pqcR%@vK{)y8!ONdV>h7s2-cj=aKCs1qpAjL{8hO#fUe=VE_
zr8MQ+DobgdHmxE(@P8m%D6)v){KRMW=Vq(04Y|gZWUIWh2aqwQIVdZ20nV}yq=kVz
zlQ0|+HE(^*=*Xm0{_nYs&2Km<p^~7GV2d2T`ufk;N)HZZZOoB1i&9ea%Z!5>DZKX2
z5m_&Ise!6llJG5i(EJGadYMT!1Ax{-&Z?aA9E<%4sgCC5^I91x^4IGr(dtG8sXZ#r
zhXsABqlLog8?23fzl60<ki&J|`%j)YVWe`=F-kVMc#PPFSlY$(3pm03!6nx%a7V^H
zh9ky_tu5_zDPgx!qT`A``Z9g%AAG4ai_WycQ^Ylb(%6=v_*A4Di^H9c&WuT&tn^6b
zymzFh(?+>7EygIH{Oa19Nqn<w$|dL;Rr+z2AoF)Dr{?P{-imVP*<4l(Mo!|bI2CK(
zGd!A^atxT}G)!sp=L>PG>VM1Y*?P5988N^7y&mdBB_ONePHiLSp0Abxb@R~&BqHlz
zPHyK9Th-Bm%GVa`9iU~Z6TazDFy%Yw(?CD=c%xJ<-PokAw2?P|d0{0t<+LbmSRL{@
z@A?-Od1_%@Ar3%hH-i1aw10iVJe>?YY($q3=p<m|T;L0!0B7wY;J~aY`=i$sqwB{~
z?sn*LbDBr08Pnd^Go3kdjV>JDPs;0@nOOl7rUlwxfyfbvY(&S>ia6y`{DE;**3;4U
ze-`X0y*+x`4TR2DXn%J*d^31BHFNXNxJrcBEO{t?lYQm`$G22|R2Dh}<jc_LXcx#!
z1Vs1XMbiO41j>t;?+^+Dqk9;(PM$45DP7WRy`?U@kA8HqVh)b&pCa6S^$dYhP?kzi
zC_}6vMG&}Um(BS;!Us$=s}M^`bCue{{rJ`+DIao6sktW(uKR724&-3s$Ue2PSKYjc
zV2<u-Z=fUtRe#Kf_wVIj=7Vn*5*%kuNG<B<!^A~A+8ntcu@*HSb-!2>yLk4@;?(q9
z_6&Y;@tw79hb6cZr%tI1W`JeHd;pTbnB}r;(Y|nEfh2Yw{aC@S@!)yRWRA@aW-OkC
z3;Wj58sACvdl`B)fosf(0sd@yQDpa3)>LN~YYp1Vr$!cJ$>)D)u;hMC#R*gW;r*k(
z-rS5l?1(YQ5z+G+y~nYu|Ck4Ovs5D%3=2YO#zMJ!1o>uJroUXS62WR3aUy>y?=Ilc
zT}!0s1gT2<+y?AqYGA)YIQ}HN8qfh(m~LobXt=w(i@<=kR<H#mP6|zvZ@Y-q{PD<)
za>W<f!ju&F%f!LGBk>e?O%P~%29-dNQ1A&14D|5<RkhnttcHn|KpV}CL2x8sR%*s@
zmjrRLJ;7p=bG*YsBUHrBfBcA({H8^3X1k$>1l%ag2j3$A%evL00Vd1=AgEmY$C?%(
zk0iNRfQkZ(?@*MySI|gjgOmQsDBPGuWX^TK^!m3xu>`-Yz_Ert<Xu!KRf$N)L*_gJ
zJe}Z=A9I0l29W@yjRW{PWgNu7$S;8Q9^Aj5xzB18ZkwEgRTBG2VP$nxGLCYl)DC{0
zgfH=`Q06a~bq5rfBo6@i(Pz53m2CSo{&TtNn>+1o;j;p$r6vLg{LOm^-i|Wmwwx+N
z7plG)dU5hjMuP3hvb3zgQ&V``p+XCZlMrN7!Z&VMLCyndkyGiA5w9<9&|L1(BVp?a
zT(S6T%J$n*za>J3?_0=!R7D1sqfiANua51*=)-PsB7jSQX}`_PTxXtdvXGkE5+hO!
zj{@+5`K8lZb>=;jQ55J~SL8QHe*alD{W083fqr4Zv&+)?4UeevgQyl#w}g`hpfV!{
zGFf_n<JNLyB6!<$_W^cV*!^GKy>~p;{rf+Bs;<%y$|#|nr9x;JAtSRvDnw?Rtdf}#
zU8zKIiINdfvdP{fr8MllWt5RoM##9IuT$6O`~5xc@BRJj{^Ne!{ZnzC@9`SP>o}gr
z^LQRQqhd3yoF$i7nS;j|MoV~OJ+3ZYd=8&DIvHtc13d$Dk|UY`z<&|<a8(RU%m5q#
z3r1eC{rb13hzD}_bLKs`mi+hA8oDVj-o7Si5mDTfNg~NJLKUd!?-N>eWPWf%Tbn7&
zE34A=ftaWP>r<>zC(6OS8ZKOzCves8-_RXw1qKFSVhVg{eIM(*CAn?waHEVvufNk+
zuak8(JYr|_GIKFKxQFmP3}h8@IXXKW2L-jXZr!?7h%g=Q+^O$OaPg2B#FA_tngp1g
zO&+n`iXyZ3V^yE#-=XN!0u@z{#-Yi(+>R!QRt=>dfb4!Bg&&Qt&H_2D!hwU>>)5e!
z)P8p$YXg8BkZat$u^t$DF5;r1wm?HdQ5@`Hm=p%26#RR%`yM#EALjtxMezc_eBjIl
z`kVbejH7>@XDK{KoLueitnP-CNfa6B-!xt=ji=y<0mxK=dVGLjp`nM;brR~q=#Mwd
zm)yPkDv*1aLqUH2j};Z_K%~VjLrTs~*M!Xza)-LQ24G=n`0vP^XsuSd@GloYl1eSE
zvE~znTpUI~X1d33V)x^?UnEXbRdhtxSli6CT#_D#=A3ujoO$s{hL;-q+T;pHY`Tu7
z=<~O`sAc!>T)oxq=SNo2fhbh$fT}(@zAJW6*me37f^}3N=EG|}_Ehrn1FWU&FLySP
zn_!4Q{Od)d-wk`txu0y=TL*+nOKYe&MVB<9TeWHxL~R_h*Md`5P*9K_dO>if=U!mk
z(9bhPAqc{Pq?%-XgPW+3ralY^(6-G)MT3&!z?cvErnB@2C@E1Rl$AXb$5U;h-<qG&
zI=ettY9ia4I}92uynK8gs?Mc0p`9*^5+Ylt5GSw|^-uU0&QBy!_qxtIzu!28zH^_>
z?Di;y7sH~2B+I?AD@ei*qamjPxPVDXS70~hr{2o~Pt0iLL>TXc7#8S@*6KR-13%tk
zONK~&4wY`%la>nS+~|ZqiGVyW_k8_oD%FD(If(jTK)%yx*$hvJU>?e;9H{V)*on~J
zb*3f^_`T|O4OP|8;MOhtDqN750UW%-Gr}aYCIsy>k(-f~Eu5WoPAEB-oCnwiyUV_v
zlK5!d6Is_AM~)nUX+>lQi9CeBz@<t1!D~r@4<GJD;R~kJ{q}7j>*m;bV!;V7T?ag&
zBQalLA_A&K;ecf##OHwW8Y-gGtB!3ZvfIot=Ypj<BD{Wo+ZuAD(&zxCVxE*~>N6nW
zv7L$|$}SCgq)e3bDS}$Y^*}6GP3m0yguR$a<ukCobrcbe0%_aqJbLiGhJ=KK%gi`J
zNmO(V*2bV#^=`Z4U;-i`388OnzVJ5MuD)R)SqVVt@Z_*G;i}?fnMcT4-}gY@k^?He
z2QX7$YY?JQzc9`w<fVT?<zv;#l`$ppSU*z;>%b9IQBuO$MScmuG;E)-_asppMtLXz
zPNa24X7%93ze4#%tCh;}<NwsBV7(&sfJiX6YYjv{mosvQ1O+u?hvEq3gcuOU3=JP`
z%D=%3Z~}e?5*-p4xRHe=4cIVg@rcl?T%yp<bpWkT0Xs~<nh64NhzK5pO9PCE`V+Y$
zEi2A0I3snQoP;V17e;HmDU$Vj2xC#yqvL-Q?t=E?%Ng;KPKoXFSB%b{NQjDkdVT@$
z8Ygfkb7tnJ#{vQaulJNPCs<)e=qe{<`OF|CL6)J?+l~%b^e`sfD!XF{zz|h)@xle7
z?nFeT^RB1*_^AKMO01OynG$guRtz@eBLoL9*+5~xhB^&@C@&|K8)cWQCpt=dU;Zc)
z@9-!Eq;Rt2Wj~oRST`K*7;Layy|D9fWc0x4#_D$Jj|>3c82?HCwE=3BU}<6kk(c;5
zok>a=n6nXQa|*&ST<?oRSSzF*$h2;T`E~dUNa~U&MW%Ky#{=XTnzi|<4<A2%`DhsL
zjhyN?tOh%L4iN{?`w9r2PRQCKKKAnRdRkZ+vpO;KrYp+B5F}iMs{qOd@}m%~i$vJu
zZLFnbfO0&<;O5CXfJ?)ih9%(jX`c{EERy<bDrJ?Oc{qF-IaWJXHfe|P0LM+kAEHP1
zyB0RjS%N5pWgZIRYG`PC@_7~JD!Yd}wr$&GRN#z=4IqzSiaR7!Dmk8Vonq-l*kFlp
zezF_MEU=r26AkddM5PbF(pj<K9jZ@8ktH{C0oH-&G3;ej)CkUSpl6BN84Z^AtUrfK
znu^5zuNeyF!=Soh0|0_TlT4u@?X;m8lquO7E^kMKDkmq0Cl)AI`Mr;bEj0TOt|Op)
zP$bdQ6+|~z=EJJ0C_V#a$^N@k78<m(3+mdS%8l?lU@b&lG<wMiOZ|y?*KBqDq70vY
z#j<56^+C13ThnyF6XKGRIoPeu8t^qN+ViH7y{HJ0c)s%^7{DQ!dY$~obCx9S9N@Jt
zedOU0cI@)`YcY~ExK27v$>F`s0G2W!Flc!hhS*h^(G|R$P5_M3U||2$8bsjx9y$r9
zhE#@%u4S85ppW~T&GTERSQoKBf@9>Qu8rieIVp(-)%{6ZG<OYSN2mKO!2+)Dx50rB
zdH(-MR`40HfBy98Q)y`*QcH$Lhhx42NCDsV+hc^Nhnm8wEHje~a!*CoNW-HncX5Oq
zJ1mBJ;9e!IIs&k!C9xupZ>Y4nAv<S@Fct1>e}6w#lP+A1Ci8g|bEHoqU@>&6#4baM
z&dlggXKc96=bXG$>N%X59BR7VhrK55mu;s$b#a-f-39lrs5gCVk95thl=eG=n>KWQ
zOk(`6<jxMYAc7=800s&XR<Q5dEJ~|K(Q19i4gt@-7fS!($WO65AF%J~aF?VRkTc-I
zP;}4uIkF4&EQ;RT(z?OFtoagx+5j)dH5I;0<^RQS+anEt#;#qv#KnJs6#?|$*Vh--
zi3GPNUA@uj_i)qvG!Eg09|F3(*|1<oqF!hjTYsH3p2Zu&79iD$vnk~FPS2mU8tQe!
zz-iUKd-e|N_uQcfQ*k+x5Jtp5h7uqm`i`_2jx@L)_Z~a|(De<*H%8WVqUxgYz1Snu
zn>x8zzc<qSz*w8{v8PoLO^<$GUy%@$ld&X;%WVXd7DSNx`mL}psVONy?x87xq5$TN
zr&F)+#<oW}57hGIwsCLVc-pEd8~z)$OGnOD$;M_J@F6+YmBI*@vo&iEmD+W41$T6R
ze2{2d&t^&d!59P>-MHuN&E^X;0gFV^eaIrW_-<7Jmm3z;i^%QFm^EeTFR|aL-B=jX
z{oY6FtDS_<d<;(3GiT1cp&!K#ZqB+fvxMPx;&)0O$MqkD@6x)exr8-|Yh6^g4P0Rv
z8N;L5&2$b4{>|7+d9cAaO)eQ5A1(V;VMH{IZ6pnj9rG@`B*4BaomrslUAgC{`o>sN
z?gk_ohxBQ!x=(@yi){GYi1jO%G9EZya~giYziAJl72YTa{OHdI;^hsXRSOeIuJ=j-
zriN%&rI$^n>ustp(1O@pLWIwtt7*RBm1G^$J^2IH6KiG(v}`)L3pm!7+wA)AI~81;
zPy9XK^=*H;2P=tT^q+FR??mLj4pg<L*Xtj9pv23}o-*(0?Ch+YW^y)Nx2C4%lk}sY
zAbzTMiUr5euFO<|uv+=MyDx);B-lCXGSe@r*Rpw$tPFK8d;;Cirf?Q}bI>K=WbW5|
zU`M*hK#gOGg*|#S0iwJaUHkQgJpbNrpi1li7}A7>lwQ7}D@^(|@w$7ZgaQiZ7j&AJ
zBngGQ3f#a6Y9Tpd*Kb$N!q!y}*?+<|^T8(ZcO@lr7!F&+t&C5**Jjgy6CCG7c;Bz;
zMv{qj*i|S(yN)&pI7ZS-&L$LRvvn?Ax)iB^!AUDXzB{R|zgs#!ro&w@QnbwOWgAkl
zsLE;;8^O}jUQ@veUD^XbOSe)<a{;jPbI;v{Ek`v5zleyY)HB1m)0ct#>I+Gbpo3Zs
z(|hYnSK`&7hd=+<tVoq*l{HEv4RGoq%Gtodp^J=__heZyt9HLZkEt8GFAmgL<jkS4
zkuZ)sRvZi2Z-r!&cMG$X=&KgnD4>F~vvnyo+i%<buDF!yktP<n#Kv;UgapIw1atrS
z>(>H~p&`59YT|7#M3t44faVQ1rw<~Hnv2?+2k!)0-xB?1?0Aj`zds5jJyCthd-QQ0
zI&ACuPQqJ&;kbfx2FMUtNe}DQ1o9^#He2J{2t|Mca5X*6em**eL#5rr$43@9S76Zy
zfX3hwo!-T!rS)s5<rjEA6;JjkD58o*T7@CowET!)aIh0B(Pbkc7B;qWl<UInWAjDe
zBWYi=WEMQD-s-^)QKUn$jgG0SH%wMH)FKexPDQ*6g|xs)aU!nH?a)d0E!h-VcZsW=
zZXP7H&So7aqpeX-%J8|lLjwZHy|wi{p)sQb%mW!7DiudkZQ|4$Ejjzi`{&X<JOn>Z
zQ=IQj8m;^N4a0VIVo#uIX6rSt$8%{*M}uG<Ode>Bgf$FKG*7hPbS3L9PSz5K6#N}*
z50qw<fW$%^?mKsG{QR&3m^1*p@b1Uyz}is8pNQWgI)yt-@;!J6RBj;u*Pw080a^(h
zq)ofe$icY7l|o`2ixSGUh=VggS%&`)KW)Fz<)r!e|0V0<#Pt74*5#B7uvLUI3jo2I
z=kAst{Xue|(I=k6_ehM7-<5B;@LqP_3W;|((swt#TqFEUIPQ!#a1#%kiLr^I5Ue>j
z$(KVBNK_z5v=eD}9xH3UCBW_G3Uh_rr!5GTL6+*FyU)sl@&<t49s~;D?8rwEEoC^l
z-r-7bL0Cc%hyCCHr^_B+ar$NXyy}U+zYX9)O=}MDgfaNK5Dfw*YV(Fc9L(K2w{N%M
zW{Ara!2kgx(a7ucW%z_Y!Ky(V-)xx)FyTIj=l~?`9YDq5FQ`E%a_<n<Dyr@Dew-HM
zHd9(=#sho?M<pD1ApDpqlyJM^5uwlR>>~U%V4yJO8WhN!ZoTf*7b_In`N%Rl`ZlsS
zgw6wQ^?F|f@9^<<$ijv=P7F*SRTUtWUulA3xtu{pLp9~+d-m_wQA8QF0GoNVXxK=>
z9*Yiss#4)5i=kFEGv~<*^(&0a@0Gq!&|PD(t}iipAZ#+>f}x?ILBl_JrVx7oi0x4d
zzaviLD+?{({C}il8rreU(wodlt_Ed4XZ8jHBO2_#C$b?tJY4eWF|E30u2<$E$QKRS
zIe}~zwWE4>pN-^D2m;IxXyLN3Qh#{K&`ghX*yGqMTxNZs*MM|_6ko{m7?m{}_g=uX
zz(MEBY`h_Xx$6KWFa;v9<gcz&#POgu=Q~c!fE>{#tr5V&!h)0`>dEB=3bpN9?T%ju
zrU9aR^Ou`Qr6a6`9zVjjmoHxY)1D-g6JO2h;PjSPuP=Rtf%Kg0daNb6^OyDpHfLTL
z`7v&&F+DvE&cfaeYY>vyM{>a~5pEU&+zXd3)sMBtW&q)VmI%O)oWvE^^xQIAhTz7`
zAk+6$4<dI6I}X}2(2ay!glkVY0za%jr`;=Ob<e|{W4vJlgVx+$+>o+Axyk%e>UAXG
z@H*fuG$dLi)xwEFC<%9Wz{*ZL<HCoBn=G@!g@(86d@l_V^8y!a>_BKofi@ll$p}0K
zxSX5T+=1c3x-R_rQ9$=ES$!`5;?+mK{KcGnSH;sXNt9bOSh#VBT77$!9l5XRIRunY
zxkEYU@>^Jh$)2GYOqe>5=4!mnoo*ovf^xz(Uf#MlI@qM>h61M=z{EVP8}%n>zbKvJ
z-L?&EG1qAj2Jw4L9#W?Ws6#_SzC#9QY}|Wp%=gi7*p4EC<S;S+KWBK@>Er+B439<w
zd+BBUwt-gcZqU}4FE^$1!d{H%+wYN)xq+Y%2YCo;AF%}y6LAAL1LwV>g!zkA&iCr2
zJId)~(&KGT2k;APlb5b^(EJ~hJ$5OJ$sQa?1IGKcwL1Z&u_y%|xPA@CcZuG5zzo3G
zfL(C2vq!}ap!kv?X>tKiBc)#{UAxPPOS{dmq3rESV74h(B*4qgL`U(T!MFjw1BN;c
z&%)_U=fxz3L8QnL<B-I!qmy>D#!L;42hVdpSvq?1t4Fs#t!}^Yv-T}{*iQbhrV;I+
z28M-ou;vD;z63+14Tz`E+7&+koltUwy`@6OB5Kzu=6>uy8_ylzXu!EPAKoL*ubm5k
z8%Yf8=aoA8-Htg~#bB3Wa?<zNm8{>%U7mSo4RJe&tc(2pP3b!`C-GhIE*hy6S=5AL
z-;aPx@$wBo|IZZ_Ny?ulMvr+@M)Y5oTixXhpiQeBSIlX*Fsw!hbOCrg^&<vH`A*M>
zosYs?p+A8{uHnhP&54y4AGPxsTN~aK=AR1~`piXGl9c1(jq*N0$F<t}&6!WQxYRv6
zlTa2Pd$){@`JUKD1qMg;W3;qYqG9Xc9gfPUQgJ$;spou5wqMK@toAy(E0&*XphMCG
za*yCAboKL57&ZuuM$@@W@(sP_UD0{;bMw+8m1OP7web`^U+$kmx#DB~)x?I=EovFT
zJ}I?aOAeUIg&El04YB^u&!q9QXX9%nK<{KrOpm7i_ByJFT$|ZN5yz{vl;Ks|twqn9
zJy$*Y+nG7}_MNzyYSXbDyXvjQf<0^|C#tk=+}%PhUc^8@8ZG2<s-Mn5H1^@rQL8_r
zo1^HaVvB9To~)$5>tMwqq946^nS*9S31!6jZphXm>y<PRyp~r%@P;uwMsNvxIe=0-
z4ZS8}DL=G1D<Zy}{FqVMvNhyrCA*i0;R(s)CA7h^jS-`}0fq(qH{_cTTG*eJ89nJ)
z_EI)jfNFaD)xftXQMShKg-^hKE59)Q+?|U&Yb|d^<#SydiKcU@A0k(D)*Lya$^3<~
z_z?L)2}Qzs8S$1c>8?b5vg341da+Czjl%I~K(}nkENFu|Dr0T|H5UWgc(hOdB{lpH
z8FROscWjQDUIn|5j9XIiXTpszB+IqxNaCRGLdpv9x@PlO$5$^D6^qAbb}CrJ@{qbb
zN~{mXN1&$n8()bM=G$XkWLop=mvz>@2CTl8^WVd0YgAIj?fs*EPVpiYQF{@|$qM_6
zG&q)0i|I>Lwa<xPHWz7MdYLh23IfQu=Cg9<POFPESOWI^2i5<gRpLaeFLTfj{5Mae
zvY#XUXs4RzwFI9J8tZAV@a`5K<hZAo?O)SiZ!u^yC@1V`F_pGRzjw&kXtToU)3nQw
z)z!U<meuL4UOM{c0Uedy-m~_lenN&tS&}1tZVYTgAE_Sle$Qk!v`)yQs6pC!>Q!v4
zmOVhzClDc+tGp<6q*B~k<Rxox`ifQ_+D4vu!$5ZfBrQy(EDF1AC5`rcdDGIzDKL71
zeW|mlRkFuU>eJ{mi=RdF5m)#x00`4xd_Qtk-{QNS!SH3~6HAw9El?5?bhLbe1gY!J
zi!2|U2o0f4p$G@86e0cUI+9xYtv(09Q51OUZ9SpQ=wj)8F7=*=;Lom~zi)P}i;ML<
z#JOT|Gz+pgI$;onsD!BsCaToer6Un0;za=+vg#`x$xB9G5lVO*7wZT#k(m&_T17TH
zu+Q~Zud=g?lIp1zTJErR(SS%>@sXw^ko@sF<lvAMOs-__|IEHwek4mlsJx<LsWn*@
zZpt4bDOdIeLWoM~0)S}CAjUke;c-6Z6Aa2uHB1RgtW8EIsnpkwE@@R34NQv*339Q8
z0OJ)ez?gxfWnbgpDd>+3Y4jn*+R~kTacU0#w}c)kLMxd<vSkoHs1%}QU<-^O=scEq
zeRTiYqLz@gMN!u%w6Qzzx=yeXiSaf#((v7Kke>vAh<ZYl&V<8Ud!C#g_>PmqfRlOR
zT|XsH!%bL<(oq-iXBEt23`erM*S<d=T=p1>P9Y)sO0G+smquI&tLEOH1kBbyvB<u6
zElr*)S_Ptj2_9UDM<NAAJtLGK|M#q2?N-Pd!oSk{N_XY?e*=e(`s99VGA|@d*Y@{4
z)`1#?$bp^NY0506pJ3B;IwJPuFvZ#G9zXTpRjaK|H1s-A_#U$N;*q&ED;oMgt}Tl5
zE#rKE1n<gGa)IMX1AIkS+Hxj{+l#tb@+pZ&&+J!?B4PgLBNh76ONFUo!9xhMfDVr#
z5D{+pLU;gO-3kL{yRm-D_4;Z&xp@1s6=-sU1IQJ(L&X6F?od*SJ@(CMIQxCcm*eVK
z!{u1RbF+0t+jWDrv=|>hQF;+r#bjCObd9#uDyJ}#!1xj(e}N3w0Wv&NcJqiqm3rWs
zhy|*iinE2UfI38tGiN$s8^Ql>-F>lEx-C$@#FUxYNz+>8yz;%K_*nTx!!WuGX*AFs
z6kh;xKrI^ms#WU<oI^NUsQRk_okdAetB!PDBA5XVbU5RKGxtw0jz<YJ$rdJkryQV#
zc9duH*}89s0asCxYEl%;_bLV9Nbv$<ux206h|f=@P*#@#e!iNT8lZ$)MgzNy$s)Ed
zHk1PQb>n2LJNICJBAerJ2<HHoAVieb`N9)Y@k&Y>LBN|(mt#%7ht#=~BA|Y>!Xy*a
z0mK4|-o3iW5^kj;t#r*2GFj`1_nNhzbKtJbZ-0+y$9!p#&s9%!HL27al93NS^d$!p
zODT6A^`|tZJgWTOR$VI-Z9u#;+>%*gE+)mHprD|%ZQs5cObny{>Ti#VS4RjI4jzpw
z3ro|f0D%LX4eg|zp&_)IZIL6xD+%A?20YDvfdL!H-@w^LyyI;Nb|`h-#*J`IqF2m(
z-!(+HI9^3gWK#fnW=4HIh;+XoMUeImaQ?kgY%(-mRN5vWFbEWo)wi)xhyAXCbmheW
zOaE6YVz8QQ8n0<9>A(j_*n$XX)&1_?A-INMjwXtd>JP_!;y%Mq1{`6AIY?)nLM%~D
z-o@r37NmV?<f1!Aaqvf{5=CT<Zy_88ym^FAilYsODi9H`Lx{KG7(!OSgY;g697Q7I
z-JDb4yPWvW%<gl*#J;p&QSx`t0nt5VGE6?NNi1})GI0hM29VJ7!xX1tr5^ly1ax|^
z$4Ev2K*OYc;qNwtzWSS2+o~n&V!?WNNbl5{;4?56PHXXbgf*Z3f#sR?n*bSdRkxsP
z3U2nJKeBsl#-E|miGic%U4x4B@s!lJmY^$;6|HM^&p=Tz=hZB2P`Hat`Gvw|U5+>%
zzrNB@cG38{88dZSt#rA_`aNrL6COysq}K;2hY0B`c+zvTvmtu1mm#DfI^g5@@)M{<
znGy*_QHE<q<7vj*&A4_~5y8Zqu7V?ELf2k(Qf-+-(#u5(%gjvd!PmOM29K7O@KV>A
zRvMkRF|GnYjWsrre4CpcY`GJ=F2N@i^Gi8O=Mz-M4iB}zn3e>jjD`dWMt|+;F=3Qf
zP)J`vtoPXcP6;shwRK;6(=zCkP&lfa+1O4c@^0Bu<l=YpF5g9PSct8PDFKN@NCr*F
z@n>{HMrAkk{i@kWy3j@2TQp{xu|(Eo#)kcUFwj1)B)1$5$CY`8S5~jK7ctfAQ<IF$
z|Iib2kL(AEDO)G1DSEpcmN|@QX-2cK0KgH^A9qX`x?lJBa<b>46@DLHXdIBBxs)cY
z`1qSQZv-qVDk{W-%gO}-xuCAvd$Sa1mIz~XKQm4!hiN%d3?JlVx!y2pqH~z6pm`g{
zb>G*__J7=^pLIN{=H4M#ERsAyRtixj(t;O^jgwG4LWtkjfIGn)k#ePD?aNvZy5_{L
z8~@`kg!Y;;=R$&rWYXY$t4mL4f*pf2sgzA$=is0K6<m-%#7C?;$n*UyTFQuSiyd6n
z%1^T#?uKaml?$x-0qcKVk2E5)M3n$a<L4(c3;B-p8ow}d_;MQsdF-sLgg7j;_HK^#
zkF^p}guatPJ-z8)E`U<=t@1yUuoE^F$rPm?Z>ix*yDFn{oD4!#xIR~5z?e%e)r6I?
zGK*rxkJWIL#X2^cGy9tO?MaJ|r+KJFTVc|0$gcB6gBR^K!M%G=Dx!BT0x?^~06#&o
zVEr6dnf$D`5wX<jlDmtnxKT8`az^^qV-`WHmVHZsORfhXvm97J520Kn=6Dxfa>^C3
zn*gIkc|c@o2P7m4^uv&ib%p|qcloG%PRi0z)1DGanC{y}H?~s{&jR`rn=kNDLPgQL
zB>auwe|yfo1)!)%P8~b)QZA?6?AztTKK}Wd=o+E<c{Jw?z5X>24u*iV^c)I#I1f0C
z0?TX!Dx`D*<q-h+x_0V|qtFzD+^hp<Yw6%HU-0r6=swSAsoT6gr7%lTafwaNDdp!o
zqG)d4Rp}zHQW&{VP)mW~<+c?dY3v#K6<jm`R^vF!vt-x?{uYu|=M!RMc5-;(-}aY7
zvAKkC>b*nkMw<<Ke=ViCO9MV}dVNZ7&+OiXDsLd^K&8rp3|ie#83mq{HAoZ?RunYM
zYitOxt~ERopk=qzxwfQb8RzeB5{nzMy@)|zGzdBEc%5XZ_pe%oY{_?!c=2xXPPd*(
z8Ju3}_@2&Tv)><{QkH+qk~{^!?P@4x0Ojd*=RCWL!Bx#ysVb5kJlYvcDl&JI9oYF-
z>#N&_)<|3dgaE?^`TM>JN2rxh?Q*5}?D+u~ZJ6UP+k;67MT0hzqlZ+w3nI=jR|y>1
zu^9iku|xTu;A&uei8xJuM?u)Po}R`by95KRI+%2LvNC)2R0%G9J>=hYju$FT-^0I{
zHod&JBzTS;fOm7y{Tu$T$R?!-ciy2afU<uqSS%qLZY0nmSyZZjGY%HH98od_G^lwZ
zxzjGECW>&pzNWd}a6Jt7ifV9rk<d2MY3im9c==QVitxoPThW0i8RBs7Qyue(=nJ^T
zOulHtU<}L?k$z^4alolpsZv_6Y9=eZ4nanXJcGjuY$f>kk+G)G^a2}C8g8m57m66Q
z<8?qCQps(T0qqt^=$8JE{EZSxMF=b2BV_;2m;X14kY$v0d=WWMC53Jp|3!0j^V~t>
z180O<jRhZ@Yui_OP;u|xy~yiWUP|r1rtamz(Wp4VxYI(T?iMdh{mTXwEA}n_-YxIL
zg>}Apd3mw9q==guv9osHFQJ)z*`7#RibX9Qy?gYtlAV@f&UC8l;@xhe*aXR6Y6=eU
z*jy;_dmLhI`Tm9Z2qEBeYc<VZ-~7j9W`cu>e7Q#{C=wYE!}>Hajr1yU^q4owIA?ji
zd7TfBjPGFBsp7>rdsGtc)Q(UGs-&c@BHD}2!=qckSIR45V{?F@V$135dfeC7hp3?(
z1PViMsUGuk*6$yO|LI=0c$jeG3aHjY5)CF8bSbb0P|1Z<)=8&WWTmjSHUqSW9bi#9
zond|&l@XuT-;wG+7PgML&Lr)wJAy&M@$E{Nzn%ZtNF!HL0<MV`c^v|>4_`$?T{@CA
z?f1WlUl{=bD1`qS&WD?bQ$AaUt@Y4{3Jp(BMa2gm4^w9TmN?aNIZjPca(ZS1pCnD0
z+3<~YfV+~iasi-pWK2Jkd=TOhfQzVTLMP(wP)-4C3j_CJK9BtxCrlJP*7(1<{&-+<
zS!0h!kbhJSJ<#+(*h4b1LWba6ijo|^SXmKJ&-}ZPn8nS*RhC9-o_Z~m&=@b>xfrO7
z@{8)-g!E9lSN8)B(T^qCZn^%CpJ}tyfu)$9rK8B=TV6m*V`49z1kU6*Zo>XvL#qxS
z)Z2L~PG}_FqcQkS?7QlC!}wSgf(7m2xHaLZUQUBIuJUD6il5>wftin3zFe|-ElYro
zmX>1I{zdy^`0}yRcV`!uy9}G`K*7bl<l4(eZ}zN<l|IhEF1s`a0ZQ&76NZwf3h)9s
z&sSJC@W9bGoJfa3){D@HkRLvD$ONfdE;VM`;&Xg1XGA2M|0`&9vyW+Fn*pnEi<H!F
zSl$~5aFLVM)M~@d3Y#B~LL6d*l#A=!pb?T*`YP*hJMoFZ10kn~X1ufql-g+73koGb
zyL;a#D!f4aBc3wqpFwd8#X*D;Ige6;72>1?jA&%vW(&0Qz|~GUuFq`4-8wggTjY~G
zzYFs*`<}^{Y1a9aN_gkUhn)OPwSzTztY--&m)EGvi%ocwdAUYe5gg^Q>Z8I$x(zQZ
zoIhR-K(|#~FH0TgRLD(P%M6yr@5F<D;b5*yyVl3-@KyN{I|<+nq8eg2SDc9mbCJ>W
zEQKMyR337@&9))^O2<6J>l(AIqXy5vHsI9jD>?4i+Lr8EAg~BC7X8ke=Ew%U5%e`F
zX=!Qj7eTQ?NzZy;tV@zt&PN&(-px?^m*ZwfZnI9MJp@u<;PNLn%sXb#wzMPiD7(eN
zELcLwv!H}pu?Lcawj!LS@!vTl_@su*{~WVr1Fue<#yGmpkA*>z81khpLz6k1=+n>I
zvLp+%$tSh(7$5id{^OR&3^R@*>lFOB1+<re(jmeq?Dm#=hk$PK%W)~sI;#gE2Nl6@
zU4O~TBkk9o>3{jHN!=Ng1rl0-+5)w1Kr|X!T4rljt>OS>3C0p@@av}!rdLi5uK`SR
zZ}N)rpWu?nzko?{5nlpJ?9k!ERy~uLa~%CYQDIF8dp?XeD+MIVi#FZ=*K9eauNeia
zDiu$8rDND+4Fdxg>Lm>9+e95k&pfZ{d$}>fazDQCpv&ei=OQ?lr-Aj2DNAoHfEuvM
zc?X9n954t}rA#Wf+zMYC*Y)}V%OLU}HJ<;DvToT=J5*C~`alX4M6YtNP*Iz7JgFhq
z>0Gql#}c!i`=_*|b<VN*UQ$-by6Jj^ERs>>bp74Ge*CzM5kOd^tJ{n;K<t7SXjbW)
zn1>@JJsbNOHu7Fd1$@N5ij*5u+%jV?x(V~~A^Tu8qIo=wC2HSjObLL;?pYdbgF-Qk
zy+k(;04DDF^=F|H9ZSRPLKhPkokQ@iL3S`u8IzOF&02D0_e-|08t!#8SgOdZ|9~`|
z!=3Imc^e+YDDjeVJ_|Mdd777G=t=1~^vE(?+uu*@_r_o~#hj_>0oIZ?H8kHxe>q)&
zZ}pNq)7>u~KS^i%bH0dgEfPP~ywTwnPkNT3Szp)rpFPgx1t?n7#a}$<OD_eZHf+Z?
zns*Srj;`RuNv)Dg>xb56KU~LrLQyhh5tlW*PwCjeCl(wX>k%Rz=uN`82Onw`EAG-l
zj6C+hgcJO^3POTA-wMH5aQ@>hBq_UH@lTWLQJkC+ui_>-&sKBV{k8Fn`z&Vtt7c^i
z<%SMwFJi8g5%OdI6xJ!}g8Am9LN{qIF06tdYI-RtevkP6>j$eZ)2)KrRek-3X?Ko=
zTIGF~U?oNGsTGFDZT<vbCiL{94}8xhdzOYid)e$`;Jk<{bPkappL%N<wzVkr&wKgc
zy<$#zd3+aKJ!<tQW-uf_#86*ZN2M<2G{)~QKDHdWTU=z(nf?8PJE%kVmHm$W_;0`S
zkAFNiQ|;7Uyk>bt`CsHK$GMhGDSt2(tPS{TC;O+0gJ(;TxUy%RZnW@R)Gv*xHRN-0
zOXVEQ?z4DBRvyo}-ELH2Onyk=pcLypzmU!{0Jw?9!b6LBH%iA&gKNt}D`=k(p}1$h
zJ7l%?Pe3cD@aG9V_-XJ<<ui!z@O~=UH##qwZ~Tq6RHD}|amGd2ICN;^>vqSTOzo+o
z;okdb=u724<U3xGIz~Cxp)|4!#e|7Ap6P6lxm^B_C-#r8mshy#_N!rxCP|N{8?7hT
zVm4mHSr76QiObmKHky00I>>T`(WcCVLVx7}kXcWUnvD*p7C%RaA(3J8@?aS_nptVR
zoaRSyEGyRQKA-Gy$T#%Q%lyOC&hfNbaNO}0K;weerL-fJfsU@A>O`90xMaS^)WCka
z#UC!EEBKMDtE3nwzeviZnCvXQ^1#P#xuG!=e<OJgxi<%InAxxd*9jd-KDOh}%rmSY
z7u)rintx1-h51L#V&`GEJ2z+Rr*smOE(ZL&ZW#We-C)(HB?tbzYWS8yQAy8|lhF-k
zJ~8r(Px04uLe+`*Q$y<<a4kRvv^ial;eW0xsI<I%5kOw5;C97^B`*EAHe2WE#VJ|-
zH>IO*=P9%B@bQkeLUUnSXkPS*xpNa&$vLa|qTeZhu0_9iE%uQ=ba;vg6V{tbvhLGF
zUvvSdy(U(K4a;DOQ|H`7$e$U%xp?_x%nErVlQWoJZ__>O*Zzz=2QF*fztKU4{FYL=
z`=1UDt`gIWi{3N`=Mv}c^LvbTor~(GSq%!q(&>aB*%$r_Wu&+*Tp1~U{vvkm0ZK=#
z<ui6&HY@fe)Gb2KsovB*L0O+2l&zlb&sxLpX?1f$NL3$GK-Rk~rCN`p9?B{|Jmgv4
z?|iq>hkE!bk4fuaoNP*p=4rY8vBN>Kq5h(63}R{rf6usd7}brm-LQ?ZuXyfiD^8(w
z9lUWgL{s6G3t36$RNT}bYbq$1n8%l`aImKRNZw8TiRyOu<9_p|^!XBGiCgg$@~SF`
zxgXci<&fX07Ad)r%oo%8CC(`?3I=WpujJ9(K<IZ&o0m8jU)amzZnD<-`o}g{BM$~<
zq0=X;m}VYUpCQMgn+|j5dI<8UN2bPTz9JEdTs-~ghld|8nH?!?R!BToc8R14oqw+!
zUFr9sHnB<J7$agJx%=cprlpGO0nm{rN~RX=P~EPl@v}}Xa}{P=?!L!W+Ep6Hi?sc-
zev^;z6!WTP6ujfal(}tK*Xb8|m38K4S}u9cO;r8j3#p?0qI8C|%fsHS^ovOnoHM0&
z3DoMtz;r2)fjP+Ne4B?Dh9Bf&VtN|qFi#Y@*~Ocg9*L)ogDmZ>ZH;~vt)@L2PtlH}
zv1jOAN`gL+|1Z8wS^bNKHQ^0p50YU@d65#>eDXJ0a&LVoVW@3A$e8!<v!$?<-(39U
zR$Q9Je5>|}u3_@YJM;ngWCHiUue2@TsHm8-LZ1eYe`Ko*-m{VE%erlk%I@e3amaV@
zK1%T-BaFd2{IKi}Q-oHZnj4uEc0QrQdnb6@o2zZ;!&|37!gtg0?OM03?w_*Alj$)J
zndGzS$Xk<9KE&^rp!TTeWXu_o=W)y8vlv<Rs=sINgl(je%SDaQ$u<VuBIJ<OO3aV4
z02I3TiKn%A!ABnyol5_rEvqhNH}!usF@z`A<!8{QERTHW(*@Hb6SPryND13=+qyG>
zHl)H!od15YHxM+-O6~ZKM=@w}yD&!&KMs%`=^;QTbMRV|YQo}F9mbUx3RmTmn1E>q
zzR9wDaVd8xxv4)?burMTP4cdlWEkq7(565gADdl}7_}sehxYUzFOAcl!=DniEl_nR
zNMe-rawaU$mSZonwtc|*#PqS#Hi(kvQM`$UpW#0DcABN7jnl`)fASXXIJ{j$?`Iv?
z;=;sR9!|N$I#aC~n@XEEvIDlsy0_DYy%GO;9?l~|Ch-Bubb4?d6xy>lkV!smRX*+G
z<P%$iKfI_3w2^c_es(<>+brgubz3!Ft<0l+A`72*!%eo0)%H?erPwcVynVO+vqY%c
zMhmuX>a;Y?o~=bYlwUK=cs_hMnLrxN4jy$=zi+O*xCbyizA3bs2_gR^GgHVRafOvk
zdn37HaEw~g%F)ZexG*S}mSw*<bkSlf9+~O<66bTZ4GsQ1cPnWdfZRq&NaNM2eA?cz
z+De{^M=zkBHuBo|^o0U$d7~3u1!PU%-x9DPQ1vyZ!+)+jwP4-J5KabTBb^N2$)>+4
zS7_U=m45WziQVo6=gP>lcfn5GS6v*WsdZ$K63T?n8UN>#$(QI|_}%wP(+n7%=v|Oj
zU+#;OyaZE`!=><swGA(hD$uj?T1w(av!&M#`@#+#t>JP1=^qRbok=!De{avV>^paw
zH4ne~8?P;QVCCqCf6D6b*a$sZHA(W`9y{n{Htr2HnRTWQKQQ!?ww-qc&}XY|*L^MI
zew>H4r2g3zuqJTEgfaUJ?dfP2`B3ZcuPD5@+Er5YsXbsKIlC?HdW9v<doQfMQ>jNz
ztu1$!O<jmya3uDG9q;|tgA#_T&GL_5bfhiELvHHs`eGV-rnHOS#}giZ#hABfxVIJU
zP;gl^;>p|bZ!wKu6EdT;?{&fSyX$Wsn-IJ9b@3_5b(1oEYiQG?f@i>_z8H(E%<ac*
ze{B84i?8d=)G^JV<1%@!3)zaL7x(@pJ$k2-_TJ?Cs%n>Stj+t+=v^2vt3E}W%00Ak
zYF@W^#;86%ihFBlJY37;|KjA+#dSn`?E$kRM{cRnUJQ%#j_OM>0}dL;)<Y0isE?Pw
zLbF3p)_q&Y<5yN6Rn6n>xy~MDZp-7h*^1GZP3{{m?r@DI&PznZ>J#q`_5PVDD<Qg+
zz#yB$<0okgUUlUtmr-It!OJueM+@32e}ZMS#V>|$I(9|DIW=u86v_{nKshlDjffTG
zmD`t1EsIV{I*e_wIA(s7jvu|SQ`Mqgjr}y!g!Mp&sZ67d{|}70+v2}=;5sfYH&`7P
z2jGa?$kesKVs3|jAAMQrrGe_jiAhMy7bsE=XulkgQ|&pbWI-Pu@m5<550kvD+}1e;
zno)6LUWNfttyaX&Vj#j_Chu`P`=3_`e|c8TO`P_?q06Q)CgvW7n(<9Ut;Nj&#^NCD
z^;=ht-dkLc<X^{YMlrFAyX6N~jePA=oFDwmWFop?V9AlLbgu=ro=JP10aa0Y^b`R|
zJ%s5Bx<B1M-crdVzf(Q84G(F>TV?BP`@ksSU=r;S3YJZIj|{4H4)G9Mp|`~Oy>`u1
z9rrcXei$Lr_S<56k-anR%_S+gTHIy*^_^w3u@t2JcM5LEqM<(;F*x;8dX$y6h0PD!
zaD~yvmJ2Tak6!I?;eE6lX3%cPJ>|dnw2?ywbYeNN4<B-U{4}tMBR8O#_Nhejsk?fU
zGhZI(4-5>D@}tQ_Vpl9KsEXpRw{A3uX%DT%nA{p#K*Y3^2FU*Z_E&P3owT={uI{%v
zf4=*r^Lt3>pe*))du3zelOvI0y?XDp`5%4ky%wYSaYy};iHV6Ob1{=kgM6l!h77zK
zqZ1Mm;^J1zWKQ&S>k5NP>1jAsS?_hJ-Z5KWXy(o5yUWwm_yazr+jBK~5c>A>&2u~5
zueMsfpBs;owHO(w+|!1V5v0D?@aSLrzzbyDg#6F|a2RP$Mt>HRGZe5~AwT4TUI8iX
zJAJwdEqyB6ZNtwKx^DPgZVvekC~J9Y-`><-N7wb935OTOgN}xfoRcHYkJ|<sL1@cz
z8XolTuZqO3ap~LIP5@<Rm2U*20NMAJ_BvcP<-!NlaLCgmi`{b0ORzat(nb|!9>YMO
z+2k7`k0!v+Px~QKy1HnFEcp~F>LyMNx4rrNJ`j#Ok9%xvEK+Bg_O@<QwEeVY`5Iyp
z|K}%pM%(<F+S)|%AB5%HoSYn2P{E+~1ziShh^n<uP?w`JZ8yqcP^*#vI10M;BQ;SK
zDBWwhkl%`0H6jBfI6spFJQ5mQLJQAd9<o-tA{(?(EySk!a9v&9;LfY07g<)f_v@Pb
zV$<E57aWk#CAxwb0o$?E5|^u&9ppkO9cKlN3T8b85=liUeK-(1qEEBwO5PrH@bMDL
ziN?F-o4wC$#x$VZQd~qt#GNSv*8$Y!47CST@EUoeyn6ky{X9nBQCVqPRwtsQtJymF
zJg{ncWTep?YtLonO@OWU+1lC)hVl>JI{IemDp=-vR1=7RlxSer)C-O+NuGnDk(e_O
ziv-9VDhnlmST*O`IXk%^(GHmbYXyZj;6npo?gPvzCL(eG3;<;9LP1sp&NgH3`pf>r
zhd%Xo!mb6Pq$WgIbF*e@zHF?Sb0(KW>_IusbeO&3`n%;~fI$ZG;ixPSm+ZRy&JnC)
zqP^3OQnqx~or}*HxU^#rG$ct|nL$)LfCt`+x*4Y8j226$54;er@IL_V2vPBJ(*Ea4
z5=zQn78Kk-dKb(tQm02B4&k`y=iZX2`moM!U9Wz?S9s&xhn8IXk(5g(V>C6DH7;0q
zHr^1~r(2tM`4Rs3`?rF81koW#i(3EWl`b=hMB2gd>g&ItWJF4cnm@AqdL8+NPK|z)
zs|Gc8^7P=d2y*JB*PYU1;6AWF;1k4qbLV2;Ws<GNF25hkBv>^OC8TpdjTPvzqXAI%
zp-Kw-N;2^!Uv)%R#^$p7^+e2y<fh+kG*&b^zFd7@{d@2<Q0hQNuuQCF*Fx{H^@~de
zG2#1hOv7EKJ8L#p_c+*WlUyzM@lq;y+s%;B3g#REuY;Z^Xv1cCQk!p`t?wo9i=~?7
z@(HUJpG(RDg6K~HeH?JXOHr^qH+xWhpY8}sISXAMq?gGG(}q51_lg3tF%Q%8aF}ZC
zluSVvMZ4(rd-j|yc4zZh=vQ}L#WOzAAv?2&!>So6c(AqRA))#UTuL%UV3%-VJ4t<i
zr7)lQpT&5DWZ_`<K*?Ce=ooZD|G&Kt3PJ8L1-XhtXr_Tl9^4(N+I#zhfso@}6zHLJ
zYGbo)#?KyjJMv{hf`V=4sbzrk(Lo{tw6Ld+12M2>9-fyy)1tG6RQ5d=CDoMJyF=i3
z!mbD$7{07*#rmJA)7b8}lE*(FKwQG`#0h439W*$>4IOkbhG5+i*7Lh}@33C1P-o`l
zt>XF<)yB3c?<12fC`giSn;rYYK@wm1Dqh$G4)OE0N2333Iu5LIBN!>@pZIl$USiR<
z+N{sPT?tHU531*3O;R#4GCut13)aoOQ+<RqpMlUtDoRhG<N|bg;f?I9!h9{7hvhXG
zI!*M~?2t`NLQNeQ3aFZC`XzA|UK7jq!-@wr!4_LUBA#%c4{BTx^lZUWE~vs75rG3}
zK8qW5zNxw2wYcSjUU?Y2XAw;NjeR5{!OcY~S@6<2F^P%uCBip;cp1%}Lo=L?xnX;v
z+*&-&E#KFW)B#P#po))Po<uRZ$|r~_$Zz?(>hqK`YQ-+le|dImt?7-A|8<W_Y*{GC
zNxQ!SiGKTgS!lvC6{AF;eXT832C`6gC|<jPQd;CcB(Qhwn=ePzFASg3vtX6iL3xi*
zz|qK)LDa!W3rynqHhe)0RuEYZ5u=YzTnEZbg1cn0L=HS#DLu-ifx-`YdU7WCbtC!0
zz$0?HaK);++FE_b{z&ME_G1K{Yt+QkpR}8qnsOu0g3>$08r!3wREl>)MQz%_#DC59
z;3Vqrzh28;x_1ATr<OCuK2aN$v|CpM+*u*QQ)I7-we7lD{t1ZjUk0|rx)YPsnmtCh
zg0u35R(Fdfj$3rH!@1OlV!nv{j$i>{VOQ*FRELhE9!o@G0wlNFOdu7p%tiB6>sMhb
zzRSrUEndyzj!BK0=2PG2nV7{W#K34G?>6$p4mluXCyB@j>X?p#BSwD~MR`~d7Tf)%
z+kX{k!+^B2_OuM5EO{*tr)}+t2WHIjXd1RKuD-w`pEdn8`1vt6CV2y-d7v$c{jIH~
z)!T5rp5=049~oRO8T46LnCn;&<l}=ptK3*Bma4(xQdNwdezHgM4onX*eWdCRwG<d`
z!rnp&KNR#P0}5v&MUS2ih6i)rM_TMizhLvFU<)+JVR@8?!<!r1pGh9e{rdH35W*gQ
z>cugJ_L+rvgI-r9jThMwnXr55v~v-MEBK(JV3l1-c8l6*VO@QAhnBW>2lf>?EB!PU
z4Mq6FPeQrz`GZBuC$zL|J8!P2udOBY5)iYAP9D!)&TBZQpze^KANMY^0o5KipeB^`
zTuw`?j#-5hqZ3sF2IM!eWwIdGj`Op&zJB=Co3*Hk$;C_rtA42YcXTX}KFUC03`!hb
zQL<!Cd#*9aek2;^&v<P0Hfg7L{_d<Eb(5#L)mxT-ChqkA>~H)pPSO9vug=V&ALgz{
zN6?}t3=GLZq3d|fsFN|+2<Aba32!g~Y<1bF2+AK!2`7q&NeeScFvF<u%IoqJ&|UDt
zK?CyJ2x{6TAQ=c{Xw3c)ycb9hd<Qi~M^7*J3#*xliHA%N4(qdTl9p2{_2HXRwj?G(
zmv|>bhP~3dZGInSyJhFQK{k!pOo$L4jY5#lhZsjLhNk!yb{$dMfY0Fw$3!1}2XIRA
zTUyXjJde7qFnL}VryXWne;PSf^AXQ&rLb0;BY(lF{)kU4A3$xp*$p#@st_#(FgNJW
zLRIH3mz@sQS!`S!v0&^*&Q`Dg{_E)DT@t<i8c<Uue$Ip6z@Qo_OAg4jNW;1i-Bbva
zG`6!I_Jltn>}Z88Z;@QWmN5N@kOM=k%P&J)0~O6=)e0~C6uxb|FYq%wF{>&ib{AA2
zf^XdpC)2*^VnA2JCd#u7m`W2bl!5&ks1tDCjp|4^$MN}bek^dt9jiW$8|vvf6|dX@
zM@;k7DYA>ys-5V=!B<8BS^i*Rt~@<*k-gv>7`3iiN@4M(FjlgGzu-|!I&R&7XozLX
zF9S5e#n#;Jl*>?6S@{axE+-S-ph^jvf}HaE;e4RbKPjCRYTG@i`R0~?f{vTp{hDkh
z%x<4+`vJ)j4Gq6fy<)Q`DRNHmc$-Zi9lnCw;SBv3IATGfW~a<<5OD)<HhvVinvSKy
z*!h&BGI_Tv=<)oIj+8+!8YE(*ibs%GhhiZ12r4`0S~}*%_%#d+TrfCLE{~3gI2UY@
zF%8xE1@x;*e7+X0)n-@=e1oX?{mLI{gTrJdFpC3x{%1fTCj_Zr?hUoH(ou6tW*?r9
z-Pl90Y_XbMr`~Lqr^iUO!De7SiIu2#-i0DB4se1&Wk#x0S4#^S*iIx$3&zT&ZeJw+
zs3yNH_N}t4^5$Cg8$bPQjjn^?2@DDCU@K%@+jwyn^$?Oc`mdpQ>gcA}OtB${_hSvD
zKWzE_?)#*i8;5Irym^3u-A43+_M4cc66PqYw!Re%i6zq2@cweYr+mU5A>w5jUyW8#
z%vVWJ>fa+0*(ARK5r<F!X<vl}&ne%C#e`m2eh+s%wK+*24oRQstt46+|Ly|bLXgpQ
zu!O6cpn=!J2+u*({c4?-wvupA7D8aqE<2EM{u%}ClP86x=oiPqg@UVckPDjfNaL`4
zsK0NYxDK5sl=uGrdRrr=7VO-%7gs^>r6--kJ{D@ze}NI3dRJkM`#_1qd|n6xoCBLb
zR+5<a^W)cxq|KVG{xl>s?t4D9Yrfo}^ME1V5IsW~M@V_|*ou}x?{7)_P-KL{$LH6-
zfMAA(yK%2(5oNr}5M;zGcYI}JqU@y7ai5uESw5vN^eZ)NZr3Ua>=XmPhe~Tkd3kwx
z+);y1y2+t9ex|A72GO`;lIMR!jskglFt=tl&yU{4HoO7fDMpju{W#VIOm81``Z(8L
zKXJN!!{JEI3y583!qZoYnuMTk@kf+L?b`f}Y<DQJsRi!sg_L;*8vn9!<$Vn@TlGnW
zZXUD?oLo(IEf%P+9spO+;lo4T!TH@Fs&gA8z%)T$7zJlPNw5K<rD){k&P3K*@y}`3
z(j4JU*=X$gy!iG#$NYHh3C~TP;-eUl)HiFT+?k*y;|wiYgoaMwRH$%YzcBt1PUI(c
z!yKE-<7LM#UhKswPlVs$0UBACiJ}VHKywQ6*;iVO>1jLu)s3KKQ_aoIurqtykAvR4
z;mtKAhEX1Oyl%>?QLH=PkuCYXkM&lmKgBRq{XnCniYnh?!yl>Tk^<=1v07L+ZKte+
zGag{N8-|5}$N%wTLXjKof{rhQ<BrpZg8}v`^n{zRprCZ$vs*#-Uo&?Gmkp=Y9OH2Z
z??!Zl!FwhjKKHlc=0t6`LJV9)G4`H?+BBAe)zaDJ?LUeKIUPE0!k@X_>m%{?Nnb5W
z0XK6IwWWxyU^=Ou{ebgNb|#|>8WI*I#>*+@`atE(|G{QLNF$AOm4;tfxPc8sQ_NE|
z+Yvsq=Qm<+2jyM4kudRfECgouJrpmL^K*R>CJ)0UyYH=U*K~2(C<Kk2>A#R-<bpS1
z)z}ez9eR3tn*`epqmz?=Xcx|C6$IR9BT8a~*y(;1S~Y9YBzm+K)c*yXN?YAWGtn;(
zS2QoH3(_lazCIoS^=*r(i3bMxwgUBBU_)f%p71+XF@{>b+gRl_o8e5{Mfz~8jul)s
z$IpJ?vTOYE_iCa2I9?(`ohi>;0OD^N_(x&wMmX>EUvRKEAjB4br^k7<i0*cbl>0xM
zTbwHou29+<J@#{b1DcKn>l=easg}52egn)R%UaTXCwkK~kvmOFV&n<I_?HQc-P$~h
zW0;P|{p+V-$cW}m{Ek*BBr06+!Np~zpz9UQm|f--Eh-|SC|up2F>9Qe9G7WXlGBZK
z*n!R5+S6%YF!r=h$+9eByh_y*Uf5Rw4nwOFRuX@s7`tVjha$&}$5$e-gK}FrWW`ag
ztyPs&h|pA2;vr+?^zQmuO>3O1_FnIQVfvWl#eyF^Rv?Al3iM<NdwC(Z6rnl=M@(Fz
zz;G_@#IKQKQbU@ChV8TFkPmpCw6fiGVXodN^*D@w1iM9U@w>$K9M(Z^xNonfUw`dg
zm=0UWPD$D8z8kNU|Ev9XY+RxvRESYZl3!4_oFc|g$Nh{x1EoMF?<UZ$dhAoP1;=1r
zerwl2Z}=7m6)TaCiSVP3B`Brkw1CVLq)*h!5|UKMh0Nkk&T9s*zn0dhCy?~dr{0%P
zD+B)<;kjkdKzfs7+mBi5?LjdNGuTuUh)vyhqmp&y`t4|u+%Z(2&_et}ATYpvGLsW2
zawjlWC01w-It<+{9(OpJ6OXhFFOLkQ#eb+A%rm(?#w1_M@Ee;t^Q-BD2M?wnwNRnt
zQ4)KSo=0~>QevW|OVUPAA|1xN;80DyD<<?FqU(tj&nh&Pc_J?Ja{TjoHBOibg6JTS
zRfik1bCPi45H1)$wXMgA2W_M52dC92&$N~~xOKrpg}vkF<GXBPvLA>|aE0=`a+y?`
zSK>6)*mw5arK(q>CUu!0BfE5;j6UIE5o|IspRUE{z8g4>^P}c7;Idb-xB;IcgyXi`
z<c#Z6y3sI>J@M-B@RTF!7(eZK&$$ncH{sor6BZ=6hO>9M*yMG9!$COh4dKA&3(!*z
zqso}24Bp2)u&e#F(S#U+1sCs_*IiaetV{7zW%Uklp!5#RN6F6d5#!Ekv7Fli2+4}g
zs9j+~nF0B;so=#<7Tz8r8((tlJB%Xs#E$rjZ1w_ALEGkW2mDqn-KZ;qXX+~fnOjVG
zqIu9evl>;wXf3KUb-Ngq$j3lIMIdpFcJPCP_N}m+LTUD*FdkQhI?QzuMrYJi4E9YJ
zpnA%0E8Jg3+1b%|3HtPCC3k+Tb`F4vic>h(wNjocUlZdgB1j0I#%0V4%20VI97OST
z06@_@n%kR$EzoRbTCH?7DGsix9W*&7-*TTzDg&oHk9)`LSMk9O%jR<tCr1VtuHyL)
z|A4UnH`JLWNwps#9u?Ym0^<k@lSH`moLxXWu{G_{5=sylRh}^r0y*Jx5lGzF+pH2c
zJw+#<Hf*WesqX<$K_V%Fr)g6iL5uvtW?$@nBMjv(RKvjWe0q$OO8Kk>Ii9MUG!G;H
zI}E$6SO6KMFdXnegW+J2`l&mbU*W2G=gdIM?k71`Zbx0J2<4gxB%8EEv*`#(f?i$r
z+x;X!;wLIV+dGU97Q<n%Mz@(NA9BE3#Wf7u=RJx@5(FQy$=UcOtcJK_G7f6X0aH54
zf>Fyip(T|OsBOIV)`MAa(jcm)ACdDJoL1znj5kJax&F>{{ap^DMWTOvos>;TK5Z{L
z*q>naNiY*$iU1%RQji37nE5f;1#7HL9(PDO^k!aFk33m8Q<nsNWKxW867U-<egb0L
zP%L<gW=b>+nK)C=l|Hq^r=^X<Stcz_WFcCJ$LK_~>jQF8S&%Jkn~KvDu=;E_?JbjW
zzxsL4dwmF~L3<`TCPo5&c4Gcy^H%Tu=df{ABTr(%g}~F6Q8EHPPLw3A;(l3VbtmkB
zIRmzphg}ciZ`HADdgjR`uwN*jm@!(IFye7nSSxWM>ZzUSi0Kl!E|=~lN)#5)8J@dn
z-|W2?SH1t(nvur_6#%2Pa8=06r%lO~+b;;|yxL+UGf?;xCfqL*ZzLR1jPuUSPh;A)
zfg2c8-=VaW8Gu=~2v5>C?Dv6!4ff0ShCozN!S`v%y9%TU{ipnjt1c#CP_N9)2Icg>
z4AzZ53IAm?9G>mB$Bw>D8p&<NI-9om5*!yFKTGJLH_;>h6`|DLe&5pTozyK!k_fA|
zyN}|m?`1SZQou|v<pzo|4bVs~N5+V9#&5Sf_p>QzpjczqXF9~fG!vu<g;%7b#5aW-
z;Cl|k>en2TvUpw}3tuC^(6OJ0_aa<^mU=GEUGyL5g}Iq5-Rh$XZw9fu+Q-c{l?MT+
z<aYkO8n2WIxHCRB*0l5=o~V#(AC1st2VaWRNtc9$HN^80okYZn1h0Mc2z6SL0Fr2(
zcmRTwI0g{d4#lZRa~!||$EG!a`*rG627CcBpjh5*n1U%JM@+VAC2@^RIc6i+A|V_3
zCW3u}G2*nd$D@$YmaOUg@soJB=G?z2a))~BQ;K&}f{T!|Ai*CJArgh_rMQBn)>6^}
n7c*LC=)rmT|3H3gVa2xH7Zi?HGi2a#D2J7fDJIIF`TPF@j&F;m

literal 62192
zcmce;1yq%7yDmBnB&0;TQBqP`S{f;7X%LX^ZjqAikdzh>knZm8ZV-@|bmx91e*eGL
z+WV|?_8H^s!x-vxPUibQamRIE_w~F1?`6bLknoWp5D1EdxUf6~0w(~0JZeIC2;Omh
zEF22{L9`NAwS_>4{9%6{gsmMIKp><L3E{VjPDwlS#&(!zG5dGgO7=?j`L)Km&oGQ0
z6(c=Hc<N>TYqdC;C5p57d1vTq8y+{oE!7*D6|xP^sI?^=^5;}h=-l5}LI<6BRu8@$
zwCP>pQ8{-w{NjC(*p5Mp{`O;p04;(!efgNF(v7y=9IM**2g5<fH+5f9-)-f~JKE0|
z)||bIH9!Odgh0Zlf7`*|zk)D?J|>2}`b;PA5DoT10Grh7{xj02;Z2WV?}WpNFFk<0
z=z2)L5BnFQ{{X}3{zVA{k>~z}>gWGAm#MIWXR1FwJyna1=@C;@8ecs?Ap6q%eQ{5A
zO~{Z@MDFnL(2Y-<97Dp9Zn}BdiZfY;%x7^A?TPm$Di!yinq7xl1y89iEsIL^NqaJ3
zPtQhOT-Ybz_oPQJudYUk#3%M$>PzU<W5;J_6L&w9_;s$nmbNuAqMrBChgXfQ*D6AP
zfe~=-)-VPio|>9^2@l5<Q1$=c{Qo~%#{cCqpXuixkgp{>E;z4-aLA<a<Zin-J73hF
zEvU=My*5eIVLHD)8r)l`ceS#zYW@P3uUt@bwOflNA|csyzl&0ZQnE`Zt>+hmo(-mx
z3}imDP}Ii{uTG(}jh=UvjW<wr%X+p3Rax1=_5MV4H8me};;@bV<yL<*lER{*yW6wI
zJ%`#+Y~K5eMUjtMrY%;O&lYOdA|WAFm`xvoyEh(>ND{D{{i53j*E5%um31rCzS&Zt
zr>BpJiMczqygQw<9lSZMYy|I^8yPuWZsv^Q{<X0vij3#y@2QXd`LEx~MU&i}cRs~o
zh$fea%$&7sbicVgNqhh2=4>&{d}3<K>CX^ZT<?iD$u%iIzczxLqGFlxD5Z$pg7Zdl
zso`MgO?RvToOg5dE%W_YHN1iwd~fF+7f-HkCN&-Bccx39PjZ+|S#=UQ_Qk(`dAj(4
zQd>bG%Bj*#S9f_NOOgN|-(hdT{R&#y=`j;aug-2Z6&@9J)W<S-?#5R+Td&_8Icrf{
zke!`fTkASoVJ;R%4Bg#LVKE%Y;XBDyx2UE#c<au$m|xUT<9T;mSjgbt3dU#$uBYvJ
za}r6ec|vsWa`C6kTwaT%hGFO{HQR**<+~>?^;(wKjZ`v8=z@wWDq%A}YierJ(#TH7
ztCG1LiQm|eQ&Hulr>B>emR3|$6c+9_!;{R_*qc&gzG7q??Cv%;Hg>)~=(N|=ZuD^9
zsdaI6<*=BWT3jqLp82U-dhOx)v+wryYOfL41=S7CeIU?@4whM(ot^bXFa6Zd<$U-%
zl!!Z&m^Y8Q=lbFZwq@)G32uglIp{AOj@;6x-ie8ct*k`66q_j0zT0m_+pZx-e!^is
zGd4AKKBge6q^Nkk@6Uhi#B)&LY*~Mz%;V*`TfG6?*0O5Z&!5<Znu<yc;ox-ElK=L6
zm9cKWg~)L(&3I;f4&{EU^c^1Zb8=Gq7@3&3Y-Ys?roA__UTko0yt~<JJP5Izb35uU
zj#*yTC2O0qIAVyCM0m~299b2gn0S@E=OzZ?h0naj-ES)X>8Yt7)IA*`1iZvy=z?<5
zwYxQ@lZBe#=U~so!~{doZJi}Gz+g}CE7+~9ii(Miqm$!fOf0Mrmp`kWVV2V+20NC=
zvUkU_GC4;j4jl%8i_1mEYCAZ>kx4Q+2CAE(FSxr7wQG$z2}7rXp5kE58Xw4$YACwV
z{g{QWmwzJ{kt3H_SxT*&N~Gk5MoyWr_m)iQ#r1qDEgpqv_<<$Ylej7!<WpBwy*Tr{
zJ6l^@b9Z+yYP`LKS^nbnI!6#T{VN)pXM(~no{{;qF5{j+i}krvE;{MdEBY)dJMXw{
z*QLmWB_t$PC~j0sKOvGKw|*^ZE7=9%A*Zg6#{=V{>t&b}=Ecy^5C{pKg_4n(c_#9W
z*ZNTgZ&D1%@YvNw8@ZUH(y0${5F#TA9|<b2mOJM<C1^fzn4yRu53miEjOnlygs9SI
zgjE!ywQLL|U#*hdIf0cy_0OoPI^Wj5yOidb#|#mzw3x@ArE`Gtq9X)fo=%mY$*HXC
z#jUM=+~cUOuGS;r7zRO$#>eUAWU}!vrikaD?YZaWXzD=H8{76qlKXw{?)kHg>rgwt
zcedy-Yo1ZK>As^!a<zDOvj}#h^LG2h1^&sCYmd7hmSF|I*VZ(h$+stq?lxsTYs_aW
z{QUf&vz3;M4VnnWI?Z3;iJf%D^OW=|wX1E`K=9yo;Kc|4`AEx|r&~dZ%^}Q!2{(8g
ze<>YPq`SYLh>$RNe$Fe%hLf<EKnr#YS1qPznv<FgD%#p4j^0$=P=`+ZP-J9e@&z*A
zr&e6ad3mVF$eG;XKO^GesCr8bRKd<|hEkY<P}0iUU!0?h8%Nh$tS~B0j**WgT|CvF
z11V7vrP_ysgY#(g97C?2szF}?F(}qRA`-LKgYUGyqOnn1YQZ|Mq{QZEz2BqpZmHQD
z7Ig~>c&%63vZWK*Ef+MDl-Nx4MC3Z?&7n}J^g*}FN^1laIZZ_dUGK**ih!JHgC%2!
z+!gM?<k-6NQ96(dNgQUfU>`Y3K~$(SEYq~BtkTd|jyFd$CI(Y@X?>_)y$YP))>2lc
z^x?ExYWmT?wY}|&4op!6SCkc>*ZtDY(XkS}WzD8Qz0#u94+W$MOUt?gQL?r~%LZp(
z{`;-_9MtpY@hz~7D)f~Qjs+%Z7V8^X&&w$EkHhIA#djb*C?m+Jt0y|`?Jg}Xr3v}7
zO4n{?Cl)dVFE1~{lep~@ES`%3FJgQbsi>gv(&wvahRXTTeR!<;0sI7q?!zy;&5`x>
zb<f)k&(Ih|krxp-+HTe$Z`Blf-X5ifnOC@798yTe9-p1vT~0QJVu|_*rgaj#(qRO+
z{eDSOeOPQx9A<dBGaVuRyPGOC7%gjiYs+OVM#TpmWTn41V_NI>=3BFJKaPnL*7?1n
z{;0``x#o!$!a;$h`?l-i)q?9@{TYay8yg#ECntD*Y1@}VXuOIaKHx5zt#*Wj@f>!6
zxVg2p)zj0HkeJxIY$Yc*9M^cgCV{w`2;4a%jIYN1>eO-3z4F(uA*SHnU5nll<)0r8
z0%YAj$`iu?rb3``k2~)>vD$(RB@4?K_eX*VLI}>&HpM#DHQtZHW0-lp7-?uO`Yk<<
zMkE=>Cnhf48dg?Te4paL2A9D#+y9MaBY-Mp3giP#&H01X&U&rA8vE^CfDk?y?$B16
zjOWp-SJYh}_V)Mp^YZdC#F+wTu3T{0?&nzS8#4uCxQ@}K=AD1%nm;%E$<#ceqwfv2
zU_4c##L@!|;(#adxzU#l$9Z`!Jj7(eOt|`Ol99lLdV7y9+=3${BU_ihj^~}<-0CpJ
z?^>037j9cra-Vei?5sJ|+H2wng6&}uWss1`Q5Y?olENl$-9G8)=*n_hH&s4z;>k8f
zMM25p4!4VZ_#i7o$=I^ozM?mWyS=KDT9y41RzjAo?W<UcwWsMd^BYH5Jn|9$0(Q(d
zNH72qZcKg`UXUSeVX4G56TLh7u})TxKG6hwfc;%h@QQp!RbR=hfvSUZy&yuf$U{h+
z?>(Zeif{XJTA^^bbUry)K$!D7XNsDE?Ojjbv<P;2EQAv49{)0-(%?r7CMkc6o93R$
zT;ha-3L^Aw2F6C?*=`WDQ>^!?3P#EkQhna^zQ5)<Ee&R5EtgpVJy=J#8ij=K-{+49
zSOX<9M6yJuc})GN)^s1nXWyf4rUDA3p~AdR&CYPp+_2}6zi;-FW0uiE3nDFb<_a*L
z&V)>heEH%4*%XCA*w49p*I!PRjlWdqWK&Jj(F9?~$qk{BK8k>$c^#8TQNNf%)T|8p
zZ(khrcCU@5d3qzF4w)X1s@}uNc&?9d(4O2+SsBv1&}yf*t`gs}6u{TH?~4J_L<;P(
z%H&AOgrQwKywV_QBlz&~`yquJWQq&#W`8GYDigXN1zq@up~9qH%;%V81f@tJEApk8
z!C;IEX!k3|9KAyjJ}WcLHOl~g?OQv=aQeuH1&e9UMmQz|i3z99w;B}7Rv+F$>(wL4
zr_I>Oe2f!1sz?;|f~RDHL^JBRBFTG8(y0XzTA8Fqr5rh@$tp%)M2Ocs>c$UD?<%45
zSt|d=1Iw_1q8Xd0<fc<*6;rdUMg70QwR%e^j|f{?MPw`Ow=(TlevH|);Gc%o#{5np
z3zQU@=6PMI@rtH5{fmQ=k}rC!LDpJ1o~s>XI!Klp#GG*D<?!)w8kzF16Hyc0pPnJN
z#fo_<$8E%{#|Bj&WmO*uH}6D?Awo9A9;V6*2?^0OSY_trb*?@W^l8;?^t>MY^*S(}
z+S%E;wY62lO9HqA$QOcN1xV48zK5WT4%^Q|9nerwZC<~T_*>qz4Gp26J*Vj2v@=Oq
zKMGllAkcHJ*OG_?F+GQ3aP=TQ`wc6rUG5?^%2mv0a-95-mc<*wjfUuC&zJucI`|>k
zYsRtv@`k^=g(QJhW}2WX{FWOM9H#dIxJM!CsNz71+>H8KD7*T$1Pv4Ou5#=s7CjA%
z-t(EKpICNKAe4glJ1g;0q{T+8c4YX|wprAydKRKuc``>DZzE~@+lK<NFJRSU6IS)u
zms9a+o@O5hNR{2aWmV}He{sm)2<9AL+2Ui_W#7Omi;Go7GE9M3iEsvk_(qBd-l`P$
z`i#rH83IhDD&eoO^U)B-a!0)~;#ltw!;S%dE-h;j=>7O#-n1cpOQtTCIJb~zS4K0R
z(@AtniP_%8$P_3E4WjLHe(Kjr?*b2zVuzKipEq~w5TU$=o%0<d_O>zyKfXBBcD|HS
zKt{=8*Mj`xa@naqUmVI$GIa#GA^MN*zh=B#2cm#-H*V2)%~<I@sf;RCHAR!yeilO~
zd16Y4j^O<a6_lCbps~_yPYnqu3`DJS<upvDiC|6}j(mR=nzoYP3#I=ley>{EYrVbF
zkd%`rq=YOn>#IKWJpA3g=W$ls8%sZ1=X~^o5fmvKsh%EZyYt@O-cPaVK9rk`gF2;P
z-{LU@lFifl-laxN+K40wtXG%Abaaz{>$3LkVO)f`{!W~Wiwl4Sb0Z@!eD<Aq09@d6
zJGXLozd38XbFZ9DJ`@M`T*z=Aiy|3~INoijNg2U^ri+tCy!O2A^i1NiH<@Du#aW|?
zv2pItpVzyNjfttLi__B=QK{FNX=%%wIjJ7!%l-<4RA9@;1%NS0*{Vi0?afy1fqT?<
zt{z;tWi$Zb|DVb+hx+;-Ka57Rr2#mcTm<D9s5C4meyV0Pw3qyRPpw}$tAGg#$R-0O
z>FT15A~>jd&hO@7!?&-MLr3Ek%rN8dDGb(kULegp(8=owee7XnwGUESSARc9Vy>xf
zhuw18?sqq5zUXXhYypnFIM~<>nl)u<X}VwuD;GU#(O1Co&6gPT`k@fm*w~~vuj8c-
zZaCD&(W%<ZnpcjGjR8z!Ia>k!Gn_F$KX1F<cQKa3XWvU-NzNBIxZ!#-u5z}wxHvw(
zX2ls87--RWbGlvBsFm)441q`y|K~hF3I~OJZUkez)xKTR1_cf%f-q72J3UU?0-g(g
z1yDe4dwhHx1qG$Jr~^7zebg24^39t!SXfxe$;svA<@WaWXnf~O69wu*Y41VAkdcv*
zl<Zv%<1gI4ygb>~TLk1rt@Uch?X8FO=%@aKwX-WdgVVD`&pR+fcY6<G2!u&9NqflS
z9docX`#U-##ze!{X+)6hR66mcMW@w}`C3QM>s6AX+}r?7Gm@;*QUwH%H2F<<D@KNg
zBdVN!cSW3RPx4>yx7HNGdp84+$;QskcQU2`hRA=tj0P$vU0nXVi$O=|%*lkhW&88<
zjTG0y#>P880t=@<Y5rh+eb6nz3V}i2wmfB=v+g8H$PRHaYGv~C_wT&1mv?LDJ6Nt#
zC6y>9qL9FFrIX<4XArtJj@=a~5p~l(n2{ouE$xj>NT$5-2g!BqbOQ0|jMDh4r04Vu
zY?2^tGt;JBGG^Qs^7KcxaZ`S#jow5ArRc_}EWf6P#&t}Qn?bty$K1pJbETj+JetMS
zFI}@Apu7m<eYZ6YbuhVQ><-3b86grpwXW7Eo+OI=ed00-{@2}~6>^&O?TziH8lf&`
zP~;aRe@d8`W?ApX-u{hPy5=*!$l(EDs8r-qN#W2=;Lv7BrY%YP&g<~W_jhhykLmL!
zgRVc17np*l=Bhl;{XMIi2!zajCAGe;={BL8M)eg82m*0KvQJOXe;_h4a=O&86Gu;a
zFW9Gc5|pIqE3g`ikuhvA*7J9S7{d_`zyqVg<Yo?K>QssSQT^*6o#Em+(spwzz;0hD
zkfOo>kOL%M0F)+-4d*MxdYyb=w;8qS&L1XL&+mG9c?sy)1;zcX=fpILqYYGpt5aA`
zL$kT}>W&Qgucm=(G}vwo{Dr+(w91&+*jl8S<))LNnw4~>b8k*n!pJ4q6QvP!6S7}h
zyd~<$$TMNZFYl;Le5Ti3LU^Uq%Y;12I!Ywt@D*ByNrGOIkWO9vk{u*+n!pzl@~m77
zPFBI4pGw;Y9u)4E8(;WmhO%!f262vmv@Z*=zI$9kcW5YK+nWU!RN#C>5jTs`g3*Hc
z@fyk!)BmZIF32_*EbJ?vUv#weMg3D5kGizJB{rpUTCc3dX=~Lv9q^vb*)G!ql-$(x
zd2x@VoLC)z+1f)L7_~8)dTR%m9pJwAm8E+v`bWU2q4DfD<tgQNSTtHKw|L%6dJ@$x
z%+7K<ATKCo<(cec-&k?-IN<1N1g3)`6$K3~N`s<%*{W?<9`W(RW*w&aWqo+f$$5*j
zr6tm8qsTc6#~tOLfZmb!C4U2BH|4JzqQ>jAipV8m4KzDs2-wox=i(}|06z48J|o#f
zchjo=5Jf>Bn%eyjGKdDA<PD0=m0Tm%gxJaTscbn&iR3!vK8P{d#>BiX#{=AZgeO^-
zmzUhTRYC~G2EDPMVCX1JJv#|uAm85G!--P4Qc+c{c0T&lxahgR)C}mp$%%;+PV0{#
ziYIbd7Utz8MEe5>q54u+uy(}ZU4oHd?WjQ?L!iM<$4>fB01_ekKl`MYh)_F0K`%=G
z#JhQluJVH#aqZI|KYyA5_68o6u<D?L;5i4N5pe{g<Knb+$%l=~a&p9C*S>st7}r~F
zF%Jrr33sTe_EkHh2e8CYjzza90H;513Iq!O-aYq$L0;ETDcM3*d2D(HmO#nAZ>6wX
zOT~g)n*vJ5#6*<7;qLY-jQ4a3P<R}c3xwJlwGL3zqIzX@bpj;cmHOQ$@bG9vT$#o*
zfVZ&ojX*|2<3C^W&dUSVA9N649Z3%I!_HK(_lzLOzD&mYA2zq%gKFYbdCe{-UPrL@
zb^Z{j4<vii4UXu<WO`a3L!_PplcKV`I9!vAr871$xjZTIthZhj6_FE68`2Bo#SReW
zyBrSdcxPMW)4BSM*Y$LBGzXT)Pzc!D5@sFk?duPFXa&E5vQgrk2=MbA;qH%YW@*$7
z6Gw(A&rYS7(ll9~B85bj7^zmP%72>9Zh3<RA#Ih8;7AN1;NUy#ddcmu^CLf>mX_Ak
z!UEO5RYXJt1`(*%Haw3K7u}AB#Q-yMyIZ@5jg5`(H#|RYt~wEjMau^&PpKXu0Gi5P
zgH+tfv!I{EX~IV0c-d*$EediyX??c%Lw>`7B&rmrkJ}RkAYXNck({~IgV(hn8SAO^
z7R>=)fHA*^y^I_j9B0SJRT_oB!scDJ3&8%Bwj^=@v-tzS)u}QKCZS^;tjxK9%MHbV
ze144aJXk)rvXY~nn}<)70T-DW)P}{cZTM2dUfW0nhD97*-GHJ}-Yi&SD-3a(mQNI9
zo>bUJArV3S`fr0-cT{8?wcfpTtzNUr<moeO!L63WAbE)sAdLGGeu-Hsf)>`zSbbOV
zy}n9V^VbI&ZPI@FJt9IW_aq2J0~1!K#H&duym1vZwv6GPF)N_TYU-*|Po_0W<yfSS
zrzh{;KItjxbAJu_94rDu#w973h`sZMz#7y|gXR`qNi+S90k}*+`6Ub!t8U6aoq1_+
z^!WMA1ITnRtP<1@IYOFo9SE%r$uj;VCsFoe3}!sNbiSoTkS5G#b$rls(fJFCo2t}R
zWmL*C`+}v77{R@1+I1g7!tWb0q^05!zvhMu?aE-~l*W)nUm^+pD>liEgp@f<ctnW6
z7OeF9d>rL>Q*o+XU9Yrhq);Op#=Rp=9T=hA1Q2>t+5PxOyuoN842XMQ02o)oK$;n!
zi2AsK2*oUW1YyF4tw&X0rqzet>tEmN3zV2UXpQ0wM!83+8A1RoH#q9UONsUexBJ^b
zijq1+o93E{e09HC3~Yh#Xn->=rNc0N6FIQqp6*ZTV)<a1uAX%K1IPX0ptkfP64v2h
zY1m+p*-8~zXmj%u0OAmu`?4~U!dK8nMeGm`D!Qo=ga`$g3*-J5$+|Ji)Y-@BlFXx{
z@!GQh`Zmb<P6YYfEe6|}0mB3v?XZzyi*2KrS($F{2J6Zao1~7Uj~P{C)HE!HIIuj|
zG=5({^U)dcPyYn9M6x+-;^AJf5yxi_FdFew;B{~kfC@OSzce0vLEsWr6f=d8-&g2w
z>%P|m3PFJcfMMJqSY)E1Py4P3YzI`!Deu9yFkswY(E0v?CYeUUe70&^eCB19H8Sjy
z%?D3sr2m)TMcNA6J0>jgm#yzq{6~hf%PM<vxb;Apfxbbpc7M5k*yXyE<1~h}<UeiR
zzg2Mm7I6R!JZC<YF;eNlnZl;Sx2l0{fsP5R!AF`Q5%G1-T`fTK)1Fbxmu71Y0(DT+
zb5Te?zqz47#RVQf0V`_sliI#G{CB}EJo-Va5t8rrqvlJ2AIojIDzvmcN-`}EG*Cl9
zq|$#p%|Tk3Tw-5ALu77aIrtb3nxOjrTo+dHOoC<l9K=NwFahsf99&O>$%Pe4?omFl
z$2iz4tzrf#!H1cfaisp+`%LmtkUP9sN-_z-9}Uji|6!VU^HQ454}Pu%O%_1}GE51R
zr+x}}#GiA09`j}GoGPB(D)I3QDqyxRTV<XPmqyd|3L<DunzL(>O2{d(-J%7&8R-t-
zttCYDXIoHG4#wh70I&_f41zu^ME@)nP5Et!Txc}s_zT|-Mk5?&t$e#eRW8}+{_;;)
zNMx8~3~bhw(`J$6*88$rI|v;zwd5sqvaP(><v;bBhi;c6ZeG3&@fz}qh<K{M!!Ds_
zG+ET$UEQ%+)$z^N{Z=X`9oO&myaWDc2{P;5_;-G~ZIP{tfmDju1yNKIb*v21JI1g}
z0w64*Ow#i+?QMSjIM?{cWHa_z8EM5KhJKpZjsCdxSjnf@k}?}H{0xkJk}>15rFP#a
zWlO&r5D_7cp<JAypiQEn>=7KoQ{IxxMP>P-cACWH&%~6@K-HiQZ!l!;My^;_Z5>Ew
ztuN!{>+9=j7MA}_+Z|us<huRY$)bLzrIj+c5hVNqXAxhJlMpp$l*VWO>ZTd-Qcn*P
z22&hik+&D>uwBF8^Jj05qeJ>Li&P05^e><DlFPJY@6wtpIU)OG)R)g#t4Jl*ptl6{
zoJNTxCqPO1FG^2houj|C`F-8j8*yq(b}Al?S}f<>LX|hiuWIXC3K_@krHT1vV-2Ra
zPmsu7M0_3O@EGq?iuo6#Q6+>6{9GrYXPV_7UXqsvPPAiwOXKZT-GykgS@Sp(Da$!>
z#zCv-!AbK{vCf}|<y_T)B_Ad~;<?7OHhh?*5}$gY6A5!Qt8ba&A9ndN$|`5n)PJ9{
z({SCIb8ko+4Kj*@u+)KNiMfc0%;DQA-8=-x&QLa;M|k?mUu*4`%AZROX=&9YNcdNn
z>r0Z}(=pw$Mq$zY&p}zuEU;q4E>@KXBI~^tpz_A6X*ytbAmE7QbR1QzA!HK`!u$|s
z^f=a9Z(Rqctt*RJ`{`-|K@rIyF;)`l-*V!$G;Edm8aEH+@g>6)wTD-s(lk^F)jgoF
zpltT2YwCeMxeOzSJ*yy{V$-1GI?~>Akl%{$Be*eE@z3YQFHUuS@u72JB0r(D;LdqB
za*f6U#$Vr~TX}Q|7mfmp8|W{ThZg?HCci;i#wT(twS|KottZ~66H<l+>INE23uHgk
z$6uBMHe9Oooxf=o+5Ew(X=p3d)~l2qqb7zh<AJH@Tym)ld6cMb^6~Ee;y{}5LbjF&
zS>X*V@ew_MWD|n8-E?fz4F}CKh;5mr2!B#sVJqGIUV&Pk1eqAckFu^MWyqlr?Egd=
zQf*3&zynex&X*4$Quc^&P?K$m*9Sfh;F`xZJE8~UsGW2(0}tY#z$Ty&j)~fvOah3?
z@^>9haL^Jn7eoB^tl@O|Qac7Xs6ZfItUVnjA3iF_6QL6?eS&tz;c*HF_mI|Q3O#tH
z8(2)wM(F0(-!=m5R!nFGC^@lUgwlFHN1$X#x_s}lxO05du{}k|ipmTWHE-&Hlq(38
zpXnS5zu)MW?hxQQ5=c(=zi9`tIG4jL0R_y7X$%)>9G;<AuCBFEuKOWmW0m{~BwPtu
z=*_6eTki00Pf(7;sf9C~AXV>RQ<(l<N#$adI_}1kQ?G=(s?^<65}~aV9sP~MV0cBo
zcTL9mYm?m=RwyAc@xmfMdhb+}F{ZZeB#x;GFz2?ys9%qoD9AvR*JsTt7s;Q0nGlAj
zw_8_kmZz#nc#n<ySL7u}M=;2k3dq?5`Oy&^#5%XpFVfmX4<<Qx;MfOv9xMe~;le@m
z$y?utapaQh$k%XT#NDY4X(?NPVUz~Z`Ld5SDY4ebU5kIWoKx5FK^LTsEPFp0YczvV
ztGxDW6c_;|5ikOPPF}Xoc&XWY7O*nAS+G%wsn%-cs&ea$1_vC$LsX0uGtT{VjM|cZ
z1#8A+`l4R3y?GOpCkU289Huf6$PT}98NVU<0e>ce$g!4Cy&!&q?>;YPW>JX_8U{J^
zd^)f2+51=1tn`>Ld#6Au8fpt!NiNYao}}Hm#Lx_y$Kt2um}KowOG{JiK!%LMa17}@
zo#dPGbm*+9DPJsS-8?`rKO5Lfeh7Az$hdPeNr<uq@)IM4nKM1(5Q?C+O9ti|P6pc-
zV?cdYG_o4m*RCz3oTq0|i;5UHLAhQwl7<vNQ0Vxq%i6vhC|fA!=BugwGru_a{+n|5
z-vj|+{E_|4ZM{4xx^|gOQUq3h%eX1tLwJn<u<F3K@$lOpH7Pc~#)pK<_k3e0um~$S
zuv+1C%=pe!Vj^vK@lJ9%1hI<M&!Ai^J%nTdOPd4ZPrXjYImUPexr0hwYXfEHvO)AC
zh(6jt6o5;*suk;cGbv}$p%u14k(kMi)Ee~@@f`f_5f4-;O;kG>v*fPbH1kXDPVldQ
zd%f(<4aSBr1;ECt5{HSDivnLZGe7d3vBAh#&2ID1s^31FgCERNmCQ`7f7urAN(4C0
zSot~IsS;`}gIotHMBy=5+9i_+_{*`H`mMAZd2$SGR^IA}yzz)Ulipo%Ju?AH0_MI(
z4U3F(k(%uA;zv!GFx(mM_D(EOtx}A4f*NCYhYe}NCx22jt^j?m`DxN#B!2tGtDcXV
zL54f5f$)DhUnXN<m7%Ogz$#s81`A%j(_fC<-!koA89DZSq8|&EQ#x%Z*Xs8}0xI@E
z?X_PxBfuE5X%_(6j}pMgQ$`~iI)%nu3mu3^R*Jn+@tf4J?BCCP%})B9{L>DO1t#Ob
z;FPHiA&W648U$Ss77q%;N8Cku_r0L|Z)tQV%D!!cXp~aQE16G^v35Mk?EN8n;P0Kb
z96-=MOK$uiWaR-w0*0|PuzTfY%8jm0XA;5!d`{Iwt&PuTPKGan9-EgjB@5GS3k!C5
zFrg#M%sI$JW!XSAI%I&^3q+QXdhr991*l6M{tTr7mgO6ZVQfsyy#1u6WE=xgxhfN`
z+wH}M<K6-<CubGtXDj6^&%2+^BUL6nXwr{*p*yRM4>r$6Th?t@9y24Ix$IN&n|zpw
z(nL+cf;Sfil5amEi$moLU<F{$0u5XqHUgRkRu=l{B9H;O4&Vc1Qu%Ixdc<wlHub6n
zrvZKOXHJeG=sp8wjvzu<r|tQ@@@H~#((~?W(Xn>kI0&pVb~k_K#NO!XUd%(U^Yk@4
z-+X1Uyh%0GjuzclM|h%of*I!kB>9}R!2*Ppy}sSUy`u|a&ui;eCX2d*jv|kfe66BM
zrr=4SgMl^owSmZEtv9Ynqnf#54fbhlY%C~+VI29z#RVWS_%FI;@$vCfJT5o=Njx~~
z8{r^Q5B~ZORJ_$$ZyD*6nDV%ATv1tcum{Hj2H85Oh62+ZCMpM<5Y7~5F;v61aB%$3
zEhzTf@Z5V~Jf?5!a+YMA#az{E2bQ{Y>eSogoU6-AEmvsyBG5P$KW->4J{U?9Vu;&@
z=RbeD<KhZBxqu02xIHOqJR0CW1N1A<(um&Z_m&8zh0j~a-Y~*JrrVgtR;5j#7Fne?
zbLr$F8;nRQMQrQ+vfBDhzqgTn&%%PoP*I_~1XDN&$HGzOn2j^bU6pd>@Q|u4paJ6d
zxCZTW%iDEJ(U_7V^~&AyNv$AvC4>Xea(+q%6RhmTGSbm$)Yusd3AOId*8<wrfvl}|
zub~#y+ki{>{rfi$&?oVw4niFmRZD(X%v$8<=5FrpE&z$vaE4fIZS6+ix1(HH_pMlU
z3kkRPse@-*`9-^njg9&Fy~|cW`7-Zu1qJ(3FN{L)IaeGBZrsnMz@d`$PY|P80=8xi
zkA`jiX7aa=sm~G-Q`EOc)HY1GJE@#{B;(W`)tZhlCk(g2OKuX&7{0!~1_Y(tc3vJ5
zSuh&>4?rdfBFJfK@|u(ast~9(rc_oPcITV{1D%wV1Zehwfq|Ks86b56_Gkt8g|zoS
zfP@7|owBml4mwFD?P2<_;=L^>biT@34=4h&je%I{@am?R`3k6+?f|>dfZxas$qeEd
z%9(&Fp+VzbM4YE@c84eygen$PR5>E5Zi85pAIZv3rdMP$1$EO$gNRoGd4cC~21!V5
zgW(`CS=$%5ryuX2J3E+yzc)6*X6XJ;6Xkz^^MiR8N2ma)*Zg^2;9}i(>cGNn7(A!w
zEkPGlQu>xp+g&_ucHp_s3b>G*tn<%Sa`Lk@Yd8G8orxHoEvi|zA#9V0STT`wIj`r=
zhoweYZC9LU#51V_$t6ZFlr}l~g<g6+dg;}n^GBSt%99*D;k<iQ`F9M%q)BDz%w`g!
zWbclHG^Ji+)s*X+okxOZ<au&p30>(wDy2>(ZS6blyAy4}uYmMrsbJ>eq4U9=HD6e>
zK_n9u6~(}@wY6<LO5SS=jd!TM0i7d#J-v){Y61=mpaBy^5Ruyj$~Mhf2S6XSHv*HC
zKsfj|=tdb^|CG$4dM$QlMlE%4a&%NdQIT~k4TuhJ4@mB;&GrH5DyyR6rlh3j$UrJ_
z(8v~PSUGD<Z!N)WASKFbGPcKZTQ|?;zFS3wDN?ABJLgoY!Om#z71FtBm(A@%7eymm
z8jbr^l)z7H%kSD=sMX%Xb9KMV7r%YhEM_2S++xEd&a`~CoR!Mo=jM<)u;oohmOQ3`
zT0Q#%saUgjOG0Twl8p7oJ+wvf+0KjZX9XXMUsMnW2>YOizFrObi1>lpPHS)qxE=*1
zE>Qhp3Pwjq_oeW1H{hDO;53S8yE!$S1LhS_ot;Gwc>zfmbz1vCfa}98nck#~?PO}f
z^7485Z@X^OLWK!;;ltnAAU)wFr&%R0YIFR)>$O_i@9NU_6Jz*wKq##abtrJ)rcHj+
zYh&{&+3o#PNEb6`!!-g9PL}`1PF6MyNZvpj@6yrp%<+KnALhR0{YpHHxlgs|VCd~^
z-@WST=iMW>H`A{9zaj5sFwKDg@(~&jlfyhyaIv+P(g8tP`mhmsS`roxsYK+w1)8tm
zTy`RPVn256DF&+lbN-7KIO@Q3W3@mugn-P^<P~pC`5TXC(wxkFNd^Vg0|1r0FaL`E
z!hIQ#U}`%@3R;OkgQX-F4HJn7P%5Z!mU!yx-|ctt-HFfdTRD=uc!e0|Z?O#|nWzka
zzu9Cq=uYZJQ;Hb{YUhc$;l)=?S+@VwSg0y<63RG|PLnx=lKC)6+m4B3kKwwuKNWE3
z{&L?n1VSG-AldOj=a{o<jcYj0`ts#V8%o{3xc-xn;+VR+I@&N@5Lx(dHf1Z#XD5b+
zegTaRkdi$m1FhKorblQ$)yhqFP64I5csWtoQFwNGib|!i{_cU2(kh}`APJW%P2k7M
zS?^|25@&K4(f|CMCiayEZ%v_T0y{JFN9j4C?Yv)f>PN557qu?*kE%r^nAc<d0RY-y
z#E2wsdG>6eimDOt_m4gF^?e2dB7XylMxtlS+p|vY`rp-Hq$y^j@TBq#)Hxtyz9nsW
zU*wQNbCN=$gpM}fQ7sT?9!6QG^&}z<-o!Q$_(jAf1ENd67HEAz&~H4e-#a-y1^9yZ
zct{9z_Vg+PH2L?ow+YG@78e=$>LhM^l$4b%7V4(=&PaHi_CeMM%}F3YodS|tpaag&
z%cEQ)gDh=DwCK)8JSlL9o0}xiF2ij9@dLR)({LR}W69Mk8gHEX_B|6W6AV)|X&cx*
z<Y#6+{F4!OdV1;s0_;r(N5gA5&*+kp60pETKy21QU@4_h^zU{6?*CppK!7sJ{gH`#
zvRb1oQc5lDlub>GQhw{WsgF0aPG;01D;m5IC^7Y@6-f|?5FWQaUyNOeTr+T}?@0p%
zJVvcYNNZt4F{uBdK;XYS&f&MWvl|;5bGw>dY-suf^r7z1@<|g@Q_zrgBklaZRI9yL
z8VBOD*L+7nmz<E0K*a4JoPm?q@aQwE{D7UYy{nlRzOdkMl^iFPMm>9tUG@9ur*4c^
zG06Dq5$}KSq5nkl+461%5D88V-wD$opit?a1?kb%{<#eRFht4_WYETo^+dt1KtnhB
z{Dd~J{JVOC+eIXWB)FZ&&9=7XT$P+$G!RNp>4gC)To_ni6+Wx3)dLWJu7I{rmha*>
zXnud}ht}5AXgRN8kqH|ZtYO`{LE`oO(ugfbOdNZYp&Qr~G+dr?RW1duEj=z@wA$Hm
zVHBV}N&J$Q{XUZqmj@qN^ZS8NT0f}uRHjba6V%pzF&1x=??51J1{M!3aYbESL9bnm
z|8lsBpbzNIa&uAwZ(RQ$lyK|+TS~aXBuu0{loK_mh4B!Mc9|Woyed{<^}t6uLirc5
z#hr!4BnXdd8J+Ug==8#{idDD?OU>qRA3~aZm^r-oT~Eypy}zsjsMGdb)@;6}9B8D^
z4_38`96^T(L`=&alZzO2pt1m}ZY__CH6SZYNg)OuD6{DjAb&quPqb97g2xhg5`J|x
z=I%P~eU;@?Z1u@POW(jg)1=&>GUqkBl)0Y)(No_rzt)iE_wJ;M+!%inoumYS%VKD1
zew>`-B-COCOnW3{mdlsn;|Xv^VTA~E(O@c}$KK7wQKw7{{~UrK(fb|w+lupJ`74k6
z-o^+_A0*rJ`t7FC=xLpyDIX<<N3@&<zN)*qzMOCP8twH3UJH<;)+GH`F<q>L-Cw#Q
z6uIytXwm$0Hp2CwfXoNI<-hg{0ZoVSEI`<eZ&9Q&)efNST+jEAC~DgxPzW&yFe9&t
zd*3!QROejH<)sT_GX(ssUWivnY8YY~M^0QncwhAz7c~A(ymbD23|YCa!SijV?%kdo
zXoEh}z5I_tSeH&Cl%?dO(s@F8>}+z!6cd|<e8xFmphnugt&_qIo{*Ye+tX>R>PA?h
zRc|KasVY6sRWqOF%v@zo4i!!m1lGtITMR2EKKHdv6GQZQdP~?yMo*ziGVH19(d2p-
z&Q&hvcDFB@>3gb#<L0Sv@%`&RhXq0);?EbnACTHgEdO#bYE-LLmmz$bV=`>e+d4j(
za8WvfOHlGvR}~szwCeBJu3aTp_d?>Sy?H=3eX*$$>Qw@vf1EQd;JBXvj-pGr#8RN3
zNMU<B)2HJWh^7kFh;p?|-w+xHqkLnWpl+}xcVfbv8Qn8Q@Drs>=}XXrm5A{Ju-TPT
z6PcUL6Q6v@5bvrAZh5o3RX&{63pXkFIz`e1b1!<$x2A%8!g)X!)Un%&x<G&iBk4$4
zaQ$5BB=YAe937I>qci_pi!7T?60~P+1|ffgF)FKRzg}D4my?ie*bv28XqeK8=Zc_r
z0{HmBjA}K+xj&I6H4L3?MsW$4I{|Uw+j0<mlMdT4mn>St{$RYP(%yjgfJmVLZY!MK
z#<QrvqR7Rn5i_I>r+=<a{A8x9$3#*Ie0>DaT5j&7PL!LX2l<4jf}oqfY{$hW4SL72
zN?j0;8DRS!T}qdG;iP{D|0cv=q{Vy?6vO%_<q;xSg0Z8xL?aZh^@<1(kF|cuxdoU6
z;r!iE<1c_ql^0`9?}=!=gx~AsHC!dyq7!-j>Q$@h1YRl92p*w4E#NZrD%;?_PrQbF
z?+fs^pvJ^<;Psbv1~r`rY_+g5bsJs2%Pih`Aq2PK6c8Hfx93@%r~w{L&rCezmoc0F
zpaHD#FY1=Bc3{xU0W~oM9ZcJ*oW`)GtcJIuqxHpds8POEXw5C<VjUc%BcH6Mkl#gI
zCP6!(s*S-7{}LLI&qshnZW{kfzBJVLbgu*QMZb?8O{GXFmZ5SmUiaiQFo1<iKN;+9
zYCd(4J$=gkT>my4rI=|j<q1%Y>|Mzh1wip${-<PZIbmDMvH|@|P`-lD$Jy%a_kR>Q
zZ2*on2bmNS&NDy^(HZG_bhFj$zU9S5g9$d5!O;oCB9{Wt_5`sgur1hKn{?=T@-MkM
zPJX?jZtK2G4IsDEzyo#>|M)++7yJMMUt7RQW%MdDVKYcl0Adh7kP!LF71hAK|3V63
z8oH{7e!AQv-^X|52ONNdtr=ak?AvqTickR4i+lM|l?p?&mI(Q=w**HZtc+3=1r6l<
z0I-h?{;bSDiMhT*7im#Y+HvUTiE}mZv~26yTBACq!t4IBHWBZs&0vKhM4uXVFIK?m
zZ{4Dq`J=rXHgX&8I}y#B6joMN&r9f-vXbL#HIg#I^rXXOrdn_UM&Dk;0GLnE=>*j0
z=ieV<ZSq8BD<>!1=&Vl@L5ys0aEHZ#_Nib@OYX`4VDV+lxK+}&ZQ1xjfgTRe2R2l}
zKp%5Y<Do5w-g;iZ63GPz_6#N1umAkQ+TmBsP%C$B;{S3{6p%HL@M>USI)Wp=w0Ym>
zfT_VfJP@Jn#M4r0{ZNH$Jsq+V_h;r^EpJ#btRaR5=7j5SlP#G)tH(WJweX?p%XEYi
z+wdAKy|1_73}A$en4<MFoRlKh?*ky9)Hc3D!kbFW`a8?!xnl-S|0gO>+BBPD;S}Yz
z&~nL}%yU(ol*_t<PbflV3JL-*l7GVpj8K9wWdRQBGrFL7+V`3FMdQ1%Pu#(zM$;wZ
z{7nlv>RCo{XYD5N0xSsj>xVW>Lxel#Nf)z!md&bg?Gy*Q7iLEmfkn*?d`#SAM}>Br
zYwtQP1d<-+@bPcaJizYc(R2Im8tZ(R7+8)lymIp-JPfN@AhzUIff3SbzM8Tspq~q&
z0hm+gQl4vOtcG{1xZ?<E{^`GnzLKPlsqBL1Y5Nmi&E_`^yGE?m?jgMwItNLz8?IQ5
zi%bk-AAm{d<VF4rphbMW*|PjAdd5UbBGQaK5xs>~TE;mFbJlp+C|X4<QeRaWQv^X5
z*Y8_6qG6%Y7?)ks%tayRsJ(@pT-(Un$Yo;YNHq-tOfLL<4x10fSL*2O=F#28@pL7;
zTL5F7Mj~|AZSRhw_PdLpdaME2<9|twj<3T6tsbx^(x0;To3c2wC-#@n4ZM!b$;(4;
zDfwAe$F3KKO6qD2$CpLd`|iLw=X#k!xVNO_WoZwjxRm4c!{G{vq(QR`!D)~?$=7-A
z_MxR}ky@djYV;F}<UG}}=*sUre5Yi3kLP1B#6%x59v|P#e%o#EIj}52pRCc;PGUPc
zx?r)5KC<v-l8$=J2e7?lWZUQt^RLlJ99Q0w&d6x#y?6*S2BO=$i<?`_RL8u$tpg_?
z^a$~`y|bLdw9>eAYM{Lf4z_}$ltA@=>5w7Dgo_M2QfcV+?hNmw3BJP5JJNbL(EmT!
zO+_*vT>N*cJT-;2^Sfgt&zU>JbsM$_+>CXs_QXl8>xVvFX{2oI9q}BoDk10wBX{1u
z`bV_9zt1#z2`g6EqzT?FKj)tPpES$e#F9Y43_}{axT0zfYC8f*q^=-z#*ci|TfR@;
zR16@Ufutq|3ZXt+$?8ji{b`TI-bdaq5miz#C2E?#*)~qvCe9kHhJj$lL=0R@Aqo?9
z7VyI5wQCLkV&{#ZTr4nlURw5%-@#%-+TG}^Cpo5!Mrx6<<62E)_%_Y(k3Jx!twM)g
zAY_N&{r2|QrMvw8T>bQaWOCf6aq`&|4#;Tyvl~|=jr?XQ`8Y%5n|c<LH(G;Y8Zr!3
zKubps)cs9kX-S(STc=Z@tideU7MnI{)W*T)_*;R(o*YZA*_PkwZM1UMC>(Bx!>)oB
zG#EXff$%8#_8}VS72bVDpMB|Ami$jf-zx8@yVwje<3dWD;BIZbo8~|FCu@utN*^rf
za&-AL7sb6gLudiS8jvbNkgZDM{|hW;r7v+)((gZ6N0!CcOY{Q`L`GC(3n<B-fRss!
zh92EU{y|%nUE~}(+2eZHC13$pRbqEf;SA7YyK(6o**D)(=ETRh#K%o~EoAk~My0z-
z7H|(?OHk^`X-%nBm_D~`?Kz%U+cz-w0?b1K93dv{1%L*$9YDkqZh=n!JIjADt)l@&
zR|D&k>omt$s#Uip_KIlSkw?*n9zz`EXeui3?-3BDx=<r^sm1V*XEYC|cWv_nF;uI_
z0k5d42moew^4nIoevWgg#ileJYHRmUlPj~rjR!?JQU;`iKr;Lr=xo9_Gb40Cuc^Yb
z2vZ#Q=|)$;&pJxCe*9kEm%s=(K%b@breP7~{9(YOW&KUdq=CV;{;GaxiG+p?(CAdD
zfoQJJkabn&@_k~$>OJ#7KN-u>C(**(07WHrc7~(uu5U)iSssf=Hx02?<ks$Kh3<Tg
z9XN+JN9M>rzy_+eU>6|%?=YyrLNpV7v7XcUG~ePK!J!xi7wVtLW5<7)(JclnPy(R~
zskiv@FQ~Y;a;@a<b_k<dVE>N~*lM*4eW@`ZRK|pg|89Kp<gR{k(&Lum?3Mpb${b$m
zNnh7$BoN97^Z-BVLih)go=G(O2s-A_YLs(n)QfW3jNr~5Kdbn8hi6GThb(eiX4nx|
zS51qH045*K2aIS$>dk1@IJB6^7(E+LoAsY&3<D05ii^Us5%u**)6$XqBX_@uD$w*^
zOx?IJE9`6*-~_vG2U~s0|F(ck6ur*AfL|Qia}&DTCXvx#{58GXf8_RhP@!;`1|0$+
zRRFs$&i3EKR4{G=gfs0&-mQond|iTbe{2@^PTc08@|&0lVDP$N;WI<$7V~$n1bC~o
z15->xKCQ2dC@YI87gX4}0vDisg1p-3E+l(~a!rBzPJ>b2+;PLDA$o&m?S9#^0ZO$x
zQT;b+kRq3Q2~>nH^NT|0^3Xw!`jhQFB~W!oh<j<dsKc}@`(0wB^Y6k@fO~6*{54!O
zhi=~a4P`6}H90)3(t&uTaA3GWqY-~RQO{GLXUSZiq;nrNuOicR!Dj!}0%+_8U)5E!
z7(a*2<NjJ$fAPfLoxo$!l_2H+QY2+RYAta<4dVfBxEju%VAVZO(OcPXwHUSC(7h(5
z2l73n3K{Gq<+OHZ3mY*u;-l_hRo@8H31M0lHFi{PmC1njPTNgOK{t;KGggs(YH1z)
zqQ1a^9o%XA#`;;}qv~OCM-o6AQT>1e=#Ez>F4Q7OZWJFHzcxWqogfgpg(dk@G<C#n
zVv&pJ&KoRCi!4o9bt`b=c-V~#<BjpT7ug&d@{iBjC+U=e3O`^KYMHJ(kK)eQ4dEi+
z3~(H~-3e&7P@IAF9CVpW?Mc?%$aoG$K;s2MC{{$e+*+e{DEkdX!$)VE*2$j%KdU-Q
zSg?0r;j{Klx_RFX?{VWIP!<j<o1gu5*&RIU`|JN>;C#x`p+xid)p2uqSJRzFWeL?E
zF7$fKg%$siJ%Cj!l5ajS8~~Bq+#)VSau<wS;Ta;&=U(y)f5JxBw*meBrPE`s4V)!9
z*5tT#Q)kP}_F@@D7k(*wXGDm87?U`EW^il_-S`+CV?(@P(^dd#htIuaXvf?hXLUS7
zj6S~<rEBl<4_C5U_GBG36vB-Ih+10*PcLp4!}%kb&}HH>8*OWM&s_LcY;2kVGSuM$
z{X(nm=+WfRxXO24;5={4D(l`=Nu3oq*_ZHz>*a79yxZV0N|{J{z_#ShB`@(A&XG<1
zeawoNfmt8^nqtsuc^r`EA?Ui;JW9Pu=gZ-oe=AL+J9<gdc$anC487HTxg1%m9<g`a
zSH(Z5u=bss6a-B>VX$ZzQ@1u<2mE*R$U+FZ1FvwaxA}UZGLG*RI*Z>s%{DDPb$?X<
zZ<nv;b`(XOe*nO8_G^$cx<^i-<BJ|bA7as4sFQK3j))~U<oC%D)Fu-Dlc|)N_2lsG
zHlD51kC&IhfoStN7?$-AW3%e-3~z4kLT(sqe!OtHGNki;)D@G5_V_Ntaw*2^t!Ik>
zr|MS%&4H`>l(1UOSC5v#dhe4BG(OM>HlA6$^meiG-0=j~{n+v*p0v2`Az$lx>-2)B
zC6{#*e^M2nSh+j<<Pn9XMQ%?s9{6g8btuA;V5U*qt*3Sc8%fX&ha2>ANX$aoQhd6%
ztPS3)$rBa%)}jfAX>gvdsC%SSqtNy6liS#u;<-M636SStqnOBOaINY}q&JP83nHkZ
zE?gFO<t5m`$HWW>XE2^zrEH%iBZ~nu*_^4bW@!kIOB{p+NQee9u#y9bTe8kQEpTiH
zoGnZD;j=c>?=At~H9+p2dBbyPmbw=vbZZ1Atr=<2dvblZGw<kzxi$g6&oav~-JeaB
zIr{Ly1FS40>c<bfhFldAUaWulGE!9cq6e!UfnyBv(2tVn!L#>|Grch*79JwfWST`v
zR}A!Bb++NEiI*hyQC_&;&3tqJlu|0xVkkSHuz-f>7NHlg5d;G;Aa^|Q>vZh2(J!5j
zKO`dQWeOI4LG~hLqgpUu|Lt(5JWZ#y(nY1Wtn{)n{o2VlVPTVUyDkLEpX;UgIM?w?
z2mZ=aC>ho_t;b3<XnWuKd<PqUS`)49%`Q&PN>fIOlka5Aw!eKwjuEhbN9EBMDC=Z&
z{+gCAm(`OFnON-r>-vt-ffUPcUypJEr}PVn4dlo7_g;)D?1=?QuF>M;NWNdba)y?w
z6%3lA3AZr?E50h9Hd8v=7H@I?1V!rUefeU<$hb>9(UZ;PbTCYc@_9p3F`asQ<4M9B
zi%LFIk#{rAKX+PavSnU@<I!(s26NJa625+2{SX@?PJUiM$5*Yv;I>z$Rv^9<B*753
z+WAu$=~l+aU)c@Zc*B{e_#;Mx9(u$FG!#T00TKVl4H|$Oj6FJ&0pBCvSI!^TVTv}K
z#EMgVrSvLBG^5zWgD7IYrf{a(*G&1virsEUgZG<|*ud*$x%_eFZ^2*tK9zRszseCv
z?qiF(@c)@qh3qbx#;?mHTvYSps=D^XdNP*++1jkRC`~>$x_52u&Qq{GYuv<lFBzN#
zQo7y~Zr|A2%6m_}pXzx)gNfv)>y*+dM9Y^f(`ikZ^5L}vIYySegfLxi;NcJ$(uE@u
z1Qk}>-jdbYy{)&zDW-c?eEXn7>yZ<VO0l<%$s4*#dRo^EzAMF);(cfSZ$DH0Ex!6Q
zNRi>tvcpW{hd#*90(|Y$Q9a&QK5XMt;aeTM?#@~AJKvKg;WCV+KTTk}J6NRUVq>H$
z<4>lg)X1oX5E8V2Fah~21QR2@NKqla6|vrmx6(%mV{+W=)=QWh{iEt&s623Z+(-LK
za{0*1#+lKv(vipREp;`_AO)rYq<p_t{BtXZ=#Vzf=VfniQNpYPxndYe`+*%fhoiLU
zgqC-6%MFXEW-}8}c>=iY1ilK-{Zl=t6+mFjW9=|ef}X#1*uZ}FtTjHdvZBoFrNjFH
zif)tH%t5u}2>ZQZ5tQv8(eCy$%_+MnxIm!?e%62#Phh+qH-ur9T^wWY`7;Uf>{Cls
zf(AP9-IkUEcw1Y~uD*pRy<b>r)LOJLb|WTNHKwyhzv}LfyhqdeQ|z-RLJ!w^!Sffo
z?|)3nXBJ^U{B(45<m7E~^^ngQ;w7bHi7VuIo5F39S=kJ{=F87M!mnXkB+zf4Om8!v
z7*~j$zB0{JMmZTev2nDG5pN5;e+ot(v+MJ7@O*>4>C%lRuSei$-#~Y_vq>U8n@JgX
za0@eYi9(KyTEWBs@q;Fh$TG#a{_XfAvllB$CVGRnW;(y7-Ujsb9r17APU!0svH!+i
z_%^TkDrCf1%&BHlL08%Xwgl)3G|!z<HqLgYSv+sAZ?4YxPh$ns_74x46Z>82Q&u@&
zz|O0KUYMSq-u&F09~ucU5s_wwD2nSN_|GGe3t=jjvt9Ah!fmxC!R}8=`5HdLQ)E9C
z9Tjz-SpS*SCrII_D^$hDWzONqffajvAAYFbCp^CGB=JmQ(rp1}7ru72w`WO2QT8XW
znSv97;Q0dXZf>mjAw$8dRegPZZEbB}mdW6>4&d+O7`1P~Gef|$0EY5aiky$u;hsF<
zI?R9wyza8lQn;z|EuQ}R_2_V5oA;~ZXaG?(XNjtXtMX|c@=YqG?efw1?-IV<8r_r{
z^f&A<91np|$(KClqoFaLtFlgHH;)u22QwvM))xoQMbOktZHSB#esKVvQ*?M}GWBa2
zoZJie1A1ZB*4E(o-3<_fobJwFfkQ;0cr2n{rZXXHA|9Jxwl+utR=<8Nz&6@p3s=c+
zM}7AgCpeq!n`+{*YKI^}iv->}n@Suo;mR_6ymv5M0$Fl<7uzHzEQ}k{0nWV0d}GeS
zJUId_$9`~-2J|q%xtaLadL1xP1UTMUZMzYB%m;ao6c{BI<L~)Qt5laQWd3p=$+wCu
zrc^P}T<DXs#-|CQ-j@U|<BG4I)YPIJTb;$~3c<1y#7c`oa;esFj}Kh<*Doe+SHKm3
zpE+<<@N|crIxGkeNoTIdG%dfUC}+vIrP0n%OkUsa8lkM1jqRkq{!O&rFYFm~nQkVc
zvVZ&#a}MVHO&2kAUfSg*F^{f70|RZA<{?X)AC$`tYp=JoTO!uhDAy~9ewZLX^VBIf
zpD--w>8rUj>aGucH?5LGw6eVG1w-c$h$12CEIctJ30<%Jhof2KLbV9U0qYay^wMg9
zM2g=^_pcXn%n>#yW9%lYQI3Y#6!V3jf}#=v`Cit;<OLoy#P4?gORETW)D$#&zU9ds
zJhCE|(kgTrRWmkWeikA8(WNw~ygBG2yi!GBB0a&$v^Cx(axA@DLxdXJn9$_xd_P?N
zCqOko1bo-x@6B|s!dyKBpDkPVe=zpm@l^le|FCu;b!-tIC!2_@GSjj5-egO%vUdpw
zNwW8dtn84jke$d@R%B+&=Dyzge7=wSci-Rp@wm@lpFWS{ocDQ+>v~?#>v>&Q3VY%m
z*VPA5$n>qZ_jw~!wrqeF>Fa#%oQ*opjN<D|P21oIQ}e4Rj+lh1tzz{;dS@lG<_7!f
zQU_kM87i)6QsEWW2&Zok16)5yW=9$~pGO{if{8$CLS%354V-3F&!}zv&BV4ZU#JCL
z?e6_iz`)g8?92HEr($V^04G^CK00^oeKhCvRaBt{r3b--GtZg`a^)^ketO(vG3%iB
zQTmcHYv`hqqCR83#LHq4<O`IvhXmp`;p~e2DNtg1_H6z3bEwP{c3t(w8XFrIfaL07
zFCF@>_rVVIhxxsPe2e=^f%^}a-QnslLbemD?J>0$#Y^${hl_kG5@W%J5osK)V`rvI
ze0E)vE)^EvVPk5S4_N|yi-W^L<iWZs_J^97*b#JN3;lU`jL;ND27@#$5^!ciA+0Gs
z`Qv$PFYDKNy0yH#{(A|96Z}uLw|dzl*2WRIyMMYCg<cI-!QE|yUv-$;JscPHnB_ff
z;xnVWRy<xp-_bk;Ds)g00zXW8IH3W0yZB?Fe%9~V(hqB1nmZ29vF@95%3t4U78=0U
z)VgiFP5N<46g54okmo%tE2xOk6IHV5KFV-L2{5%tm5+FTwa9Pl4EAxn|6xQ%+{dNj
zUh=Ymy1A|g@)g-7OY##dKdL1$C=}|P1!(0S^CkpkLyhb?M*p5IsHiG06~J*NZEkzh
z-dKOJ@NwBfY}FsMw*NTe>zUp0CmUdSf46ZC;YNK|$KQCXal21-P~s<xF%rIk`)M0n
z*<M~=22alrku%W=VlD~H(a}+v)nVzhla&<<3Ov3*^%>u+3I6`4LhE=w7bOSCuEpJT
z%*3rT9JyI)mL%-q_4X+3H#6lY#B=0N&_Ik0mmvcl8;t2L^pAnQIorQSHTg@u_MOoS
zeL1R==2?m<4C!CLDj^@;a*HkRdzrfTVX&R$!;#+axi{5^+$PMEkMFb;G2B=zqK)Bj
zG+>N!{`N4?qiV1v011R}l+<D6B_$+~MpV4K<)A72JyI4D8VY4?Y$#Akki4&SoL8~7
zW<&MoX?A~E#!0>P*;K`L^dI>TlE{4?9Mv^=$l$D$>DX^N78ykplhe(l<`8qWb}D^&
zc6ng%epYUgPZaV^ICdAo2SWLCR|216;UFxyL7I{l8!0+hyY#Kn)9%(*a%jr+^vs1$
z^V^NLh#2JTL|bu0F&Mdmd$n%)+&unRDHfjE{vSFObwl@}v<NyLceEQs#5sMmWbqs`
zWM71CTMKD;>aERxJ#ItJP(s6^V-=0hK`|VloSYo7Kg$aXRi<qxFbnFop^}hLM<Jej
zz8P}Tu=iJ`HscoL?FqdYP-_nUIUY2+TJptyEOFoh<2&DV?Jd>RyVx_LnHy_<EpYXz
zZbxg}xZieiS5mKDU{g69c~q$X!OjT$&1`qd_V)J5w=6|iOuVtn`1qc?GcnLo>_?_7
z+5Y}%9Q11j5hYuL7V-@&*2SLZ4|VIFI*~y3l)I%#g_&84nYlh^99~Lcn$o7jAi?_A
zbM49q*@c7SirccpHM6n;rg4?AJ1`)CfsT$p0}mek=Sp^kttqS~0$hdt?p#qU_H!40
znRTPG@)1{wcRQyR-s6i|4`N^Hbb(2E6v~Qw;;wMSvsXtkXVjQWI(%}sm+7ME(O+8l
z3J3o&$RiQRo|P&Y;hitlFq>P^eeB;osLZ^-@ZL}8yO5bmvGIx1%@(DKXrROoVAU-6
zUh5IV${n+n1Eub>uc%Mu{?km86<L2I2N%`aF1A?G&>)9UAlstE?!ds$$i(~)M}S_r
zQky<&)aKX9Ehm>p3+beskjDe`iQ97OlBb+pXC$*K{`vd%*MO*Ex}2;@CGbwrSqfZP
z!J~zdh?u+c7vm|-JI;Ik8Nv6u+d(QkE|q_6A#L&Z@-L`4A)!P*O0?O`_H+?gkGVO^
zz~`rN6kF_b536U$x-Ur%#2E0DElPc<5UURQS#GwVnUMj=YTf@!wQQ#H-QQ9iVHY#S
z-cKN(g2N&VM;H7mVK@fu1{$Qj?G+kgV!C^eFE`keC#aX1lrH_bJo&ZveP*STfc<@U
z6@m^gho678+j8#s%Pgw|xc%riD+(;|OOROh^szbl$ZR{f*o6kQGMJv8o^kK}{2NoJ
ze0^>)ym4PV|8U9+jSgz+D|oZjtl31cV85MT{;^_;gHryS#WRxir^{8RsBMO>o_NcK
zr|>zYB_-HwkQcwacpm5IBe_Fg!hp;-l^|@4OjnYJ)?y!=t(K&=+XS^SJ6}H&hKqf2
z9GE}<;o*p-9)0%TK$1Vc_L3*GW&jlTLgLN+^iw!b>3*kmL_5hjlcLsGqKYL&c$w6e
zQmHF$uv%D)y^fm=Vp7W4DZW=b(v%?gcGdu>FnDD?Sn=FS7V;oO^tDtde4MK&X==JE
zYptl1T7Z>@0e6i4)UgRY**v@;_UGtX$lgxJWy9VN!PkkAzZNw!w>>T5w9r>nR5Ua+
z6dOyP481f?;Sfth!BWR(o)<k*m_)>pjlL^P8M-F~xV|T-?b~aWJ7~03mc7tbWdW@m
z;yY}}XZf~&8Kd2_v9W;$Y>%MI{>_^=ngx2ppq+*GX=u`P8Ug2bNkXU9j7wz6`u3es
z9fYVwng>ffzAp9!CC+a{y=IU6-c-7Mnq=bG?mj%!E8a%f1PE|S`JOVS!_!Sqo1*W!
zEF*9$*aoUYC$xzz6j`n;-$-Ue^JY{Pdl-(jwU4;44Ln`D#MecNe$ClAdnA0LMzDzh
zX-A5anp?#8^hD=+ur3Q{Q^{$qX8r4_=J}tW!g<o4=BN@&XHK2pidA*{OuH6D!6u8j
z7DUI(|BuR|-kX!Vhr_}LrxI$v3B&c+kD-4Dw1<GY_XhOm?lTRlMgPwb>~rNGBi+L8
z<eh(pKdoVi(J_U1m*2f{DCj{T9}Z4ba_yx9KoeL?=x73#%LOBdqmE}%mg{`*PxX@!
zEY_&AOIKH?93_w~xr%wB`t8@-w|IetAF}JOT-uDjGdOqhDZaT41!sk!ap8koSCs+l
zQxRVgCHeHfMrC2QHV*0uUssr$;RXJ19EnY}Vqv5_F{e06;zLmlqt?7gh}_-T>8K`(
zqyC&uK(>}y;dQtIrbf+tp|_Egn$}q<OsaS|poua0HV^}}!GG4`muEx2pL~cYn{$_@
z%abSIJ1J0rhrirg!%@40gjP=DGZrf>WyRxHEjq>uV{azej1!f(oNHvm&*eQce1ctl
zvzPSkY}#JaUkfz{6K?3acCMp@Am~!6L&R8^i*`kDTN?v<Ox<Kj<i%xtQdkix?rEpW
z>ptVg^k~I6$0u5LG%kF_ehB!)$zE8}Z(2S&+>c;XJlg!Gm?fW34$VKHOUVB2E=i9!
z;9KWE7nRc5{<pBF@5X@X93y?^h-)&|iivnp?93w;65|U%{=&<daaF&{k=fPj_Xv-X
z@*eu1r-8Y0|6ACyK*QFps3@bKoJK}4YN@ArR`c7PVdu@04^fY<y;LLd#$JjFj=<Ek
z^30^+^FeDZd6RA7n(sj@qm@*Ad)~cuRDO2+t@ru(8Wdlin>{%(i2-0S6ck?GpTf!R
z<X710FxcI_{V6=(`@TtF!L(e|Z@Rq4N}ium7;d9DF+^Uq!fRWk0!%00>H54DDo+G2
zbmThQMi^GJQE4047~70KP0T6r;_c?&9P?942*v8ry>7mDWLx>L`R|3np$<Gtor=$?
z5#rv$FrTE0!Qh+N{21uX$OKbhtXV#J`22#zOGdO$B^x_kQMe9Z_LFa0C8i&hSdKCD
zwtp}V_zX(kT_!G{O2mzb$up>65%GwTlDl4YSNFs&HPvEhJqoi_j?{$C{(BU@v1p&B
zp+H`43w_K29%Wi=(8d(%1%9<hvHq?h@qO~zbY6eJi&0*l0d2s}_h-rW@oM_ey~D=;
z(>nc77owA*CHrfga9qn@)2~89$h**z54tiy0MZ4Q@7K&s)cbl+T6;npRrvSOgx?i0
zF)?U*so5)l!!Pns=C$m1Txs+3s5Psy;+3nqTRO%Uw!-8nNBM+}?wd#3@waz)>07?i
z&?!0o+kG=NG0K|Y0Afz)#z1=gdM)&10A}D3cRW-aRDeDb`b>hoHA*gK*R-)&hPsuP
zG$POdM8T-DMk7zZ=)uRKViS78X?&$&sg)n<w&Sg_U6`iD?;Gw{rUYxjHPQTk8#tWg
z!lsa+b7#+j5N`&W^K8s^Lw8wSJ>*4&ssttL>L)y5VPOLU18Qn&(hz!RU<WT++DTwx
z*g9}y_V@S0C~p6PK0zB>TTZKzr|eFbENZG(mvg-f#q!&Z*NmNU%TIun{;8I1Z|P}#
zw}UWT0v&Xj0NYpfuSdZlmSxsVLwbLL3s`~D2zVBdt^-cHS2zf*C04qv3j3{~7oyOE
z247*XJtpGdUJ1O3zW*)8$o6?RXrbu&%6ifjjT_!E=zo4dnfbr78Qf9uc>0;2pT%G9
z<Nf>h&}b%E(2a|knHjJqJ-v9o2K*j)1^WBf9_=kc;1ShUWOiN6{3Q6_sCZb^pKG!(
z5|a4KBAGEhA)c3yN1xxbS-h5k?yDpN2_MVJVP-<L(2<Xgjd|^XFf=Rd{_=UHk9LU*
z4b;2-BF}Fu;)TLle*Nv;%7b6KKJ;|9UG1a<Z@$0B^UKRgE;jyaHu!H^<Xk<iri9<Y
z;_^7kupUjN6JO3<VI`{&L7g>ZeSeCmn*vy%tK$2v^4lBE(Gw?MoJt3DVkET*S5C#_
zR}9~8ppD5l7JuI?<OXcvqsOUxzM1iRg;tQ%$6Fhx5BI7T|MhY>GM9juW9F<pzxB4P
zH;@^FQOTLlUmmF1YKFse;j-I}6-M(?DwJD$M&7eC3Dvp(EvO9iofCcde?anLU7B)t
zfaH5E5W!|-s8MzkDPQ(C*0+2hPw(t=@|4*_$Wz90?>=?kxcfO)5>peBSG+AcP#8aS
zZ^p#UTf)K0y2^mxI^SNKf+{t__#;Tp5WtCiS=;{#qT6FS?BA>~{_fzs80aVH_~gvN
zv_fe*s9WGQbO!$HQY-s$&RBJXRy0=2*3e|-UX2E;v(oPCcjfoiF8RcF!b-|-ZtXP<
z>Mo_8{8fVyJR@<;RG4bJp$`Ls7Wg}d-f$GE1zlIG6IqwLzfAvj#2Pb=4re1+_vz@;
zJ}No+I~)Q?wxyE08rUxttq;*owrW%t<UCR|SYWFct}eN}!YFvlK@hduW@?}Ilp{B3
zh6C^9&nB46qMkyx+PF^bUP_#5yF{yTo86RII4rV-@gq)jS#eqjB9l;I0orOyt-Xx)
z={?-aR?ZV6FG}%>e?ekxZ4D5C5A>+^^z_V7$<bqHM8EtAG>Z4)z)i`m(@I|e1lK+`
zx%bBg0#D-MHt7-<ycS7D2E}~=58FI@Fn#}8Q&BU7is|5}e}TouLW~J<GSr8Cwigq-
zC)f><6U1aA2>6ha#F+`=r05rEYO@*t2mM+~&eoZ?!iEtRmg}+g_|EgoaoQ))%qi`u
zPEO+y-AQ<a^X2m?!|i%#{gR`abuCl^oeOteO-*gmrA#OP)#oruzAs00IPI5BT9pe8
zZ_3vw0+g^?=>0a5T@1&g`cbiS-IEnd8)}p^g}DOHI7jZTbdrhQ-^Gta0oVej61RWr
zuG7N7-eTckq2YM3iR8y>m7GnWRpG1X(_>>}J7PFShp&t7PN=>n$%{UL;hH%wye{aT
zz;Ms3^%vJ|o5yeIVx?|`d(sf{f9kmZ@4-mzgjl>gfiG;<ElHsQ;5^@$>(93HaNCil
zJ-QX_E%x9iLV*N{Jk#cfS?t|C0p|tROzMrSm9pU{jHY%k+i!b=Aymfd)m2DHdF1B}
z-TS+fZU8q$oR`Q^wa~!L$jGSM>WvbUL>RR^hK$1i;$Pnq259P8xhx>e7lY+b@)&EV
z&^8n)_Cuqmg6d@CLlmS+>;zE|*N^Gme*aR2+4=9nC@C!cnE(AHNdVLLVeQGU^MQeZ
zL_N@j4Ctp+xi)rvrIk@}z^-~_4K>M!Ze&^6wh3<DddBOdUgdV=GjH|D_2m7fS5DSg
zJqVzPSQ@m3>uPUjsT!{J*v4Y-)tyq1v;UK^J|R5ihJM|L%jdG#A)9Edybb)Qo#)h@
zH$hnPbW57Thjn9w2V5t=RgLq{FGg^m4CzpmST<)4wGaQKqXUM|V1s6(>VIn<?UlZ&
z9bMpwyh78+nq=&_M-ox_^S1}EY7b&#|2d}cZXG-MoLn^Qxk0Tvst4rvP~NutjyI6t
zx=#IMF;~)mS^#e$6gbc}U`D9I%+#wZm=$QqUb1o2Tg{I|WT@Et5mI%w{dSlg^9qT=
z3YzuG8{PBz;q)JZn}R)oK*U-}=_+b#p??wDt|`dLu^q5?Z$J;Ax#7~Reyht++c_Pf
zxU@vjP)wc&*UayrogR%2;H<C&Ia5%gNdaOG8$5!B*(5RjB<n)`_jNv0VgE){IM~l)
zXUF>couJn<I9bSWK!Yy=C>OnS2{mQ~?R-HuU*)or93KAhRwuOn5fm)a<>5ztwsagO
zW<+4W6q>Vkc(}JDbm$%mRV*mN!)*4+j~KDMk3u_+;?sDdaEpJQ;9MNeI}bt&Vhr5p
zfh>fW7RmlWF(OAIR6@GS$l#)Q9`;THngBdqTF-$a8E$G(`A(Nv8o7mXHcLI-Bdcb}
zX=|VPV6;u$^mfs|@g8;6<+OWuXUTg(ISVgzi3^GQ)55|PDC-l3E_W$VLkidk8birR
zOXt8wCEazp#%DuCLgK$Q`!Pp~`fZi9cFVLW5vovy(&LZC8NX(NU@TLc6(w=XYF3*Z
z2EPYmSg7b(`=8TEd}c!EdJlF2k^){A#;&HaNv*WxRXjHBWI#{p^8m4;O=E~W<lc>L
z?r8(#+%5?WJ7L?@KgSi{-~YF%Up|FH-gqP8N6p(kdpwg;6l!)RY0%^9$}i%@s+D%$
znv~Us)-}vDV??w6`MPY4pntcDA6)zDdA~`3m2mfG_zUKup_1u|R2WX?KlP6Hwl>0=
z>@iImc-()kwdVC3Ul7N<Ke5}ns~xI}nR=9j<QcC*s6dqgFMf&(bgP`4oT_i-ILYyw
zA3}WEl+N#_=C(sN$wR5}ZK(DN2bqsPSZPJqS2g!iZc>%^e_+2t@Kxq|Q5D_EN2qcA
zVOWKT^5x5y&QeS`X8u3b>X)KYL~c+GFYs|Y&&enh6?nW2z3zSdapm7QS^KC?<=xI-
zCP2{yLV1^EJ+R$$z?c{uA07HVdqzh{eC1<Y(L1Yc-rcf}|B?x<mtId-lP)&wJf}3r
z1LOPaiBC$n0rmeZXvf&uBREtT;_KHBe^-p(MN{KQ8owjmdMV#4P=UpgfBs&Tlqrpe
zY|Py#3#w}=v0q9s8R1gV=PtY-{2wUD&jYvl-8(k)0zlsz&raFVF$|`T)hNWr%VFrk
zW5<?^PjEX5_4um=tjhli^AeM48<EDIL9LeDchTfHf#C@DFC;8+IYIs__`?E<o9b3N
zj$*H0)z+WbO5(q91P*Z^#7p4ff{<RCK-4sFlHY&%KWLuJ<k+Zup{$7@xKmJsf%c(|
zRTW|$e`p8|NirPixSwo^5-o6=ib;2QAS21k%gewJFzXiBboK$`%-Rh(19p_XO;ydk
zm7=3l#*!Ecw$a5ehd1}E(Z<-LnV%$-RH+Wn%KpZDsp&;71T_f>qtKmx1fcQU+*}l^
zCQu{1+}!W6{&T?QKe7nJ5jQC*HT6lpfqpF$dKhSb*QNMIIxE9Qsz=cFO%1fltxlga
z2YYUF{~<jVat|U`bZ}^>yn+HlIxxwQDkzxyEH#n#j#=@#3E-o5mkk^OmJS~%EVn+t
zfS4<4uKC;vU(y2h|F)VW3|I(ghPe%?xVv=3IQyKH0ngT}Gat$pH>UAW?7=nXO=e7A
zz}>C-nyMq?|DSX7H?=(t4gAE=uRCUtu@=DiT_-2kYT9x0EqN>G{Tz~st$UwB5))z@
z`?Qmf(L7!s8xQCgbsp(yoWC%ap;dCirQyi5!ATn_dxYVJa(=+q0%&UA-Y3cH9_%f(
z9wPIzNEH0_{<f2E3YF}@iHq>Pz9ciTvA&Lf{>+IV0G1&#dA(`HCuW>7h<{%QY>g+!
zT6zxztN+;=+>>Mx4$I|gLf3?@&$sZqH{#;rUgBO%b&S>XTbiQ5i0vNo6o2|)%5?=K
z7Z1u%;USn3{?_B0eEC`*PY!~pnem{f6CYm%5DBOr2<;G_g_&95Kxn_V&T?j3e>kAR
zrUh@fIz_Jhjr*+<my^G;BlaLm)GaX$l6a#HKntqgK9;<@Z+x1GnPKps`kn2iZS$$f
z@t91Dmmgv*r(r9VSpE!LbW_eJq(^j)KxIV!|DiHg>BK)SZSP&~s6Vjzxyvk_lgG9B
z$w)VJaeV?OpLmq><fD9V0M>Dt@6DXp0!RH^d_DvQ(onr@<7ve+ujhND{-t(N`%jX`
z74PmYA_7Ub%I@{;sDIHqjvY4~v`~qkulR9f<n7ljy?1L3++xO8bl#o$Nn{gy6GLYf
z+kNOI;qDe`aq?w2N=|TbW~Zn5=?Hv#jT|mR{so9hx|y2o{dvJ}uXdh00`VP+DPiMJ
ztsN*6{7bL-vNVsKr54qL_>3YU>iO^rb@I%>O8L9Qx#x7Z!2$Z}mGg4Xor@~@4(*EJ
z<;!yMlAzjyu;}rQhih)hmNr%+-}@XVS4qgW;$Ml<#J%_Y?|pd4`+`+l);m_V+H($U
zCLva1u<fzH3)KI?9rEmq+JEP<-Jp$d2fA=x<Ty`2m93G7m%{uObweUDTd1M*mRsfe
zhD8vTj!?FAJ@zp9W6AxWM2f#C{2P@GmQ(ikrC-&Y{0`AdChgNA4%1ST;xX*{&(GFP
z$Z>+YLvq*Rk4NqK#6V4L7nf#JYT^{TajP3>|L425JoVnxRVB+Olq^mvPQcgu{!Qee
z$DQ7`8ox@%ul$;RDd+mC%K<Y@8&QBuU^%&<GLZBa0uXCJ4*up%dDqGLIFa$82e<h7
z(Z0q?i3dr=I@nf?KQr5KhufrcH2!aQSUm2}Uyrex0hEY=L70V1LBY_@wRTDqZqiO~
zzL=S*4YqEd@c6n3qoZdV^7?&`#Lft1YT5mb!7xhk;!8T^*4mJrED5?Wv3tZ^FUdV2
z*pYjX++p9iBD8ad(dR{*=#?LXr@S55`2R%`%w1?_=<f)UHe<Jp0|WEYaiXs1FMP{-
zt+xqL_qlrp7l{m4xEB6~P^tVy*z0Pyf>kAt+3V|-BxbfnLf_ugv}J&^mSGxud`^8a
z?OU|C<ln>A3^WxB;b+Up6p{OMk>k2i;$+n&4r7XPTz|u}@xb$S85&Xly`3IvrdSA@
zb4*?|i($&E%v8J$qm%F4V2-+hPIqUS8F|(4kN@|{+P|M2d798QoBxH|`bHMA^G}{!
z==QEyJBZBaQ>SpBEP9hAX|2RV(sH+)?BxBvDiPT0IHLB<iZo>HD-d+wQIuq7&axO5
zb|Jzo*)}qu7U9Kym024-v1K?48Qdg0kY&tM%S<E@v>qzz)xW5o#Hr}b4p;f%a_jJl
z%gqxyyA$YaE`-ir*>mqG(g!U!jwppzv9W(qsruSY!#d9<EXs^H3=6sMTVP3n3Nbe|
zl><8D^9zcXk5}k=zkP!iW;yFe0@0<tbgK|n0eEwo-vcFe<{LpTbp3%rOZd$8H@CO6
zYnSFvsBcqkJ5h*c-q@U2>Qyz3yfWx1>T|T4WfvPBJ~{L!T?$@ua*=JOfCxziqGj-&
zOr)W2{c7t@LIYVL`z`uq)#CsA_Vz0CA-C6;5$Z)9fbk7W5`-rP7bJ-R{mP@Un%#()
zNL9O;IVhtR;}y32nhN@jzIWQ+ihqk2JI?oZiyVI+9o>hSO2#cJ52F~K`phP|fD4&j
zRl-RYcl*^OQabMC`MaK0k(-YP6abFNSnRdx6V^Oj`gi?pWS@xB{_U2!lp*m3MYscT
z-<#kULrBJZ)p)GN4M;Kq+Z91UA)$zfh#-iM<d?Kq{9xgLM7qvx`Ks~$DOVOaaL39@
z-YyMO8>Gp{ntvm#p?>*NCVtU>BP=BYD8N}wj~#ZE9u>F}U<6$~nqH=_H{2baSemvt
zGVrWZ5Zdop0qC&t<?z{(S7glo`h=+INpu$V`s0HK+Sd37avFI&(3E)At@sRcaWG|N
z(yInPRTDXe=H1%}BU|RNw#&-!WV)-GapMIh<JB$Br$>fr8xjAwA<Jwh1L+vn##{J}
z*55y@<y=OpnFEXf+c8qH>wS1-6np;iwW~ewZ|JrUF}obY#@gCL=<N-m3I9uv1uusu
z2(Exm>rXg1IPx_{vz6b#c0g0u{ESnKbaWUPWrV?4a$n2WpoBF4_3KENkx!4!yOX<a
zb%N;TA1YiLm~bTGIUp2bnezHIn#u;a$u>Nfr6JlHN*WqNI~CCS;g@=B+xUIFW!1+7
z-vXUHqGZ(Ty$`|gLXLAghf}Y{)qZoX2Q*G<dHr>Db&p5h2L%NcYq6cX4ClM?RwpE_
z6fiKjNvs|?Psa4`-}NT#nuAF)J&gZg40b_~n#CaUB#|=?+p3dAz4%yj_mJ<tO^m?y
zvhHKVsM_FOUYJdLA{U9A4W;uV?;qY+5-kVjmT?@8#4l(Q9b^e@eZy*bK=}hL()S?-
z`edY|GxE{z0@9!zfSkF7g~-4TjC(@(UWgtQ68Xc!!;Nl{vm4ZWfLT{qES97Gpk$|K
zd8qh}@*4>V72pm+;LSsh5C0Jw6y!GPPxf4Gz-{{1uX3w?D`<xcG$zKhSc~1TuEt@e
z9Y6;)M0UlW;ThZ@yA7f;7}xWdD~?*&(-jmWfGHtpCO!Hl${aZr6vecT5P|$0ARf5r
zQ9r8h4cZeL@|0L`BWn7z<Ela#{&^wypCLImUoLc%OtsUZ1jggf5@(!TSZ%dY<9WEk
z<`WH1{kzBEn_<k-jdV=|H#10{Ao?xPljW)BNnjv4cef6&0eKw)W@%}kVj&q_0k;%P
zf;UXjx|{TJb{DvzFLAfC_|qVF2RpqC%s#HaF>ENn&L42lfD_4AVtS{s2%icn1@`xe
zypOMSZ@?0LqV3)Qnks>BMcak_;r>)Cms$Jn#_SC=n)IfQ<5W`+WZ>|+p?iZ!`Oq*N
z;w-s1C^duj>d?8O<l(os&=<*Np)Uvv-BSkA?q4u+IGSn>aZ64S_PUF5^YFmC_mYyY
z081Wx+$2NhX6FBRy1}ijJF%6cK2S2U>fp=z0a?yG9yO%h{!b^EdV20h^)_?>#y{T(
z`iUwKZ0y{meQFEgy_S~N<c7qXaIC}&@?wxa#K}ouz#7A#uCO=h>0f|w14$r=@*p<H
zSPDX}tg}-tPOcnUN7ItfrPq1wBmI}5B|;7Gauz9Gyugwp>C<PX+RnWh(wOPrziZ(_
zK1*7Kz2<eC`<k7d4H7i?sZPH~)EB76#o$aF0#~;3y;3!=AJ+mV)8vDxfBd6r3&I0f
z-2&w|=TUKT477}4SnN=>i>;It_Nyzn4|u3UXhK}5UtJEy5rzy!bNi}@6dOwz1Vz0&
zT@UktU^Oe?n_i@!Vrqz$53b+#uHd!RuS%G9_+I331>D<H>Z*LBkAJxwmSFe=kaPG(
z^k$AjziJcOBB6Rz7Yk@<X+y|3NK0JqiHbJBVez-N1DW2ZHVAy#^{XPWNb2QPE#9Y3
zpJ0nI7>PHWppu#U=jYT^0g}5%7zmU7p$hOFa7{6e@KnHkkuBbi5z0}kf{x<PsI(9}
zGAaHs+6mG<gr6A>R(y_GTgT3&c>ReB47`9smM8OKYM<h}*l<l_kn~BuAm=o0!ZNzp
zoVEKP*}jsg<0;%`6;k!&c$Vr6<kD#N-Mg-AYCh<?i(HmqXln)S{G+hIPieafxuqSj
z$8><!Y=2rEJb6sVOjy!8peEsrVdubB5NU8A)9H!`N8xm*)JQ`^bjSk9V?dm8^R86d
z)9UJdEb2zGrwiN%m_x$9&CGXG=RK3wZlUN$CckHyKVX)EA_jh=?O?$VlTb6RVK~Cj
zhl+@TLOU2o=CCQ4!i_%Q&g#Sq@>1!wK>e9*S%xPqqibZL3}_Xu&fW=Mpj)m}o=^;t
zejrmLQg(OW1Mi8afP2Xk0?*o-n&g`4eoaHKh5YBng*}#X5KW?uqN*XYotm1ORZiy6
z3{BNYuBR#OKw7szC*ONrKUCtz0WC6kL3}*x`}4p+5jPdIn;;)?ayQcRr!Qd>o)^xm
zV!=OJRGHq!%`a*Q^TcRKSg!8IMeNROk2uLHd)FJ4JJxxvJnfgjcykRGT%#3o9GNvZ
zNtr0j!8j%2g<@DNB4oswgw*fkGMizO(kVP*xmTp@WWB9TO_*O4&Pzl0Fknu^C7Hg@
zYA2D_%m`kpf$RTd6!>?e@ZE`t2Z0xq78o@WTUq(&Rwo>(nd>W<<aZB4A_Lv1Br#j+
z?J17}YO*!%@QNX}pNWqpDi`h-<~>+5qZjX$ya%nyl$DhgYo8k>3qLA(pyvg-GgM|q
z1~DVf%b(B7215S4QwWlZ4LzO~+cRcH_njsU5#Q;P$t*c&bR8gKRhW*VBy_dAyMCLI
zA|f2Z4{!R~gY=&F0;8fB8<n<!qKO8^w;~7_6m3SoNLQ4S;`iF)?%sfmBwaf6o)>h&
z;YIyGoZu`T=^Y{oWAGJ8*myqMFCvPQTdT=xFrgBCmp)y<W%+9OEw-BPf-V!HJ&^A^
zefl)E^-I@SO8cD@Eq@605<?DWwmDS4J#S_m+gRU>G&;$|_~I7Ay%#H}gAs6VB*SoJ
z-;forY=gr~aX$jR8zw`E36$(0qr8s++u=PhWx>`_!q#E0*E4x(9^o*ObxioKg#t65
zWzjkAjLrJ&1G#YlTh_A&@91#=1{^|%G~x0JL(ZV1CKjjyn9w7(*yva(A90dV_7+L9
zxce$X7L&|1?_ogwGVCN&Xiv(^dj#h)$P&7keHt+K_V!j}4>gzod8e1d*L@Hf`#%D`
zF$<f=05mc~%ScNPgCZn?Pg+)XV|Ddan<+v^gb*ldQG)2LP^);V%_ZihZgf)8%}w1+
zFB$or5D6pFkQtDl7HyP&nuUDuCP+l7Jwkqd0FeoFqWbXRL*(8Uv~Y1SgpEMWT1TgJ
zan$R^*J%U`c*gB1xvo8p$<hd>u|=$bj2OAN4`3c>Fo^2$J$2@}{Ew6cu(ZI(`d1{$
zN=s7!T<o`^Pwy7qm*b}(A)$Y#iHOPW2NICf`869;fb&*RAZWY=;~*4ziZ4@eVVHg$
zj~^eVC{1(Nfro>9F-qChb*IRvG5FXmq_L$%QfrKZoxNa{hRqf6WL39te^3)KD!@<{
z#&q~44D|FQ-j85Z+o!L4h|ab+4_D-m1!^%!{j_-33=WgUapd@;ext)PGVZ%;dOuP3
zsZqP_rZ<$m5xAI*0cV!3#sae+ijMS9KWBra@->iVcz|rc*IvQH699|=fPj#QC}=Jp
z2A{z|LVt7o#?BOxdPw=gT|`viw;qT>8$_V(pwkY5*&*H_?1A%yh}0dex#FzhQo~>c
zfGtL_l@c$hnAKv#t_#2rI)~C>IgFv#mU<7-yzD)6ICM}l2jR9wcQTyN-!K*iCPqQ4
z!xHzc1-KP5bgyR#jyU48vSwkP2!mX#-O<4=c#tr1&U6qyp7%2X3j)bz`gG`3Lwd9r
zuzG6C{Sctar^ZIed;^5}(<czJ)x>ftP$(Q3QUMc%v`Zy%!Lfc+780XxjO|(%Q6NNd
zzR1*m*uG+zF&#7BvhSKhukQ6!2HF1hrmn7=TwD(Dh!|pO>LciOM%?2H2lvc*e30G4
z^D^S5AU&aAoC03TLkMm_mlAlHGCDT)15inTXo+j+!=*mIyuQCXn?lA|11bRo?m%9G
zH%<-^5WLRWy1Ba-Xcec!4{5Gcl)edw5`G*ED-v{o_J21daqD?t#sD-eYWlM81H%er
zG95lWzXW9!VM4)3rzPaD@MA%b5r=h6hJpx26f%|-8Vz|2BsIGmlO&xX+=0;UGPU4`
z*c;r*36uAuGIxH`nQ#+^=bO7Sx3BC!aZsG52+EGv<84x%-PwU(jw7m8>%K)uxRc(o
zg6RVghp4__SHX=#LdDU^>G9XpkMM98u23JC{sbs+2)_@I>n==0n)><EEg5izGsJt4
zDbfO{KNSdYaYrHP?oBN)iBHdnm5DGhWdl?K4G&C;WNW){^C>iwf^+1()=Y*B!vMr&
zW5bS`cdR;xN9wrzefa2XG8S1!P=$KIp#(^(0OX}cz7B}lfqdJ828Y7DA!*=zVZ5yY
z@D~maB^AiarN@PICP>}!MH0mpFsNuW`uRPAW@VVwv$?tH;C<SmQZMTpD!k2<iyVJu
z*eU3idFi92)3i^!DY;5AE(t33O==t<NwFi!IgzHug0luJQSPp;re<cBN;4)Z`>e`f
zvIL3(|NQLmL|+?oluW*c?$I%$nx0O+#!w*d?c1gN5sSW9J$-!=Tyb;rw?j|W$di+k
z34@-`?rv^67MtYUa=d%J#U7#o;|nUBFcA}^`sfm!d~I#*8JdPCL@bPukEsb0-os*{
zBPT7J$x$q5)Vx<eUci{%B7e@os4fqZ78rE}8#X;8Q~A22;`H&bMardN4`9b7L$8rs
zyT<Zv?xGSpSVUnaW?YtM-^-=^UhA0OOo{hhzLplmFyI&yMHfCAc|WgO>he!I;DXfD
ze~RDuasdqt6S<_(3uCqWaL1@nSJcE`^rJ-*M1<OUYt*GVW?eyA!BCUDG@IYBgwUeC
zGP97Bf)prSdKQ7-g={qNxo*kzpq4>FsX+U^HBt2pevkI#;T7GI5{`sUjGAoDT+liq
z6BDI>9Tf%F1>%U?=AGnd2PdcNw;uQcCyo7FFu4qaQb|MnS&kg9T}7FcwRRuqd)xL@
zWyF*U9*+m^-X_V~+ifL}97nt@OW21G{6ObQgkHNevkh8Rh&w?(b~RL=#MNIct;Z6O
zJ$RLP?uS#^v8gi3i9*ff8`<w?ww?Bt5+G$F&P0+_cJ1G#QSdd21wchso1?3wq-1JJ
zhY~8kdsj5m$oBMjp<zAHiC`UlzNX&0cDa|4iQ==>Vb+1c`tS<r%VS=i`UBy{ak7ZN
zPZb9tFSq5_e`smb;D$q*lex+sXtp9Zel?pv?8eAG!{JR9@eyWWv6}qk4<R?fCA{2=
zyTb^ox>h!9ML)J=A1b_<+%JRS$Wad;$^D>;kbR;vZ&h-1D%jW9WPRJ)fpZ6P1yWUU
z5Klo2O9|hZE*`SFYUktQ^NT_?97Hu{7GX*cZ)xOxdO5tpkZiaLWc9rFSa#iCigN~Z
zqvOy5_Sqm?2PoMQGbH{V21%To_8|uM5|w^Hoe&*oT82`-P3H9Ia{CFRfuqz09JI-p
zKqP@l8OxVMVuOT)3g-ElN~#Jbq5G~v{oO6$;eUxSetJk9B_>h@Hl0!rJqS}pk}boz
zgHxwv^t=rdqvOm?f|yhmN4`Vh+KDy7!AX3fzYj{3FE>iZW$Si4Njuo(yXnvYb0##l
z7cQKfV4R6aU`{Sh&*HYP@C-Lhyf&bYh|aOlMlZbkR=1Sy;OPkq1+OnHpkqC!6|E0;
z3EIBeg8cy1MMRq#SdrD$RWRF^uisT;c%mh#)WfNfH|2P5x3(#y+oEV`eNzto`l{E`
z&zXsdlaInlj51kY0FMYcq%BA@`YS^1$OLr8EiNvGIZ-}!dkc9GY~9J=$esfQ)NIUn
zql~Kev8`3r)1TPvtZH4MsNS`;|CR(YjKDqo2IRt2GhoRLG(G_yUPPNIqKWlBba!_C
z17QLfU<Ay89~m9Lu4eh7E0^C(&s9ITc03}#yr04y;StaNy#zR^EX90J>%o;mMVTB1
zj1DBqD23b|CqrjmNc@C675vty<YZykUk3U3vuDo^4c}0+Tz_Kmd|f~{cX(x7P>yd+
z5Vhsvl9#E*FD(4`H(P)L{J8%3u-m+pnK&3MDl*=t*rnfU3)~jmsC9_Hva+(eLHh>d
zpg@XALe@9}d-BVi-oW#yuD@I%MI3{BX(OED@3A2X+gK9I@3(3uIF;J8r0>IL4_p4X
zFM$6Mcs1H|xO^gUwPcpyUg01HSwia5+4%xV{6W|^xq_$Ip`v(eB`Il+V-}t+P~vj;
zH40(aUH;o{h0d|F!U%H#sM^4>6rxUI?tt7NqsQkw>Jwz^w)de{q*OQo8;QTcVXOE*
zXK52u;BLtr-lfL80W#9w%lIa(v}oRU65uYMsC=^q4gQ-N!Q8fDR@#$QQM&`lUFcPd
zaXi>rOHa6T`I(<=jU|N|L0G3ZhuiL9@Nmg()<CQt{p(_;`oDK93T!JBOnIOY*A?74
zshBKSY21v94u0_h)Faucz`F-(Son#_y6=)PUjaq7dv`~pE(0`V{JRe%zL46RZi^5S
z6znvA0;(K2xff7Rdr;73ssm*cFv=THo*sMH%a>4ylZ}PYD8_STdqd(|_wadiuXN^$
zn+HEX|H)_h;(i*dZ2I`|Ba|I1!JW|2(Q$O#1jw~8kWWcUDuDsP5{SKf7PrJYz-+?k
zI#A@OVIW-sSf4hNxxQOLnrJ{_r(yJ)e(;GYhjKM#0xkS^KnDTtO5uLk3&(lW{SW{X
zP-n^??t!uZc!6)V9xnU)&u{~c9tVGGc|MrQRG-p$`^dZ%WWPC%88Ov<et&aE+5A+c
z@h)I10l%)pjC&AyEJ3*;SWLQfs0M^t)_K6>-4DPIZ!5BPC7~DbiEcx+uNYa|F`L(C
zUE&bl-0GEDO8WP%DcRp)D~y)OL?-h=-0?e`QjC7pa<^7qnnT30RUkS+5(!yHlmBO{
z0$wvj((d{{0V_Pw8##U)5pU?7GO?IxO#a`2QAz`~aLK14-sYTf|G{6;^vQ>Hi~_UP
z$>CV2Fu1pT_kZ5l6KQ?1wkSyeb(xA^DPMz3SXJ(Z#0w^6kL(W_F1BDJ{KSSPiU{>P
zh`gfW*=vqK=tg^Q_pgvUw+d1^HzZ#Y8vyDSmzu!)LQF)slY2k@HC;{7z$2~RM->(`
z2WDJ|jxW7JVR54M?^$8_2R|*a1T8I1YG4N)Q>tFg>{j}8d#5bT-fTSa+Bn(1b-TYP
z1B}c&V`-_@1re)(5HH+X)y_8?Z$Y@zTt5r+RRBJP3RJj~k&)5Rq=5Md1wAR)qL-oH
zsi|6cY{w3O3?^pv1=qoaf8#9l?E+8&Vd71VTLhXVxSl#X-a0e$^UO8Sd5PfVMHamI
zMq;SUc)sz6xyN6`bmlwAxR}H;mwpBOJw<Bgsak4bkDw9^5I0lv3}67i^Or}8O#<zh
zVmhGM5MBy!aezRDdQfcaZL)ILwJG2;V4S4tk|-9Np)wEBVqn4hptK7-U~!|p?`!pm
z+M~}0`ovB*)3|mn0<{tm)b<69{_-s$rY*!LEo!!YxB5W1d64q2-%fo72dp1>iZRa(
z`F&fn(%~xS-&H0p1fdd0Wiaqv*FrxpEiJ*>#wWkkU{$6Nd+Q^G!!HD*!a0WmT47=O
zyve8}DC7W6W!vwH@Ku4UEk!OgJHU1l^?=<Bwuz36+=Rt*_3*fIjl1;=b(+|#3<jcZ
zBGjyI$=Zf}oc0<iD~d-Y?{3w@cDj>m>svCI0mZ@c@%+;oKz<E_r*Hi92_i!x6pA(x
zfGK1%^eb%i;StOMx&xU4JSkUZl8;sCuxoNmhpdu^M=x|ug&~!}T1;J*+_`GGDi^eI
z@f%-I-Ql`>cMr-D0kW}Fcz0lHpG<(x8~Ev+S>6t)EQ9|7(*v3`7TkwcM`v9J1`BsI
z<J^u$19wBZ6*o5SZJiw0#CyOXlLu(Ow^aT}5p4!l(&Pd2z$Q#a!$TR9{}MUgy@Kk3
z*zy5^C&ap@TokDJE%yMU_al-F!3$e^KG4k-%o0l!i=f~E3<z)U>|}H06B6=*Bo1bC
zYtWF)bj4pjo?H4&4;>vHm5}@U)m_*Q0t&8ae=>usDu+nG%*aS)^A!wjfRT!d3Ivf8
zGc%b`?dj(S#_4dQTZHVrcC9;4<1L6-yxx6_j3kbed#jOGT3YIVTO-eDZK^q6rwkVl
zkJXZ)a|3|h5!f&&GoPNCN>^fns=?d#Q<z$V;~^o%M1D?i|8U4SI0^}_-vtReA6tj;
zz^_W01!)k819vF)yE2_MOMD47LIML39$U(?sAF#*@#Hz(*SH7mj)!ORLb355vd6!E
zh?F)xxx4gzz-*)UK&FHH1rjBkjEsTUmsEFD>|UD<LsUodo&%vH8w0N}qW#4<!Ic{|
z29VT5MMn=nuqz%y3JQI33<l1i39K1louB*)mPKYAq<$a20_OlHQkX|DN~R2gdw|SW
zu7o{{0M5w@$X7kR?_dF-;2cUjYb|@y0ZU)M>o`~3A^DPO7v=!)2?&gKcf-8(had%q
zzOVyuL7uw+3qI|DM8Pz=N7g_Ruyr`k_clX@jUPd_85tR1M1(0OK}Sde1puH|0HPZi
zCJ9kJkQwrbxb(mi{x{@+KvVHpMddZfYJh}ifOgA}X!zvjy(Xe&1QJdFL4e?ZuQNb(
zu{<Zdl7=+41zw3pKzkGaLbREGpaJaLYrqM_5%KkbdTW7SXYIcq_X0_$dRWszcT&`3
zcTFxjxzg#)_o|;>hdZgSdi7a5O@P0#`xpmlkR-?mORK2#0xfDa@J<>-#;$j@cY-m{
z3TZ6<Mj}+F(jhY>MAq8cda?paEtMeY3M~l+vz7g@u>G)afncf;dXxO=F-UZ;_JCH<
zps97E#e8HO*bSJH04X_Oq!PIwzLv@CUEBg)*BC4W5D3nPdrtmh%3oc<jl=oYhF=O-
z5uz3-j}qBmDrHuEi;<A1gTaWPLbQds5wNU4oo9jMo|qs3i>nwuR$)tns{J$aNHKtn
zk`l6a#Xx+5tGtv_3~@K?As7TL*y_3C;p*)VOdSWuKU4OT&se;AQZSK2EZ_6B<b9{y
zH|vd8Lwl?OUZsuZmD-~^*RJc8T9-cV(wFV%a_$&WAG8+8kJqqLj$@EYcWdQFomEo`
zxHDESVnZ$=de+>RVs^`Ov!$!KS#R*4FzT>vy}_+Y40H^D$i$fY8mK4SVX81WegFuU
zUx0A|C1W>?Xn=DYzxUP(n6<A6*9WPlfkSKhBJMj40#pPpH{C0$Zf;e<oZa}lRHwtI
z9Iwg9VlW{udp>=-wyPQ<^ODAiU)<DEt!W0uh9IrPpOK$V|CC%Q8?pN$%i+Sy*8pS=
zlBGF!?i{rC1)AzA%B$4f-24|LKwv;|#sycQNrYgMwx}TycLOG6Ym*%f$btf5Qg{oX
z@tg~;htd2lfuCb}9z}PVCqRN*RZ%Fj0y<AD!-(0o?ry+!`>@%<X6?i-vp_on_*uZ4
zhm0Z2WME-uH~3&T30~=FfBbmC_6hJrun!IwT)nEjAqIhaHshTHF$rATO=zMZx1y}L
zvgbj?XY(Dt-e+g>f>48nnOUV?0gRpeRLeK{43FIvx^lBuFO>jeyKt&b+BP@3gS}b~
zBYe2|EfsGSu5D|3yMxbe=k0DR)<X8MkvtSQ%{&jg)3bDCAEXi$Q1%SGe{ekGzHP}!
z=Nt(Nan@y;BdRLy4oF^r*Xx3;yxx^nAoFD8<!fi6b)z{9>tKebkTZN;4r*;GgrjaQ
zJw#AfMO9V2Y0_@aaA9^*0BE+ILA`qKTBsk<Qc#W@KU{VJ!e^|=$j@Q`WC0}}lr=U`
zcLC);NXEHiMnxTQmdM3`Ajq<GcvF2D8LblkRzCI?Eo7sQ!@<Qajn!8B93P$cOI!Bj
zI(Z4|0xF#H4ug3l;3HJjx9awnH<oTgZVkeR;TeaA@o$-ABF5Bi?zP~8;n||_F*}vs
zc1oC@-VD6uGRGFq%JAuI9+IVN)jr-v_Vb*9s+fY%mo@w~#UX!S6)x)Oh`rue7|<nf
zI?TOv>I1urFJb|%Q1P$Awf^q-O&(H!&~;PfeBr|pxW6RgmA~R2#c#GdtuSi#f0K!t
zCs`q&5&qsv2F-QAA#z;{-Ltry45AHjQ#dKfp=H9tSHl4+G`F<GL`G5xd-BuM(=#*A
z2eTG{_MQXmV_{0xz;-uS$gX9Y9@aK`x;`Jmav%&Z*R;TQk64$3!D(tqDSw^C)DA>!
z!uDIM9!BQ@2bZJn(MQwX(7t`ol$t;01M5)bHO*&ree-&H?s6F7aLEY%LlA&mC4T?G
zbvBiC)SWamlgxTBBy;}4MB#_2U^u(LlEDZPNZ7%^lBuaF@54QKIG*_^A=d7GR#-w{
zLxD4XnA#1zD?DjRY^*Dc5x;)Np=rhe-r28@KbZIWx}fXV(zTk61#KrF&7p>87x+R|
z6%`EK?b{o0c#x^?Fwg~}+c5y;<v^dqjkDh#EVa)+yR4D7;C&NF!B?f~a*q}ga+|Gr
zp~61-%&D<amd(N%FoK0OQ_v+KF^yD1vq;uPZTM46Sah_2KAMY{A=kxH3hxWdCdaXi
zhK4DSO$A>GEUBajPNC+bYWQPuQ}9z=qax$v2x}jXF!|r!-`)by(sLtBump1Ds5?O4
ztEt}vrq$zuuCj7?pY)QyzrT~AQg>Bbn>wBYWCk^f%|Pc2M-axk$6%2|k(@W@-gWRx
zf3?78S{dAkXv*8VoGR!zy00cSvCi8-V~I|Yy9T4n`=-#S9x*Y!9!`tj26J`?wVSb6
za^jgQC-$qsa6W2G<oI9%(44rW<YG(imiipn#Bx1fDCOjilLJT$4mhvE?}v?Ht$P_3
zpORM&1B@W}>Ipmk^^J{3Ti*dMV?Ttn%>&f4auR?(Al_XbP))iCpfggIy)vVyXf>~}
zuJ=sM?2%i}m+s8W%q~|yfEA!#fb*K`q?&|1fTYDZRGrDn$|ityw&$lpBDo6%;v2jT
zE;eYxP6}^%A$z%E9r1e60#r&vt>gQ?sb0+V=h*CJT8o+;gypc*b@%lFCkHM75P3$S
z$f9`P77i2x-T_i9B`hI6-W@J~NGp_L!LSL9ibimr;HU`6$Y$$Sbq(JicNiY2<jq2|
zLVP+muVCHU6PVyf^s~rx3N#lqAbC^#tJ4WcR;(cSNMKaKx@W6LDqLB;T$r8`G;B=`
z*&0};8rNcr{g>ATqS^KLz}Pf5H}h=C_ow@m*bS(=AJ0Day<o*j%$xS?2W`S(N@k0A
zQwCF5(E%cq5^)A;29#ug84iTKhJ-o<2@w$xh(K0-53GTn*H%Atte7Ob<F;-J{T@^U
zy+Qb~dHLYX1%gCK<H2-#+lo;h`HrrxS{RiCOdLJBux>B6I^a9ZY=;&aF>KoFOQqc-
z*1RZYR@MZFoFQZoa9VJLb#t$Cg=R`{Px4Y|3$wFbf5llqhkmmh)m$*nG+vwGiQb3R
z?0?%%fN$F~*n8`~Tse!OV`W{0-`TH-dvLJPUGc*?VOC4tUA1fpf+P$i=?7m_?vmJ^
zIq<Of17**LJ(4v2g%_06w*W(1PCKn#w3KnW<gR(+^g+!VdR}>=aF{}z)M>Gsf?)iw
z0iEU>>YDnQEwYC=H8!oUXx&X3_h&Dkzi(~*u5og1uy7p4b`55vfdXmpo%U_h)~k8_
zcTGMct#w*kTVK6;1(*c@3?-D%077vx>}5wo6|Oc=7*bNLkSL+PejUk>LfQ@)Cd0^h
z{FCO`Zv60V0`y1%3IK?J*!cKsL`29q18gfe;+dmYVlFTq4C(FRQ&8q(zy}FAn5$8x
z`<?P9>{yDBM|r<h?flyWkd?(mM?1rNZAZ&*-nh|tct;HJ6T`#9d-w^QeDzb0N%kOw
z;Qej{nJsvMq1bPvIeL|t=((6*;eH2wkeJqU%DcYnsPLy5nlPQ|8{tnKPjgrnUl&k_
ztieiBgnmAC=KKdruv0Q#E8E&v8tTF30G&SvAeoX$(uLvDrAw5&);rxkM;kD04E6+;
zA!%}Q5)yKd3V^W49boOFtrUP6&{^Z)6CQ`>@!>YuLs&lYTMrr{6ouytWM_E+GQ$M(
zpAZ!t!;A*V;9bBcx9b+#?%PK-0woXWFJZfh21F}!xD5O-;t7F4F?5~um-li(1NF`W
zVmwGK0LiiiJR9Wtz^ow$eJf!Wx$?V^>xEth7Q+WPncP1WvvRoQDS52=5R3+8T995&
zcG7IS^A0dRiT7l(9D^zlQdl(&je-12J%ngOH|aat?C0<A`h`g+2~Z0j{!#ka4vV%v
zx-&HVJHdHaBBWP@O2~fd`41c0#mch*(6a!Gy@*2Yk(rs9t|H^D=g*&G#Kp^wj)_id
zs;NaDwIO@65S=1k3rQU~h48Ntcl4<n?Cc8a>hj;>T>@_NH#arG#BDG*<;hYHl+@Kl
z+8<SnLZsHSN5uc_`j7lJpCg@Xbwg}YdQZA|z?F@@f2f+@{O5KYTs&Y%{~dDEcUaJD
zEDg{H+_A{+G?8I<AM4D_474?apFPmAa_;>A|CvIjc8Jp0(O?(9c2iI2WoofRM$Dk6
zUMO1f*UoC{XkO<a3b#`~)HgmH6i4*Et3>d_0QvzQsNe0fUHE;Kl++mvAWSzafCNm`
z>_8DD7qr--L1|Z%&p34^U#;m&Z4g>GIop|CHB4<BC&e+E@>;TpPP%5r2S#Duo*|_^
zgbV)~kf@Laz?9P@08-FAR*~M1J!%oq%@qeGZzQ->pqgGbS`Q}6-L$_ay7(v)=NODn
zkQ4@kD86(oFz+vgomst{B-U1}=IxrUhe^M@5V<on5Q{XB^fiZ1R0Dnxp$8;PG&D5M
z2&t;7Ldy&T&$VW_XBr6xjROVx2}wz{AlNBS#@fD@@O9y*;;7q6$2zWz)sh6<r&I|e
zUAS_2oOavVsMvd9yKd)$Ry2|%@h!gvM8ySa*Ho|lQtM=SS=rlha`W@^uuSXk#xx6|
zhyYP4KrCCA@KzIejoq*JuOyuzYn`vj7pn&v=zLw?!x0F*j(~rMldO(`0Q1+3-Rcu{
z*45xC8Nr`*@wTiiB0_F5n5h>59@-~)@e!%ovbhBvD9?sUsB63N3=zx_-74ZzR_hg9
zffh=7cUyix_D#8EBzD7tfq~&mEN5fK<U}`dpuUg>3%X#A{3iF3+>hj&g1I7c^4Jir
z_r0_yvf6x*vWeBFXLse~x(`tPIxYLCDGP2bK!iaM$0d0`KL15sMmg<+9(qbf_$Nam
zheGm*W3CwH^^4DK_9yNiN7Tfv`m_&ZU`qc;%gL#!lj?MPh2}iqtH@q3{}K3ls!=VI
z=k1BAX7((Y@H^$Sr^ap(CuKhS5x*WWuAzmYYqqYs$@_Ne1kmF=4fD6(c`<3(k6UxQ
zI62ITJG|1rd2eHd=w}yh$4MIJ>)E<Xs|0~yWp`SxyL?}Rd4O)^h_OJ<Xf&Jll;d45
zuB!(z9ht-=LD=X1k9xM=75FGQb5-=Rp!GZ72K`Rgb%Wt9fGX!|eu#0PMIZ`6upRW}
zL(M}8-fRW1DWLEw0nHf@g>^7-5ZIt~Z~h&|Ev=lXq7dxu%H+ch1$mFUY^&YFE7F%3
zPL2jnYEX^-(+ZvZ<xOfz%F)Uv7Tqf#@Kd<*8{(3x$pA_{_L!K)kG2%SxSN6Q+uenN
z%H?_ROhRJ!Keu4Y7CMn<nn;{@?27<^<S_`mwPB_j&^O&H!l2*-Iw%POI;=H_3V`)7
zAh=MUxv)RGnD=1z>-(6O(F@&S4D{q$CkMv&Ex;ZKO^!*!7YnLwfL@a@5u#kbe7+H^
zv0LM1Xx#^qiy=p#-?-HTL8V!X!cy@FCPPtcuJFwbT@S=V;Mggw@-`qsJ`C6G+my?^
zHsASOED|I7(m|+B@uJ|v*>2|q+5L{5-0E=)I{KoWlM~hT^RX|?#>~JrH6_j{tYl|1
zp_rMNRvlpRGHdj5`03FFO?1kN`)}MghJyT@1F@TUSFV$i^zd#VM7nPSfE$ny)?#7#
zGbe4_1E;6m+Y=_KChAU&IqjX=k&)L`(knYij*4vk%fNCcRGji2Dl$Ts0OJgyMsiL!
zq+`Zie_lpWaG7FZy+vTXw=T>=z(7V8LppRF?V#;CG|{@sdtxnc5?O05uOf;_a(o~i
zC?J`bmEx^a(Fmd0E@0&<NHlYEZ%zi%{5kmzMGJP%Su1O+*D^Y`DqfrHZA8q{mwkMX
ztk(pUisFwQhWXKol_iiceMv+dB|OabLO<(`8#j1(c<AV+L0kZcU<YXZtz>YhDM**N
zt>=?rxf&BB%~4uJeSC@Um=K1*RjrABGVL)BAM(-se6Exe;-;WBuHWk~1RDn7RAdbD
z(#!)KvqYpQXb%_GO@#E|m2p`v<I5l{E3d-sxF5goM9Ul+S<5+&b|N7L&KFwYi>qh{
zU=6^>*h2~z^azJApAlk30OJ73zXF+b&+j}k@D7rM!W0+1rG#{t;u(N^W*vH5p|u%U
zXY}{t`&M`oO^9PerV3IVz<R>q1;NC0vwI))KYnp?KUaM{N(05o+J4Sll~V4cs^ZH{
zVZ7C>+U#O+1JWbuo4ExKV$@VWE*Jf<@0%nJH%Oyp@ZbN5JY$3Im`0k5fZc>B_-05}
z5)lzyKJV9NN{EW-h>D6Tb6GJk3_T4~OkD!?Y|Q#k?KuAFJ~=!bI}8mX+uxr6U6Tm3
z73fzr^z@b-DmR_wzahyS_-RlMn1X+{{-;Auv!pl%FYGHC{iCarH1FsY^|fAJqLYt#
z5_0AA0Vs|Ip8S#vX@)mcIZ@s=e=&=2hnwWb(4xZ~);TmSL(BdyrfJ^(ObvEgL*v|z
zaN0W<zyyc*3T)QK#>Q6v7PN%`^a?J;17fe`m23C`K<tE43+Z5v_6F;ZK$;G>*yjzg
z292+xaQ;~@1+z?utTGu{SK)Im${-#ook>w}?fFsW+lRK00fEVXKbKeG^sokVL4JN9
zQT@lmz7j}uz~n0+aS8yIfiyiv-+36c&D-EK%ciKhKg~EmE@w3@klHHm5}VF@DP>}>
z(+XJ|fYe}}6c+KpclgX_#&FAr1O@GadOQG}Y{h*su}`2ehL(Jj4wc|K#lcwAb5iBA
z-%HVYf>YNwDQH?6BDH<>NvzDd6Lo|h`ykk3PT-X4lRg6%3QPkZ@Vy|?dIT9!g1|Q)
zKI{PL<h|9e3GoQj4QCFdJcdfHnJ5q(cP8E`&1Zzdja%f9kAWRr0(e>R$!_fY4hYv(
zFv~zag03wkCv4cd4#+hO9V254_&8g9ammdKSuMCV)z9>4Scbqw|Dy#`B3DRCL)yx6
zF^D_3agrPYEzoDfEreEWDG?DS#ab|<wxqPwwrZZC`UgUE9YNO4cJX05q!;1p?!qpE
zBQ4+3%LV4I4%D(&1Xo2_U(EdWvfbZaAqxj^%=bP2|LX3&<FW3;xA99F+Af94iZ<b{
z?48+CDWlA$krmmyQbr-E6f(0zWrd7NMQ$rAnT4!EA|vBD-WT2X{e51~>-GHp_`QCA
z{QPkbuIn@2<9wgzaUAD)!X)#2kcikHIcgl@lyOv1D5$hfprR#Afi@xw^Wj;8afwVZ
zsy^X9)xjld_OS{^0nT*-z-Fhj>gu(z!TPzdbyz4cJCi6-!UdAxL$gja$11NllgVr6
z%YVVGy%!uzq7<XPQ%_IGD=36Yml_^j+m)I(5Q5qSs6{}^yhf9Q*is}^sS~oVmG(r?
zrSwSH-N$&*1;+<xo_ZI!w{~!4e_bWMKnja%K^t6XW+j9%QF(8AZIo41u%gIO6As|`
z*xD|9n>7&ZPhMpajH#b;jG7`U#iVXO0LQlb2nYBT`@-LPXlrY1Yz)AX%WeEK(Ehp*
zg~Nv{pv(INWw*3Ck?O;!{`mHnQ3d;0Rk%g|sHTz{$>d%f2gD(I@K%JVWw>O%vlC=v
zn4Hxd5*eULx0d&Xn5L?&jYZgxcn=W&i$=4<s;UnWQNf#+wU2<X@VMIAe4CK?SFhw_
zs@Jiwz|D)x{`mbDNfJP=L^xZ~=&=RJ#;B~QAPldA!*g8X;6dlkKoz`)e!npI6H9b2
z1=jlR-MavGaT2@ONIX=X)@T@=ooSneTZgY(J`Wwk`2Z8JCJ;MWw3%O0Qb%#aN5cTQ
zNA>EPW!~TXqXl@99KSH1=fe#ORHPn7Sor&>6Nr<W`4Q&B73X;dIMP0*rnfd=UH`N+
z)broaU34uqz0svX&z>~y9k66(WrZDvPkIe}Hnso~7`RA#WyB4+g54~Ea7M_kAzTI?
z8)lKv!w85l`D5UTEI+V`*9d8*>LQ=16hSLlxUF0(R<6_o{RsQO45#GgNz$a696V+3
z-`@@jTE1)<`2v1f{!mhDb^wKg4sESL+zGb7cj7po=HrVJ522lRwXw-)9Y(mwymsx>
zSJA>m$=B$;iNkFW`6P$|L8DHRr-g@&{2Yg`gxP_Zw37rJ+qZlB`|E=n2U@io+W5+_
zwigDrur3gNPh$|ksR&h-J7cFK8G-7E0E10z)ekJdex+5Ahjqa@=DVP#u}ISBnChau
z;{pvb$*PJH4^($8R+u~0+|r5+4Q#(TBu@wHoHqHTK6>dJ9^_iydFV-aqLVJ{-TRVQ
z;I|OM^smnxkm+li`ISww00N<>X+YcPH23M`-#UW@ztQ6l5Z?^8A@H4F5UE}B^hDy4
z6#y(JFQWZ#bshvKLYlUoJlO~k{lLM4L_(;pZi6GW$t>$_Vc;Is4%l7hz~t22q|2%`
zFhci*Opbwj3E7nlUmd6~9n6L_EaQ)kcM&eb`-97b&%RwTE&cZGTSCiR$aKTRhs%!A
zWEWsioB5$zr3Uh)K93$4Ax*A<3=Lq4u<6MLL;Qd#gRvbAP*>b6D0oD$g=D}WHb(s;
zcknZaPH-}~0*rYvs{!&h;n}kfKj|OdYyZA}{P$p?J7Nk%RiG}B_X1w6h5&Ojsc|Og
z$~*(Qz$Ci~Kq!j843LQ!SZ0^{C}f$P<6X_^7<mg1gmZHAS(+y-+Qu#=uN;%JmUnVE
z@zX5}Kq-b>aQpU>%F4m&7%%T0$<jT0_Mi%M0dhF7S`Cuc%mBO(v9UX$b_a<HaKBS0
zga0Rw!4(zpKNyTL6C@$v2n>pdD8xaEb=mU!{jW)Ws{wH#<sd^WaPSk_uH4d)!~zfj
zZU*Q+3@TUR>rzYq{#zIP2Ba8hWUf4|GV*;AjLuKp#aNqq2rY3;_gARQ3hQ5dxoGZ5
z{-HFyz+K<ir>S|btjp<(&$i~@s|236Um;89S($(iw%0r-l=JObnPMlVKP7MWl^3D`
zB)K>M3_-FK$EtVu5ExzwyKf2ZrjX+irLw4nu8B!v^mZr$Mw=}zUy`fiuukebkdR1u
zaQvj+OM#oyKVhuJRC;a<4Y=0+%Wo8)WzFYo4-IJ(C_u<BkXw?zMOauvk$sidOsw0f
zochde-{`3U`@%JWurxvw<+V{<Ib*svq?5OnUx0NOFdvd|wL%(6=H&j%koqMiO~S*?
z)ykQj9&|_J+taT+Z{12dK@`L|hR1rUUmtCgX(N!>%@vgh(R25Wrd4-YyTWWICAA`X
zGTaph-4MX5BcURyH5mYswMPP&3sCoJ@uWj#6SD)g6mex`r8`z3Bu?mGu!>VEA=@3k
zz~pP_*1uLpq`7areChfz6Sen;y}kKl_U@@yzx{V&YiKctHHN^rXoCBzSH-~#nV%Oc
zDLee^z`k+-flZKSXF!kes|%-w%Tz^Mgn+Xusf|$nH?CvSPcK|O)9_*4g-lNHO~yif
ze6#;h9OaI(@iobf95OD?1E)hJlC>j_dhY`O4XN<K0|&s3gn?YRbI0Zlvlc8ID_GP*
zFqg4Tvr|YE`8Gq%zpV>IJOB8H<aEd3DPg!DRfg%dJyqN?GOg(W(c2MUn74)|DMfrd
zS=QoMx`a+&5f>!gOT9RqdH^EMz<8pN`|%SeKElew8-Ra-T4>iNe|B?b_IMkIct2;V
z{ApHcn^u&5jU+-F)o<Eqqv`dtqP0@Cr>=IC*<qS8#jK@EWn&A4k}fC+mpXyf29_Ke
z&_|ErTzRX9WV%530&W5XmRnYV^~XvbTaYA*kCB`!xFm}wNpK174NfFrSe@fWwyln%
zQws5q9s!9b;4G*Sp4$%|OyIOcXXQSqov}a$v45L@X>coCfBw6_@ww&#`(%xeh4+9d
zKkv5Nsy~xdZg}b3xpSu$I_QB_eIe_5S?+W~+LedpvEQ>A6(8l7-Zy@9g`l++RNNWJ
zp40of)kXcWwW(lF!0rGCWMyXFUo%!2QHO-b_@oO~R!?hsA2aV9rS_M<4d+cBbg(Ew
ztUwInJVC-=AfOG3BwES7+3KBnc1wNjNoE$7de^<u{ad$b#rnYdVzHXTF((vF;moKY
zH3yihlved<`LA*vwt86&ysSxNz*ymht1nMwH-+xXfBW;-3J?u5W6HlWPV5Yj{&gP#
zraM?GacSO<-!rt$1ypJAcaiuz*A;F!ejrSmd|>42;{dsPxg^D16ZvPLJ1>f{1OLt}
zKuYr}s#UOgO|Mao^fZ$6j5xOJ*;8|FH#%1c=wv-^U1gW&iYMgRjpS^@zj?Q9iwEoA
zwYtCI1riJQTS8=?(|P+K@QT98K0fIp=YfdezA0U8ZKwL|<65x&2>3BC7h2S}@yw37
z&x*pEha!n=Pxe*sh)_hob1WPYG;9KKL-5%R16iQZK&m0BwG|Gw?44Im2dbREMa0id
z&<>~nQY3<_`k5>KLy>?}JxZOR(IBqKh2XKS)6nWQ=)3VYGs_Vec8Y~VDxa`0IoenP
z0hQ5y<ppIFJIYZ57L|=`9Y5Kt4=_i&bmEBnWRUj?PN|Q^j~kg&4UWygo%-6@iA1#+
zO*IMWBGDJVZd(o09broC^7IOoD$RjI!;ntn>L3CMM_(;_QmmghKNv0?&oJ3y;H&*a
zTNB3e@X<UVBLsXa;GlL>-t-Ew^HHh%63WYybNn_1GmX&ft{%21Vk)?|QDt4ZFfgNU
zJ3BkEKM_mLz}z8iEwl98g>fZ_yiTgARX?(Vmtd0K_yRe%B>oiqps%SZV)Ee;K5y4o
zOZ<nryx!Obg8t>M`Heuw0#0bI>G0^`4zL-pkzybNV9Wrq|3*a8^c}G)lAyuwRUj3|
zrvx-uUn+M)$;sXW2M(m>|I(g*ea+X`*C^caD@v%a1Oj+)wyU9^QsHvl9B(k2iqhVY
znq6{pD<}vqNDOx#5T};De@|$v(Y8{s<}Tnu9sTPbiE5M~yi$5(Q0P{EU#G*7t$p@K
zfICD>fC;|9Nqd~Pw(!gwUp%nD$!Cg+<PbCX;dRMFTX*x}_^<+ZeMRe?8cWQcqk##9
z(!CCDEt&boBVaMvfWx$Y9GY>w`ZCg>um$D_1E2b%1Hzm0@O7~xVE-I!Y-}()$g_lu
z|6W+wRDk=`YuB<*0OnxNprUClbf1OJ<`--)k)R`n*b=P!;l2s7|I#or;B_E|0ziTv
z=ooNOBvv`1f*pwAuN4pb3m)bFNdEwy{xmoe#G8jyRY}q_4s*p$biTHBUllEhqXa%x
zvTVaA*)%pJMuAXi;UDm`0;PtKwCO@-D_B#UnV=ZSXW=5=LnVS4_9!A~`1Z}r_4@qK
z5f2^0<GwL2CxJJ!SSARK@<AajU%nh05-K_(L)678m$|=t%0nB`bljDP$7|dYGJN+*
zUgdt{axZ|BV9g#egV9KDLlAjUl*GxCujlC*;E|sS-QAC8Z-bvME-C4ZWJgxri+k~-
zto#;bw6iSR1~|195i2&N=#ct4w%<PffRwZ)%RH2(zx|9LkxkqWlm#4w$4FE22RAeF
z;UMyxIyyRtMTBc8D=Rw-pWpoLV^Xv9H)3MAncdbo;BK&geuE8?0(NTK)(iSZ=^PSg
zK*^^3)|2M!S6Ud#&Bteo!9uAUL^Fs2;Q%;*>_gD4Ag#&8Hv3<~=3ZoNUR3mY_-sF$
z+Ggw!j1#)F9Vb-q-Mc;|IBZ@(HmM!YdPQ^pexEeA&wl52KR^ov#Ye!xEM8tWTs}Rr
zrl#Ifv3U0{WtFjTr;gME{mjm8Oo4oxUTfJM9<4i(a%5J07%$DQr!X8wWnn?L8Ezg7
zpV^@Y$Gu4@34|g2v$EV5L5lWrIYYqvcdt#>sLcWeMMcj>Ga0CnbdGF3Rs7~nVp5Wc
z_Rm!(7RRGtb#R#m)@J>hr7O_Rv)k;u-EUY`ZTAI6xpH~J6O{%BUuXL-k_aug!|0$<
zeN|TLI_GlXhR4uGU`R7rnkvs>?hs$-0NkfvEp>D(50z12BB8680TkmReLM|F-?nYr
z0stEc7B*PBTwJU#ihN-#;vTOhKxI!L^iArEn4Rjz8vO7(M}>c96_;+Qum(%ZlJ-Vf
z#G{l9?)1iHm>`iOgQ^;unn>{Hjj0|-FZa0CRp2$r0zsAm%EW0qTCCJTU5__s9Ff4I
z<Fqkikwb9VD2{pHFV5lor0zRN$+YVpg;he3<|t;|V$`y^nfG8B4N(ty*=qx_2c$?~
zg-~Xow;`*+^!=S3TZ2rXnM)DNOGi1PQeYmRnCuyAzl2gELPG)c6(J<Rl1_q8kBHL&
zn-cneaHo5A?+%AK&A-A`v9j;SS>HO|$wk(_Uo7mko^+tG8GigkFVrTW^Z=q@Xj~sZ
ze!SQzE-udS{CUEJO$M5~KSAynq_6mU)~gc;F#-GFQ}C|b;XLwKgF)!*>N;;|I0I`p
z1zHlK7juy&h@U2Zea}P4Zg6OwdF9dLFH_?d=5M`il<B+gc~ZEgNOn05tsf6UY%ViY
zM30qf`CLa^JEZj7fg?vqFBPN9MNX;+jPXveI6qs2@^>Tu4DMb+tl+Jh_4!j12p<Pp
z^L2<VeeUq%Ol9s3*e;zcvj+(MVX}cBdaYP*4TAIq(lqq_y+z{HmpBBgc*q)Ob6pZs
zdZd+bQloEnFByitv8~!Rwn$zh1d*SMYEbE&hY!DR%M8*JXaG6K$i>wr@fz?D>SS3c
zsex&`4di&Neqnl}6p%Jl1(T4L)~m@Na!hgdsmoCXl!&v55|lR4osU<3B{dp08Y4i;
z&^ENVpOgevQ8_02V22+^MOnlkC`lZ2zp@Lbk%%~a$dwde01=Lqju&;U%-VON5{@bs
zpSt+Ap;__v*{40Ld!y9R0vbPUKSk64^-Djd@}(N4PdqRR&jCmBf_;_w;Y6|6u$`<_
zBbVrPD+cTWm`48jiru1uUA-+S&{yfN@)FS+oN=U<W}Z+ge#~n!_ZbhC_H^fV>gL*4
zeUu2@HSGZsVb@lHWnfqu<w=wLW_ZRe8V9gd86hi+46MZ0JN0sfcJ6$<zif2qNME)5
z^vef#zF9QuU!s><(heg`iBuBeFCv~IVq<Uu=UXiKaRREXRZ2wk(rW|YX>xy*yTj|q
zv?4xVzNoE54eTD<Xh5)B*{S8yA#zq9>pEh2zWDlA%bWVv<s7UUi=w$wl#CSvwhC*u
zD_#3y4ZJ=l-zd**=?h+KZW{V4mX6vAS>X#9i%witS>tdgGObb8{Z+4&arTXpMciK*
z?iOw*qi9Sm<`=pgRiZ~$sH;V+#u_RuD5rSbQ}j?-v(8^Lr6_`>o&V41sCcNv<mety
z3-sQ1;yDZRcZ-UY0^P_X+iCZtR2^d&GTxb)p=k9{srZ3QxRLYWSPjUqTIY~$xf|R?
z=$rPda$CLnll12q`I@2LkM7@msyq+GP_9tuIQ?SlwO$IOJ^K)4!t+5?>aqD8`}guP
zCE%@faOBtv1%Y?RY$@S$7NaWqX8HI7vI5Z)OS1$1RIKa^z*@HM#nXBV1rYmq8!N}~
z3A}8%uhT2A_XrJGr5NumUpho&rEHcjtl|b+t;_iJ|6byDV27X<?_rGPu9LMPer%j{
z;mkBEVZ7HYN+o6;c()~$(?9K0%hGVrUa7tqo+!?>n9MO{L!wEPDNGdqo|WU0V{{Ic
zOe1+IBFXwM{b>O>rRr4(OFK!&U<p|=6?RPB)hkS8=_LEu%*0=F<bzz<ndQ<O-b5~O
zP^y+XSfvG`MCNoeH!bsv{{|%>{rF?e(PAsA)tk2_ThsGeFkh>gzmZ!<UZb{K*<s~X
z3yHhVzBm~4;y|*uB3rG$;qg#a8mD%Jo{rO~JM~oe$t!LV<#r#sm;y(tj>ejQrO~Q;
zJ;~Kg2X2`TT68X+dtcRs4q*@b3KS~NE|ST9cFa9;p)|Qw#)1V>6>odhZv^;yvrrS2
z6mvr_?F6EuyMP=L6n=kP>90F1e22Zl^j{uUwC2Cd!<4k0)$@M;?0)hAk3q(EE0zPF
zsnaj-VsEW({1$sS-g{)(a$(+D+NSnkQSo?pa-DAy{fl?Yq(tX-0pFg<zH7iJvyCb_
zRckg3hyWWk@bj)eEp~8`<f^&Pm=5jrAdV7&WFbWtjr)0Wjm*@IxI~rZN{V|P=zjmw
zT#g5wAJCN7O%%9;GjN7KUC5j~Co&`ojQf}gm8x~-nM0UugI2qr?}s1)#Zo9KubL>c
z?__77QUxYfh+?=<D?LgIOO~}yxVC8WDq-`U*7i!7*?SW<gJFJFG(h5zq+;rNpg_o-
zbKS)?&YD?3Kh*SRFV%moyO0xS0x*v`FmJG+N#*Bd_r&aluKsVLWk;4SH@zxOmX_Sn
zVcwz$3ETIE=|219<*x*MrJp^<+^$h~_kO2^O&fPG=i6ljkf&rYr7Zx??SJ#;+CYfM
zCyZr=HXVpn+99}xs^xOUFhSfr(DbRVH_rz?&-oWCZlcHnvZp?P=)W{*o2$^4@#TNO
z0@GiW!IFBijb=O_gp1kFN1o!IyN4zHN0i90sa9eoVZq<`cR}+tQX1C2ZN;d@lO{D2
zTOWq`Y3m=(S~gAcx1oi?fsO!_8;K1Fwh-{kA|AQ3r&&M*nsz}I#bhh3CzqVmpVU}-
zu+rECN`?Tqekgqg-RvYDqM8IS)U)){ONZzbcRU784z4ERNtMWbTJh3_e3tYt^ov(@
zbN(BI{d_K9;j9EOtastuidMqeJv?k3@i+a9G6+dp%+Z=;;%sI|?LD>o4ciKH5pnW{
z3^M9emFe0^-jCBqO)^bGF4E5pZm#CCA;9pW%@<z-;?y=0bEv$S!CU0q${lR_#~l0N
zGf3HT1Dr%)*826UT8}SZKC(p1+Lig6y!ia6zi-+zgswXhI19P}HJCvmYt;9IhzFJ~
zJ>4iGXB>i<h)3HvfluOXOISyNf<HfrB`H<8h(ZN|O%cDMN0HT838L~T`&=ba+wpE~
zvWB9rPigdar<mh4cADBsn=6caPVLU@nj+uaIF7Z5${Q*1V~EiVZ0j_@M@@q~RO&=z
zCKOt}%IhpVNMIkX=gxJ)BnCcs(0g@9OxJqfBnywRoqG$LE~GyfqV|?8B2XHost-?s
zVtCX*&<O4Tu7Og}KppsyMu7YPtuaznYvK++oW!wij!UCE(~#R+<3$H-J?tG9j)wsN
zo}U-<R+gttC@|yQ3|_&>i8T-f0enl~yNwDmD?LC;BS47m^3ZGm9fDp8{TI`U8gyRz
z)1qri)g+z+&@)pY(5D<7r(fT8@BmD*%l5o}o%98BN|7Yr`D-mFmAY$PnAOq&<@wdc
zZAtF}ha`nSk=`W;AA1__hldOBKp^$gS}bEh?3X*%eUH4wI@Y{7RmCP8$Z;2zEGkLw
z@2^>5wEm9^x8Mr!s1?rv^&PlpFx{sW6%_@ksz_*g%Q{xpZ9I_Q!^O}%l`u$|q6^Lc
zyO$VUMh%ltAKI?|>!&J_X;>kBrH(J@0zAKRLmfmvc{=*~;oh6KY=I-CeOdcNC7r|k
zd7RZ{?oo&nOX(cqT!eIR)-Ja?08#WV!Tv%FuSjIB;2f9(Gcttm;(SM8)DPN;e$`G|
zRW$amL8*eY1Gza<aiJDWfhg%l)TK@x;Cqe!S_h~A8@d^zh?UaMHw4zwBlvX`GtNFG
z#kE9W(B5KSGWuu5FT&3AelhVi6iS8fwTVH|>JR+}>XI9vYgH^h^AxTzq6Oa)q*A@R
zxxQoiqOqFq>r_rbW~$a9N&ZJ)3&xt?9{|gLAe)iKyi(-Y{&JdS+8Wf0sF8rfDVoW7
zL}W<3aUu?ffLRadI>5Fx1d#pHCyY(z<wYY2r?gJlvqSmN`KpRBNPa1;FN-lf<O&wt
z-0|kAoXB87LXgY&r17Ewrjd|tItTkjlJkcz+mDwlPGJWzP7wtQB8~$%m)@BX9zi?W
zGo%(s9qv%OQn0nCTT?SQ(r6K_kmRR*HCa_CEu!fD9Pf|BG%1}8CV;_I`vNq(tP@pk
zC@Q{sm6>2{pI&gSHF7)a^=EU3q<M@9-3*x54_v8*GfvV)c{&?>D1()?ADG^phRbSQ
zKK$D%{^zZ70iG85zfbedpC;x{u5k*lQ0Z@!g=1Z-sQ760yqL)@Bx`|zuI^)51eU$4
z5s!}ixLIYVZoeq6ID~0L{4NeXc0#gHJVa_TC|oF6myBZIuk3^NvRw*}w0G2ef`83H
zbCe8@tg4vn`d9S#Y|yh3j;QTn{sv*i{6>{j)w%4B$m#{VC6WsC3l?a5AJdOh4`U){
zW`ah<H}&Vb&M6|)USpBWJ?lPMyAeyl*bKKP!tv{R!FQXAe)B7gt{ejULq8<Rn<7B{
zJnw9nsxBXbMszd_u{vs@Ro(-u=Igs`ESRwPh>P}o_0sH+p%^4D%%7>+_9vpCB-Cf@
z8B*y_H?HQ<(7>!iHi|3}glp%Yo<OdFhPrwNbPdQB^C>~31PKx;b;shkW@u5Y)1ub&
zg{}fyddLJ9$vWs1Tfod9)iCMt<1jGdgdof9Pla_u%QWA8XU=>PZqQAKk%4fa0x*tU
zlQ{sStAbx|vZ!TV(^nG;;2+7U;~TmYNgHXhQ$bAYpFiIPMEyr_;gpiv9jD(?zdoJ?
z_L_ZSr{IgBE7(=nwreIkeR=Nv{7i>X&`HaToD4mYQ139K-CNg(+I^x`NNv}_xdmZ{
zZwU^=w4{~!wJ?+0{@5wTQR67%kT5rjOxuOb9HG;5S+r9g*>p2t=XpNhy|3pXN5N+w
zH2vg;wK`W?T`2`ejE&j;UP>mY{bdt{i-Uuoha{)HZUNz9O7*#MgBU@^I*#@rQ<)p{
zdwqO0EFORrrE_SGf{7779s$*I0ZCr17*e=!8Vc%xTeQHeuoJJp`@F9=J1W)Vw3^)Q
z%uhD&oVcEL=8WL_4DrYL#d&$O6(6y1MrG#+O+-P<_I<_mwQJ_ryq_c6->9uO4FP06
zkR<@#M3(|(uOKpkAoJz*eFl~gn1%i4T;gPK$+q?{eLc1!_Kk1Nj(EYs!KaOA-OO{F
zljYd#ciL!rXm(z`TSx32gsE`<ua|XcgS%@v_7i2)0bq?ou+;~nveegk*2P-Y!@i9&
zA219pX|;%Ar2_Oo_CE>(pEEQJx7Sx)puV<C3z90ZIAOuTWB8lbZTjiEc?<KuK3!mN
z{H4ESCmH3U)#KmHE`h?;e0+kN2W|w$4=cX+Qzl=O^hVg$QLmcTZf=|`b9DiDKWZTC
ziud{vK&y0XUOD!rG&(vO$v`0Js`quJ%SD>N;ipy~GStfZm?VPpNt3$t_}!o7aamzB
z>m;Y>^6Z+vJwPL%6+q>oMcJ@oaZu@I9)Vy!0Bbn43N|c~iNE*oC@(+6nFm;6lXP3S
z^E&Ro-ocZvps-?@@LX{5ed7P2CMx1m8)b0#qikXRt7zGIuW5WK*TQi`_jXWEp|F@D
zFd`8YKB*j321d4P2Ke8I+s<)<{vMY7XQV~3;-{uuTy3TK`3aSTvj`OcGk@oy91?sb
z2$ER>$pJuiAlQ@c<}E#X`%4Gj>cX7Q9q;2y;5@=hePp7IrhurWxEs2(Tn;8CCX7Gy
zIjez40UmP_XLm{EElN(}+f~|f|15mtlNwHpw3r<>0p2_MtWk3STj4}>iIV}w0Qj34
zh_W6~GFk4p%M$zIkfQF7r6Ey$^&V%u(yKl*1`476(_Dx-v>NY$P~^rwY#O$KdQ<|S
zqC-VC9yuEl$Y_&e$yYe$TxE~hA0kDe<%U#oc;l~I+-4?>sA?xFM)zFrV;ZSYgjAVj
z?s)WQ!w2pA>s1wzh}3p58*86xaw=5%ljlplD>6BTi)0WPd+8ed*@&ObNe<20K!VSK
z43Iq@0mOa^oIEuU;2Jy(mxH#!kw-=thOmCrlmu1?XP48Yh!k7AKjH6@ElhatDI^7u
zTrdYWt(pup_C%_5t3@_gy+aW)Eg*Rmd}|W#-ShTl;fT=?%nBCQlWqFOW9{ZGE<ZYV
z!cyA#TDNAPiYW3!WJN^kKVSSiR8(-wmRKO$li4JF9kvQYGV*${tsy`7`0+&AwiOX-
z0W8?eJck38+2!25i>7UlypFx31+`P&tM8ovz=s4N5Uok98Ile|k*zb(gn*^<EdeK7
z0J;U{AEHK=ef&u>l!#78N<r=wy>)Z;r6^#X>o6}EFzCaE6;?}+fZv$~g@(PkW9QCq
zsCb1k5tuu8E(+6*vgZ=SDTD9=CRgRi%-MO0I{?VE{tvNb0L)#+uNbQ##qPFrc9m92
z(b(5Hd~s--`i|wAA#t4iYpMJ~wagqb*%}u<aYBG{tUONaY7~DwEy5SmtI;GL`u8pe
z49Ms(b35BzmN$lLsQd!@Nvp*=Px5j=Lg|`tiuP%@I1@TIL`3@j@g^a`X_Sbg?5n2t
z?r8|KtxTMwK`G-qvZ|KscyqdeOGu&IODyMJ(lWc=c@#{x=5HQ5c&p%YUjQ)!d+Q>}
zo#7wC>=2{kk*0syXgMv|9}fl4)nt+Pova+R<iKeL2^=9vL8BPbe!xg2UW@a-7ulE$
z`DxI8_lw4#Du6z0UbnLQfs3Ow&3`I#R&A0e85=Cj%&x<w+prkz&zy;lBVbq;YFqv$
zF^P5Q7P2%!4`~}m5Ag>uk0((I%Pyq<nQV=q$|AIlL~uX~(ndy}LhUW0sDt;fcn$#u
zsMHfs2npo&M1#SC=s10m;qr(?ae9ONYXye0^H74vtce>{^6JZr;Fd_GF7(KqlyCu~
ztauK@4Jwe4ta{{&u;Rah;CYO-n0?-3TUpe7V&^3NytlA=B63-N^mdZ&gz6kvwMfRy
zLoJxN*jS$uyb8!AS479iZ~gxC9cR<C^EbxMzSNScC<-0*(p^5k^xP-LH=yRtB$^^`
zla+OcFAHHwANq57L;2ec=~5wBSu~6M*7Ui`EK{>xI8X`^Ui69!+tL&HjcEM<5{?{f
zp>Ipp5_FxNLI5LX{z-_qL7@Klh1f%q>er9EYU<?FY~ksRiaMZEL6guPEo6DY|2Q(T
z9d!vRNjk7abB}e7w#7yvuNIDWRqIbHhvu9`ihJj83d!m2$#DhxJkZR^%xsHXv?D0|
z(EU2<K~i>A|5#BTz0~~s$WH7mLYcUWT>v^XACCiw5mv2XLkmr8o;?tQVBW4D+fW`+
z`$?YouZ4%YBWUi&)NQybAd2;$f*WM!1Gm$XlMfOIr+E%SoiyZ`-eT0WnO0N9w5iPj
zO3?&RYO>ZQ2dl>(V?&Qw)MOiJo`Hju#|HinMbw<Z+mVFSbJ4eR+>`Y$vj+9D=H(an
z!kaw}neYZ0*U%mS-xN7i@!h1WRDH|+TetoK&aiaNR%96dZEW^3$zf=Vv#Lk;oSQo2
zu3FZFM=kg*j?C5TY!;vbUqLYdX>r=N>-V}GSGydJ{Y{bTrlG_|M}#lEl-j)c5mZSC
z@$iEQzA&{KH+br}=~ebxqw@9R%|Y=a-6MZ#&K)^Ng!dHbq}0Qvw;f)Tl2pOs6cepm
zGKcMqCu1){rju~(>*ab)j`o8ZuLbS$ylqzR@;W+i4mD~POd&&6+nCe%`~d$*TT>F}
z<P&$AGca}kQ~eq_wrZ{RmDhDmHI_QfpC+=Kik9riUzTqu^i?yLUTO})Z7*aPdU>Ew
zxx)DU(-TUH)ZRDqkFaZj%dKPVcGMlKmrj%#t_c({yp(ZSmbMm$49;amc3S(uQa7Fb
zqRRuH>+PJpeq4Bg%U?-5_Ht2UWdA%VJDtpS4TA-R(z_41_sl6>DMOYE$0OA_+6z~w
zEgQ<up|QY}DK<Kpb-g-0DM`6y?7TdMyVtMryPg%kmdt!4_rmd0k8qVuNe}hZJlSYt
zxRH^u{p#)4+&pb6*)@fI3QK8iP==gw?Q`iqC1cf3C$;7^*0Xr`P17r;#;UC*gXdwv
znE(pc&9z>y6Sq48&|oqWdURjaTzV6^UC~KgInWS-s|UR5G2UTEXssDM<m5b7EBn57
zdw=w9)25ZW8)@d5&SUN}(VAVqjw+cnm68ikHe99sHYVEXv-F|QmbBk3pnT?EXJQxn
zhNE3$ZWU~#ko$bOsJ`yz<fU~wv;c+nR)O6tlNWu7JeeZ5aU-4`wteu{n)%;&kl&PB
zWUmvar(u?tJ?9=luh@SM!De6}!zISSTsKzal{`i3^(aT#_+rtP(3EFiH;B*B<`H*q
zy>k2YXMr>CglR|*BO?~WJ}WWan{{;IJab7e476WK@Fv+;cqKQ_-H*KT9W#*|m*?-c
z86#die+gsS=bhF*Jv$dTFQB8785eu_#ad%sa$4Z=aW*m(9uzr;<r2QH1$gfDF^^xQ
zcc(=Xj1&gTm+wBKfWabt>eA0EXi(h31q)oNh(2qB;%4<Nw1RGOr||w<d3uT?rk!GQ
zFx%xTQ!F+~l|^ix^h66!VH7JNm%8Y4EHR0P>Fp}o%3xqzINp9XTW3X5+;$$OxMAA#
zQ7&J`SJ%cX$WjF7X^VDTp5x8ugugDL4Jm#Wt{c`FdfGI)%#MfY{oM9g$gsd=$m%2C
z3Jtn&Q<~Gi@#w9#EB;?Do@l1iKR^HK^mH;k+gCX+QldH68>=wc0k1!|%5&*Ylj~B3
zF3}!HD*?$|YoC%a@Q~Z<u^IzovCDcfZ8I-efonLh{4<)1@V~}(&Pt=wL!WuwX;)!f
zS7NzW1+!+F8rVTw50nM;6Z}2f{X=-{bT`r7^70i5m(5zQm{%t_+IO6pTesv_LV5Yo
zPTelg-%HO8+d76Ja;cq*=htTu9@+oPlS_W=<W>vW$y~6L>&_1qS&1Cw7HNNJ8f=Om
z`gcIj38`91>_m+GdGQSGAvgYo;kog}+BdEjiEH~9jOKQ*9Ik(}MSzDnX|JNZ#uplj
zvv37WkCm9BVxhAC{A|j3Uwo-lP5<Ho<2(T%$Ng#%>zmW||8suX7mjcI&-|v*27~S@
z1M4|>Oq3T-<{%lmYZNYBG8RQ!=3d&t`{LfW7D!m}**8yYF^XXv-eLu)FLFo?q)iM(
z-gCKt@9Wc6OC0XbFZ*?vy)TZ7|Lhbw@}1L{ETdMZxzXNvToy0#RFbzgiP>R(=Tf+f
zmaxA0{N$gV>qX%@$1Y+|m7E`N4-Ci+7zFh;dal&r|Gt6$`w-L4@dh%+9v*vX4>R!Z
z&ZXU)%)^pDw<4<;w@OH873WH|UOl&x=Z5y5=^dio?Vsu8A<ygl@77&HC-Z#!DwS(b
zCpU;+b9W^##jsBC`NocmCz7<<U$30%CqpAwwDHxV^?7;*m$ek-#pW#%%-<ob0dYR{
ztKObT#k3AX46qX$qs2I)O(un{=V7WMdEezcAy@${bsOh@C8yx~d-i*a;~n+7%S+mI
z>ikCJ%XkLl<rWri;r+6miQcE|rqS94pQ}BQ(d$^fU#|CCWwq2|mjzY$3sdw<HL+;M
zH@keT-p1@>(q-6SPv@mXrLv`1<R;~P4cK<~fc;g~;~xuB2Y)9|7Br;^j1D-Cqz&Ya
zu>>kgv@kM3oW&5;yoLbqPkD-10~KB4;x<1}-7L@0wLcj9apU1f&v|Z_oQqyM^pXK1
zD%Z>~q_jEoaQp!?vbGox9@C}nKexMjnH{(1?oDKM&0S{eL*9wqMNP^GWj5;1UE+c@
zos0=N_%<G1IL>ohc1bJE;QpZ9c59HcXTh0i;>DfW>mkVFb4cRul=Z@YKN4h<8&}H?
zuEXO5Q~Qnm_=7UpP+yehKAh>sHSYW9AbF&1bB~ni-lSX+i^b!;W=HD%PX04SY#=~9
z9H{uV#6w;^u6BN678v;!xR5tU96eH@wfQ@lOvY{GSG7hL4%{T<Tk&m*!J!W(O(nQp
z@g4HtjyFE8zf7*oOjqE&xo=P6htZ~zExItd9%6xCeD_P}rMoRGh%Wwfckj)+4g8C?
z$@dtqB18C?Vd%}~pu<}Z$}^nb{{!ySMnf_nB`!IBVM}R4xT%EM@A~zl1_nVpmu{s!
z!>x^4Lc4pf-%3#3s=&|{PH;GHHo5}iM>WD*O3iqv^4pK(4q)(hk+0fFgy49b`yRyr
z-&6Ndr481BTYM}&hqh|J6!1K1NSg|ah2v)eQ<4wfby?qTPAn&*<Q3iOeetIMOctIk
zI#6_evYu`#V!fgGpUL_n?;*(2X!D=RA|LUy(Nj$nC$D0)W$p_)#TB&W^9{>q)5F6l
zZA%JxdUMI>-rT6Q^I321r$X~h-G6=|G|cieomgNweEtnjVCka+4NDwu5bvPa4{!9r
z_P^7$XMxMpk~j>++=@7WC(0rw%v%@(bPku=LtCn%3tXxb@TRG4$^xJN+zGd~DW9Z{
z`!Y4mO+?O_K&*D^xFQ2bIc<)~D4+X~cy2#K)Xp8`G%gOn6ZG{(n|B`aAk(nvWUBGG
z-NL1}PZ4xRPO8-UO*D5<;(sQJo322twxMCOyhtDe_nmEeHdrX*fnB}|lVyLlmDJK9
zC2njhYr0n2IIX}{xF7m!K3R@0_|c{kw`Kp(t6n-uv81i!b+>Va+|;;Y+81u(0-p!J
z7U>j|Eq8t4_?KH_46bh`N#CPlZ+sQ2V7QDrKDSACZ7yQ<%FWfjL|YbzJOr0v4mBpv
zdx+A8N@anIRmr2{{bQHY_T~2A3O_cLF?-oa3{ss~$uPN{%lUFEwd~KH62p#^7kPO2
zZtChG7ILb4lQ(`;^WRCr$zZjwq3+J(wk7j#xO50_m|BX9(00X&xe=H(r%KZ72<;JV
z^pb=`7Ys(zhJB|N-alk?(-89(0SS!sR0(a2l-t*E+XyWoh6S%3eM>d|c<F6M`IE<7
z$n0#vlRP%K)=E%*kNZsB7&3tndP~|=9rC%P6vh1zQ+1WLKh7M)KqOvAi$L0b`hpjK
zy19b+vRT&FP(S?*wB@jau7IJ|D0)lJRVumSM_f@(Lholb#gew4?kybuB}mL;{St>c
zixDKIweW-OpT()_A-JH|Ch3uolL-}PKc1-lhFt5W5$1ja+>KFify?h;@&bY??PYU2
zIEZ4f@I&w1Or1GQzMiYz8$7>E4lNz}Zh(=W+fTk^m~P{KGXqp7Eo`5S_lk3V394z{
zOX_3K&!TOpLwB=m=|6tC-PUTU=h*$kOz8<4Vo~JG)?db}PfqqIDSeQk*&wWtYw8LL
zBhi1;))!fzQHHJ_qO^gIBLkgez|}stH1UjUkF7MQxGZ?&`JYj{*396d<im&6$awIJ
zu9`Q?(}SK>)iew4K-No!pYHq=@zE7*S4^%7C5|nl0h6AZ_t&C-{^xhYt(BRup3Yau
z1gBzn)8e9J*yI_aq<XE%``%S-m{B@dwS=~4thN?qyXUCRKUOgQRM)B89`wgBSk6YP
zwqG?T?(3IB(WbM<=KgTR!*X^;<d2Elf=eV`7vzdc+rGo3=^j2?b%9I^+hxHyL!yKK
zGh!yz-YzmDg{_E1OXJuxQ^%8ft_ns0RpgoOQQBsvPvs5BeoS>XH~&tQ;`Qb*UXgdD
z<nfY}T^q@cmRrF<Mw*eHhUyl3&>Z{!-~Q?7r}~xK1xn97i;DW(@A7VPauVWzou25P
zb8Fw`f{JYaG)11ka>F$~o<|;E>+9>={oqhlxrkbA`KMgwr)nrrp`*j*Vi8*ME^|MM
zQ?TU^MVT|TJnnkdXg1>YVD(Dlhl(Ph=KUg17NH#Fte5Q6X@<`zC^J=h#O*_Ud6a7|
zve!^*bFKVyq##{i&DBc#C1_W$cMlj}LGwTw$dg}6^d-cQ6ig^*2^^BrB>y`D)l@h6
zi}85=1ubUYKyGqMD1C8&3W;7fE~fE^-gI<WpOiCl<Vf&XM?=~F`BPI#=B2v2I^-uC
zGMvu)p4AV8Sx40CmkGM#K##wY8)f30p))t{|5_zOstOT}Jj3a<{9j243C6TvLg8p@
zYg>BG8u}9R7M4%(HWJe)o8`%vl$3<bqN@6Gx-Tko_a0)9#*_S?ms6~H3neS^)2C0Q
z=ToR?T5fJGk`n>&4f<>gM>@=h`tLl8Sp@Z^AjO0vG!??}Q1dJledu1gEv(8ozZpVQ
zX~;RtXC~9!Mw5ZGi;u;m?mNC{P&RkHUzqj#9e7W+#EUhmqTjQcEMB(U)SLjcei1F;
zNT(yoZ%-6zVy@*|^*YG;Q`#(0;PC?V6zBuAj2k@yCIE3k^xeR^^Tfr{Ey3y<8kx5J
z<I2vYz7Mo)STc;SKuu<2HXsonA_aH$Z+mD1%(Xko6?Ja-#Y0o~r7vr8x7ni|l`&FQ
zc%`}A_!{ztbF;Ft8Wra5D??Y%k_$$g6>WS#cM>=hf*rKzn;_|yBwq*VQba4RlkIPr
ziZ%M}o{mLgTAFO<Ed{W3El>&LLC}!I47y!Nl$#*^EtvjpsUjRB>qs6GZDE8Y_5<>~
zlxC|GqVaJiCbYU%daSv|Y=YDcL7L(h5BgO-SOe-`EZ%P2AA1bgd=(PE3!;|IKjNkK
zrg)_AUV{1`vfo^ggCL1G<v*IlKO|KGks3uLrAUji=wCcsJ~~o!q@k{EC29$wcr{XS
z;WMBLo%6S+@7OOL5(#xhy5uGDbj~9^YW3J<-hDm3X8Y)V2UsYlhuEcHq2r3q13B?X
z!@^AIr^e;2=NCejHxN1F+5|P!3TQt1@Kx(`8>vVtU4d(eq`d{hX{^MFN=<WU{vg&U
zF-!-JYMK}KLv;+R={h4R3u<iL0tzfla)W$Aa<UE@8Ul%izF5RsL@&^gIcv7ZFryip
z4!?D(KWU_7aCEW#9+5ZbpM)&PRtSfYCoM77b$HX<sBD7S9YKD`HAeLQ{Forc>*Ft<
z4l4n*h<)SvVC6344KZ*luIW%<fVM&n*1)bS+$EQ%oY!o&y4*Ml6?3GptpRLLwT=}8
z^Vx{TbXA~SWn#a%@r>hXOI-=hDsS<j?==996&R31GZB3oLfy>0q24|KO@(x3Ad=hV
zS%o^pJ@0yaY9OtFJgmcgrrRCGpkhF#<4Jydt~vY>6!xXUSiTRBU;W~0xyb^iJ#WK?
z4X6uw*PzEj`84ydx3L{`ehj!3<YQEM<H?qbpm^jN64jGdeo_o0GG$20xIH`kB?WkW
zW;!2|#sZE%kxPqO0sY8-!>zGha!5uz=;D|Q|GcCR^r5y`HD2u2`#^#IENdy>{`lXC
zI}l&P?TK1S%3aLN%-o*9@X6c7<KHMw_dQUB)5O?#9V&calsWFIxDSK>FaZyPip{+I
z{L`Qmk&G7wn=p;y46SDaq@!Ez{<FlHz<jyye!#x#jAD$n{hnpXPecVcRvRmNSVq7`
zA#(vHye2_vGOj)~Dx+$-;ERipk)>*9LXtp8^x5%%+53_<A5VLI2Q3F>#NW~lS|?A6
zBOwdf#3MP~;h{*`hBAJGjgm;@lW7&v4nAbP4WcX>$L@3cY8J^f|NE*pK4u=N{*oKv
zKI0fOoIo&Y^Qx_9VsmpRy9qM@yMcZ{uHBMtIj~a0HO>rCAlk7o%p~{(lYSw)Sw*(b
zS15VzF48}W*Km7+K+EXm{`b9#8C-tXVk#J4fiz;f>yFy3+@9uHjd7Skm&dv}!}IwG
z5N$cae&O^P;Gb<ly0?ty6WA~Go!qw>Z?eCCGQi^P+Z5d#Ybevf44CKM8B=~9S}ifZ
z=#^m#z#3g=)JI`D<*wx{I~llzv<sIE#~w=re=&t}HR9i%ytY8?3&R>`v^K%6M}dfk
z<%Vn=u&x>mQ6Oxp7w6>is4~C6jx%fCQz;cxZU|y2<dDcr`n#5!SNOj99ehW>fr?wX
zk9)Z@aHAiFth;RxPGKAZ7MXRL&?T9Zy)6o(fIYEtgOD-SMRn_->fPXAQsNB4AUuq9
zi~>>oF+@RoS{?B3Es@U5B`=0^Bb3uxG^Tyo2~I~93We$Le#O1&&K)0qcqqfKGr@V+
z*H$=-RgI$}yS|~pixH*s>|1ObVrzN_-Ny1z+{C;^t{^AesV|Z76=9#A&~bYz^8Bcr
zovP&a)Un9fvSA?ucWr$=NjW$oJ&F@V!{VWlUSQm^RvZc<Uoct|g|oj4Vei0fiAeWZ
zD8HQFj+;Yd_9*{;0-gp|uLtE#k`oeoaPT3`$3ROfyXLUu^cU8`Cx$!Nju{xZzgoIY
zaYz11Qm+l-ub2=l{#n#6g{4bM;uRDmJZpNRnTAFky2_EiHK+CX3TX`CEO^T8KK`1E
zWVL}GYq|J(Wx3`>(HPsq;$3$z-f%8Vy*TWQZ4zkPr$oyd0=NuYwvRm6M6jtSdQVCn
zp{$tA_fMmcI2I1QS`NK#ZM|wos43{F!rv`}BFCbTY||4Hu!?(gP$?8FmLyjdsoz+t
zrWY>QVMLLyO==PN#dY@n^IS0VC{X=_+l&1f`dA3-;Pghj%Rm(}3DByEey7YIPc&DX
zk^bxdn@;Q#lf|n!!_Ym)r?}GXSGr$=)M)MTH9m^aEft1eH!sG{;mpiIbJ3KhXE~@|
zJ7YQf%aW9bgrZ5z>P(sY)6~=fs4>4IXCA33P1u?kF;e6O7Wp~yR#A#Wr113U=xDW<
z3+OY8nMMjFgx9=(F{QW;>ZdiaFS9K6QTNB!%#IwH%|MZfo_9E+vxis-;fWnAsQ*H&
z@FF%Twx9z>1v&C>O;d9WaXwx%=cmY>26U5>0S(oq3UTZbar2g2Jj2bGDwTR&QT(er
z>=MdwyJ`>Xpn$FnYVfda*dWgU^@k)134xJ6si#NMG7B)9D7>H#Zv<yOQYY*_?B8bI
z2-v&G5I&{$AhJLl+b=IcORmQHl{;8qRxC!ofO{sAEuCRLB*QdA1=aws8h#4$*`^I<
zM-6@k#MShoEWpp(2L&czm*B?aliYk5)oYfb6glqu^sB2jARn+7Wgr+%!-*gcsE%m-
z31pG=C@rsXR#h!8-4>y1U|?|a<V5Z9Bf<v;xjj*g;U<UVFFf`tZj$v4?XbdWwiT44
zatoy7OtC(NmpLOJin!pznfZgL$bg^1tn4Os%mV&k)+9nY+(2%{?FpyX5_JIDj%K38
z4)R!KxIJ;sVBe>ux$+-P;0r$-e>_oLB(!DfI32}Y$zzFeEEFC{&D-gD5xJU`=$*Nx
zkksE$T7I|k{;rtbYJ6Qu1;sdT^F2;VI<T3Ebd^Hc=3Cqf;GK1mFsY$oTST{n6zlNm
zQ`gcW6g}iDD9icL>K#YnSpFa}{A?z#v4cULL$->+86##0vsZbY*Bx(MIQ~Wu%l1XN
zNE1xWcRV)cxdNp`>N1Fd#abr(KjSNq#J_M*!O?>3Gh+)^8N?TG5G7ITf++bkG>Gr^
z?c-C-mz~=~sP-iS_l2RAb9F%Vl2H_LbLcP%M~w`Is249t`N(F6a+Jx;B)tYD?NO>^
zw%PZq`93!NAI*+3$h%ChkT8YUo`_mbe0)upDy5-@ZZVcFEP>M~f|3Q$jF7Snu%__&
zh+~gMCz#P`u#Dou@r+{Tdv%q>cF(_1OVUi_Hxfw9VE?&2k=`cx@UUo-<NF=)o{Xs1
ze4B3q<|oE^5UHBYBY(qq-1LWbK}?I3d1oZGOE~h4J3ky0)I;+CQr&|Sy=9IKY$m2i
zUCe!E3P*3MNf&$^3lz}y`}7mx3_>#jRlaRZKG;}sYF)_0W_$&c{lGbm@f9M7iMtW#
zthooRaO+3(P=(_ns*iUj=|S(U+OG@QESvaGlFX$c*>Q7E!L#+DjW2M_I=`3&nO(x1
zRmWWU&<Ht7h;}S%-Xl8@<y<Y?rV-=0Dd=#b=uJT#*GAgnEs85$J=AFSD_(ZmWUV!1
ztdq*Z%h$D|(t$h!DXRc~nxr4sM9gI1U?eBdQ@qA?M&eDjA^Dx`!sk{UOg@SXFvJ&-
z<XOGnI!a<$C(0tWqtvP`R(b$L|1k6@G;dwPJ>hpU$NIcL#Y3A9`(Yo#kxS)1_0AnW
zo;dFd6jJ!vuR=$%sZp%I5q2RJvH>KAqg04X2iI5R;zg&MW0D{$APu7|-F~8Q<Xc69
z8=j1u5E)NR=D3%<J;L@t!uEo?xU8DxZPg3<r|G8*$$BAY0plx(iExnlqqI<Yxb*M)
zXZ?LzGvE>ByocDqyPy#(iBvB~+Y2#NT0y*HhbY0HR~CBDl;11pJQgsu(}}3YPz+E-
z5`vm*jJ&+O!)H%qMbgQ5UW62-ODh~E+QMT5r3lMqZLsK+yWA&hBBF0ksr`Vt+!j>Y
z$%`hmLZad>(B9?@!}q_=2zM8D^nv>@vhGF)+hjeDLOI@6vAqZ-Ux^>8y5wi*-O_V*
z(0(z>!}SjJU#!y6il|4$%~W(pkpVIk)Vh$Ek@0B8P$caMo!g!?hZ#qlp_+Pn4-Lo(
z9k7ZdZK{>xpKNzYrtXAr8-l-{CSO!2$zK$)p7P}U1NzlZP9kU_<HIP=d-r{LmB4eH
z+``(tgTNPV>VJdP_G=Z*u6L?&tUz$zgo-rL)h#8+JeWd4u<{lyT`Q#lR?I59{CIVM
z+a0zieyFv@H#t^cTdQ~@yW7;*c)Ux{{Q+BS+jDFkIC7!fo^b74P^M|Z3w7&<BW5Qe
zxIKv`kkrRHSW!<J`bD$C99g_cv+Vt9t{as~3Jc?mgjKt1;6S1&LF=aBe82|Wo+(+M
z`k*l_Ps>3dqL0*;_=k4dLbq2_GWh8@PhH|bX)3I025u?R1zBpi<etL~C@R0k4=~CP
zG=7ESE97)VVw<Q{J~UO;j&B3F+Ipn8J>fqoe;&H4rSEe+Fajy5fD#x%QR-G@>>_&r
zCX-FxIQK;(46b%TfYV4nVwUx2UR4u=b?Wo>mQKF6wyUe_2+sqva(6qOTt_yo1U}Lg
zZxD{1-?GBd6iX$+XZ@_P0fxtF>*|E{s!6*|(i9C<Jo{(3DROi;aTgJ&LFsh`%16M#
zF@IS}N<NA5mR|dyyUx@pA(4KSSY+ccS-GBQJsRCrMMTn!uP`~y-*~`TZC_J>{pI&w
z)Fez~!%PeOg|f1;P)=z5kB<ny#A0X|zVx*pmVYGL*68Dh@}}qVdOyhY^P6eU8q9PV
z@bd8~FyM{E+db%0(Ju`8s?!L;LqC~OPTj!1Sayo+azkA~ovfYNk6Rkf23-#h)ytRa
zTr{}dU%Am1r=QPCZ%um=9>@G7wHvzpd|J;dsJ`d8U@Py>;%xeP{j<hzB*O^NCN@M`
zykkn%el4l&eV^Zses8je^b{xeQsT3_qpf7yM1?&LI2XVn<W&L3q?qwhjvmF4d^!wG
zW?+I&F?<F#<BY8Rrd;uI<lERD&Uod>z`f!6G<-&F>#Xb7uQ$Ip55#dLFTBH(k^iQa
zj!s@;_stdScW-w30THnK`7Y9C9qV5TXXH=Tsix@i-%e(arv2FMKAuLh;v}B;u;llQ
z?Y~e9msHmz=OV03IG%vN@ep!G7dm*n+P5_b)=pr@4g#;B;CE<+^2`C%B^k?8dwGlk
zo6i_lpZ#2&WQnSkGw)-M#xOQXRQFA@T{y;Y8dXbjwr0vhy%@8D0uTmS!HsYENNvOF
z8_lV`SWTZzbH0Q=g4hs=wo}kALK*u5jxdZxiwdXXIfm1~-ekvLt|m&jEQ^hlwUTt4
zk*7#WE3WAt6>d*tz>;QeY>5waq;|I9NY9Mb&T56|w$^NdR90n_$FZHU-XTM+84*I8
znWO>(yc+zZjNj%lxO0mXg)?086z+B_F7xE_$!2~VT-}TDf~o1xu~Vk?c&2Dp$mQFZ
z%1x8`TIG>%bn@A`{WGk~kP$8Y^Y1-mQ=vW3HAeKvy!+dqo?^_0N#xc;or7wrAs`ZF
z=W6YdlIeE<aRbUv6`W^deTDiywL-R}yFs<SF`6(qs55<EL24ABbwi=4({04P=tQ2G
zkuid!wkhnbqH}u6$@Hz+OC1Xkdhmrfai|_LehB-E;uBjuhoBdC+O+!eEYg=<;qNFg
zV3sN&!OK?s`Kix0y9%2d!7PeMC&YQJ0=n1p20c>?T9xT2e%PGXu#4i2i=V+Xn4D>b
zhZ>Xr4oOoXvd&VW<>;}*?X=nX28s3DQ2VdjXdF=uT7C0C#T`@{G5yU*?K43)M*Awo
zI&PwP{cO3cf}UQ>dZVzX$bwMhu|mvkdHC#M>Csx_gju)`Q#g_DxTDRSWA%MUbO#Jk
zy~|&VdV<^C;%j<i{3ZGzfj1tRRuT!#hb08WvS!+Wz>Qgs=5B9vwoiUKgO#dk&+;5@
zbu{Y|4x6fmmYtOL!A!RS9!j~y*ZXG)aQy9~Enp6~ACNMV>T|@!o1Gb)ok_8D?MLwx
z^6hY!Ve})-d9XS*yBH|eB+hPRB)xzm_OrWkHf?q?4e#p7NFbr6SJHv|riOzH*>|5S
zL4X#cH-M(B&5(iHw>lajuBO4ppW#UPvG7}F!Y9JBlhf!otTmiWCsd45zS@cerF_7d
z(b9htf2I6^Ap&mL^R(Q)BrKIj#}eUT4+4pj1XN`Y3>7{LO0d95?m9Md9(f1FRF$<^
zj%}WNtxT|HjzAHxP?xOSu=e}rY#eGRwqu3?BeTFEF9sNmNUpOiwo=ySr|CvAc(_!H
zbw=*JDss%I^!Ao&>j7?0U{sx4lQ;phTI`w(ElYAfx4pT++TEwpk&SFQ;}I5u&Ztko
z5}Dd;&mQ=Mb7t*bMZRRB$|gr<`sJT+L=sa{%x%I^0OuPDyudXKD#h9qNn(Ha3;hEo
zlPc4mhFTWnG;SqwRIGAyb@6P)sknkNth`H2uS7Kvt|S{wU)U2G%yy~}V&LJjrcPhW
zPATN!OfK8%IfM!lUzeeb=$DoJC*7;|F2QCnJR2cM8Oj8VF86u#TO;vH&3+w4E*$Bj
zLU1cUucg<fAcbwiAQ**Ny=5`iz(fc@W(_>=2QE2K9rI38$2=2n&VdM;fYtC@QP=ik
z!VId6<|4Wz&?*2hf{G^t!e7Pi@W~{88h$_y$|*0eOInTsYN+iZ36u7Fv>!+qei{qi
zX91<m3c62};gHlxuqWks(HRuPhE1)S$|z0`q)`A+!wE0(#d{N33Z(os=#bAW7xWP!
zs9M9}BiSYpM>iwfAiXoON%0`Rv7|zQyP=uKN9-A56B(HrS3dr0MNz{km>E>!E9G?l
u7TDzT@K-tW5{kC!|DA65|B~36UA<s?m<fA*$ziY`l>JJF6;tI;UHM-`YV3pn

diff --git a/Perf-improve.md b/Perf-improve.md
index 2a20bf9a..533387ac 100644
--- a/Perf-improve.md
+++ b/Perf-improve.md
@@ -64,10 +64,10 @@ bench::mark(
 #> # A tibble: 4 × 5
 #>   expression      min   median `itr/sec` mem_alloc
 #>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 lm         860.49µs 897.15µs     1094.    1.27MB
-#> 2 speedglm     1.56ms   1.63ms      611.   75.01MB
-#> 3 biglm      767.92µs 797.64µs     1189.   589.1KB
-#> 4 fastLm     987.36µs   1.01ms      982.    4.48MB
+#> 1 lm          881.5µs  916.5µs     1045.    1.27MB
+#> 2 speedglm      1.6ms   1.66ms      601.   75.01MB
+#> 3 biglm       772.1µs 809.23µs     1174.   589.1KB
+#> 4 fastLm        993µs   1.02ms      978.    4.48MB
 ```
 
 The results might change depending on the size of the dataset, with the performance benefits accruing bigger the dataset.
@@ -100,8 +100,8 @@ bench::mark(
 #> # A tibble: 2 × 5
 #>   expression      min   median `itr/sec` mem_alloc
 #>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 base         1.14µs   1.18µs   796519.     2.8KB
-#> 2 fastmatch  990.93ns   1.04µs   929883.    2.66KB
+#> 1 base         1.18µs   1.24µs   756897.     2.8KB
+#> 2 fastmatch    1.02µs   1.06µs   904225.    2.66KB
 ```
 
 But, with a larger vector, `fmatch()` is orders of magnitude faster! ⚡
@@ -120,8 +120,8 @@ bench::mark(
 #> # A tibble: 2 × 5
 #>   expression      min   median `itr/sec` mem_alloc
 #>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 base         23.3ms   23.4ms      42.7    31.4MB
-#> 2 fastmatch   990.9ns   1.04µs  917170.         0B
+#> 1 base         20.6ms  21.95ms      43.0    31.4MB
+#> 2 fastmatch   992.1ns   1.04µs  897837.         0B
 ```
 
 We can also look at the hash table:
@@ -200,10 +200,10 @@ bench::mark(
 #> # A tibble: 4 × 6
 #>   expression       min   median `itr/sec` mem_alloc `gc/sec`
 #>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
-#> 1 as.POSIXct   29.97µs   31.6µs    31256.        0B     0   
-#> 2 as.POSIXlt   20.65µs  21.83µs    45281.        0B     0   
-#> 3 ymd_hms       2.22ms   2.29ms      432.    21.5KB     3.48
-#> 4 fastPOSIXct   1.17µs   1.22µs   779859.        0B     0
+#> 1 as.POSIXct   29.74µs   31.4µs    31531.        0B     0   
+#> 2 as.POSIXlt   20.59µs  21.79µs    45434.        0B     0   
+#> 3 ymd_hms       2.25ms   2.33ms      425.    21.5KB     3.43
+#> 4 fastPOSIXct   1.16µs   1.22µs   766780.        0B     0
 ```
 
 There are many more packages that implement a way to convert from string to a date time object. For more, see [CRAN Task View: Time Series Analysis](https://cran.r-project.org/web/views/TimeSeries.html)
@@ -264,7 +264,7 @@ rowSums
 #>     else names(z) <- dimnames(x)[[1L]]
 #>     z
 #> }
-#> <bytecode: 0x55a2a63ce4b0>
+#> <bytecode: 0x563353356400>
 #> <environment: namespace:base>
 ```
 
@@ -275,7 +275,7 @@ rowSums
 .rowSums
 #> function (x, m, n, na.rm = FALSE) 
 #> .Internal(rowSums(x, m, n, na.rm))
-#> <bytecode: 0x55a2ac309f60>
+#> <bytecode: 0x56335cd3f7a8>
 #> <environment: namespace:base>
 ```
 
@@ -292,8 +292,8 @@ bench::mark(
 #> # A tibble: 2 × 5
 #>   expression      min   median `itr/sec` mem_alloc
 #>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 rowSums       821µs   1.29ms      761.     859KB
-#> 2 .rowSums      818µs   1.28ms      837.     859KB
+#> 1 rowSums       821µs   1.29ms      859.     859KB
+#> 2 .rowSums      817µs   1.28ms      833.     859KB
 ```
 
 **Q2.** Make a faster version of `chisq.test()` that only computes the chi-square test statistic when the input is two numeric vectors with no missing values. You can try simplifying `chisq.test()` or by coding from the [mathematical definition](http://en.wikipedia.org/wiki/Pearson%27s_chi-squared_test).
@@ -360,8 +360,8 @@ bench::mark(
 #> # A tibble: 2 × 5
 #>   expression      min   median `itr/sec` mem_alloc
 #>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 base          904µs    927µs     1008.    1.57MB
-#> 2 custom        708µs    721µs     1376.    1.13MB
+#> 1 base          906µs    926µs     1071.    1.57MB
+#> 2 custom        706µs    722µs     1371.    1.13MB
 ```
 
 **Q3.** Can you make a faster version of `table()` for the case of an input of two integer vectors with no missing values? Can you use it to speed up your chi-square test?
@@ -418,8 +418,8 @@ bench::mark(
 #> # A tibble: 2 × 5
 #>   expression      min   median `itr/sec` mem_alloc
 #>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 base          638µs    652µs     1519.     960KB
-#> 2 custom        355µs    362µs     2736.     488KB
+#> 1 base          638µs    653µs     1515.     960KB
+#> 2 custom        354µs    359µs     2730.     488KB
 ```
 
 We can also use this function in our custom chi-squared test function and see if the performance improves any further:
@@ -477,8 +477,8 @@ bench::mark(
 #> # A tibble: 2 × 5
 #>   expression      min   median `itr/sec` mem_alloc
 #>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 base          909µs    940µs     1059.    1.28MB
-#> 2 custom        418µs    427µs     2314.  594.45KB
+#> 1 base          912µs    940µs     1057.    1.28MB
+#> 2 custom        420µs    429µs     2296.  594.45KB
 ```
 
 ## Exercises 24.5.1
@@ -601,6 +601,6 @@ bench::mark(
 #> # A tibble: 2 × 5
 #>   expression                min   median `itr/sec` mem_alloc
 #>   <bch:expr>           <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 crossprod(x, w)[[1]]    871ns    922ns  1001700.        0B
-#> 2 sum(x * w)[[1]]         460ns    511ns  1796479.        0B
+#> 1 crossprod(x, w)[[1]]    901ns    961ns   966191.        0B
+#> 2 sum(x * w)[[1]]         461ns    511ns  1759060.        0B
 ```
diff --git a/Perf-improve_files/figure-html/Perf-improve-28-1.png b/Perf-improve_files/figure-html/Perf-improve-28-1.png
index 6b43e41880902ff0676446dff2f9792320de2d0c..27963a50bbc42f4a46469d8be97372861224bdb4 100644
GIT binary patch
literal 52143
zcmb@ucRbep`#*eESJ6TWMOG=5RU(y{RLW|J%m$&1WN%$5Nuq?3(MGcORtO2%d+)vX
zx}PuC_jmsu_v8NK{{3;EkI(gScAoFoKF0HS9>?o+yQp|>>n7$+6bfbQd0Cmu6v}!j
z3T0h3^;&#l_m9jA{MmS4_PQB`!s<%?TjSX^rc0r4P|nMox@sLT*ko|y>Qsiv=*k}*
z+JD?Q4!uu)sBfnk#Sm71S3P1=Y_f7gs+~HY%DpHxulk0MKK1t^_#$`?U1&NZ6O*VG
zc4k{}`zGoG)K{hW7W~C0(swG{cJ|WQ|MqR?gpF{}iMJC)+OhU#B>^T}Y@Sq@3P;?T
z3UU0UMBiR3hCe!M*0AAEtTTlcf38b8;~{5Wj_h?5iWDPV>l(D!Og)OPyw?9?jXwoz
z|G%3RoS&gwzcJNcHD)cw=gNPGyf$$f)<%DbjTPiOs+0Ed<Hvvh{a5&2^=CUTLuEBJ
zW{<NSW3&BL;T7U?iHG?3_yh!6Vs+(GsR}RH*x2mbw~x~$Ha52Sok)mX!1vdxISKb_
zsw*o$+xdRKdinAzKR+hv=qG&FE?=g1lc|o<4jw)gds&DX-E8H&s;vCQbZeT$fN$jK
zbyFwS{r|Vf|Kq*k;BHzssA^*~m+Q7&uOuMlX3i6b>AIfQN3;?{@fEjXRL!*Y3SaG$
zoG<gFIo|4g<o2)lUjf+ZzqW$n*->f<cmKRO;?5*M!)e}Et~=bAV%AgQurQQ#?#1pp
zy#Jb|sH@vmEYhN(tXy^18$u_=^LtXR*CeIOH_x4k-MaPROru%pI$E*&Xf!k2WLFY!
zq9fnaq<TR9&EcDU<zFI%xG-9L;&YnOiJsD6aXZTp#lW*3%vBNUNpwGl(?%SdENb#@
z#TZn4jnGJmb&Hz##bP2h@#m;!s<}b&O-RUl23>RK;3K!sQn!uf&_+2W*n6^x&(Dmo
zd*)-ob+4J~qHpFzBsHg2D2fJR?GeHU$cV8R(VdZR6X$N*-Va+7xf8j9nm(9z{EE;_
zv&?<Et%S&+SBBup>?qZkJuY6!)|r{PzFkFrBclw<i#?L10VgKjwJdhGxw1DiZzl_H
zDCkZ5Yto#4!X<jDsaw>de}{A}#HfQk^H$!ov5x$i!McfmH}mWNZfmxmD0Uafv>q68
zS{l%<aGcKk{Fy6bolaWB$`+Y8nLXR?y?G8H#mijjWz(k3<W(VlhH;=i;a=*w0^91t
z=N(^g9TX9%VP3n-VS_~3eBzXep<Bx%o<oP2=W6}^{bfuVX`N<s78e(*S?ICada1L)
z4@VkPCQ>a1e*gZ>Y^W4qXOd2}{K1vYc4cWMYbwHIYN#Q}&9$hsbfhk>r@G-DUU%>M
z&+Z9h@Ho5orR=RI)?z`vyJ9tOZLOGcie`P9cgu~cIvOXVD#jKS5uq@{as41P;8Pjh
zzruzwxy<t?%5Panx;wDoA1W~dZH#h@yow`9KT=z$oF{+xQnc07QjMEZ72X{at_`X!
zFYqzMY1l4~<u6PR+5Ot3C$w~*YWMx@D@x)CQ`1%9Dp8tgiG5Mt(iVPA#`W=G{|{X@
zUt%|;x*#X_sZ3vtEeXn~rlzK)m6>C;6s-UicAKH|@$=(uRykv~uloMd{M5K!(d*Z*
zJG;6b(;YLcy|GpOD-M9(?YnlCMAIXkFZx1g&nz7p8L8O8oNCK1D07x-Hqf&EpiA_f
zV*d>d+fUu<3AAc%rixUJ`IwkEoiWqQa?)J?QAo4n!qAULRIYT6IM<Q1k<5cZJF)v}
z=g?k!y0*hZq}h5PBI$+1eD8ci3m&HXsbx2mkU`7K%iCy_)WmGaD<GhqW@)6QrL||S
zr@K4xX93#(Jp?giH_hGV&8-w9eDB(sGiRDSdNK3EcGH9D>FKvU#VzgFt!LX@O`2Fl
zZvT27b31=}q{RuU8*yBS2>ZX!GdC|WWsW{}T$+b~PbXz98MSNbF)%RP&VLc{?c2%j
zH@Ba;yAMnUoV4h#P?F5J-}ZBI>t+h&AcN;Q72fc-V>A9)8FWqyQw1AlwmHAx|5Tdo
zN*9?v{yRJ>({_HcFDh$!p3QD>><1P5)oa&=GZ%(?f}I?fnuf7A5ID0^@tTJEH57fS
zTSdN%Vk?H`Rloa76V0E=e4A&zIF{huuAacVj#UOyO3OKz*kGL0Y&#&5Ip*>6=g-HR
znTF#l6i+|dGB*|F)MV|t8-61yV5Wyjd|V3slp1;DoAs3P`5%{vJ;Os~esNnwv}nBT
zRKIz|WlfL9CtF!63lii%o;4oeU*{hhdDa-Np)Y8kde}?3vzi_n*RQ{2PVqhey$#Bx
z{60Zm(<4CG@QCn(Vuph}JedntFUQA+hAvi(IFJ7bN$8Tj+sScSXMPu)QSZeEJ?T}l
zwzQm`nIfeNZ69va!(nke%3V&ern1ZJh@tgMEaa8Cvu*WL8UN82A9}kLj1F(~G9bRJ
zfXil1v~gchR)>T0^f@>_=Pj4x<UCo2ucQ@4{_m$as;HLFJ(Xdu`;e$U<F9FD5@+N5
zhZDw0YC~94id}lBb(&KCtuAfJlrb0Dc$E`<#GlT}iv_JTzAmsFzQiQJ%PVQ1uMj0(
zZ+(E{G;yY02gg^IYZ`6Lq^UU>c~<HRV-mv0RhD?PE^QmBO!=rc@=aj(j#2^HndegV
ztIuzHpqJ${vvS9UPMYho<<Iy1>sCJPXCAsrbmZH&ZyH$=#S?4#y2VM~j#%~%F-hUE
z<nhqk-)I*-Crj4pa1bwF&YNnh+}3&Sw1NWlhr285csA_z9(Hq<=OT+adFJ=;P2A!;
z4hqn6R!$CTnwHavyroJ!Pxd9=?Wzisg6)A+V<YMhiA9RKr%Rvb-|A!ARB(}?fzRhx
z7u)+xm(#`i=FPgkp4`pu<>K4DQph&TdM;*Vsj4um9Gji2j}#5D+r#l`Bbn%il(?4P
z3r<lkqXq^V#cU4cVsu_(q=o*y3mwj@;N!i*L63E6XUQBAKMB$3A0f-Ae_DQ=V;31X
ze&ZKYsRye~Hlqm_>Gp5`>tAaMx8um)TcGh}Y0~)GXCWF{2GUr~9gQEaHeq<#-~W-D
z{TegX=Vt`{L~vYUPFl=)k4LwWM=vWG?Dl%Fy1+N($KMODp2P)=b8QdVu-^wTHBZto
z>;tBzy!xOs9!ywWoe-urK&HWQ9ZilAL?v|@Q!_eECggjCnJRwQ>V#wscK`KVoe)_K
z%j(SQ$b(Ac!8?~Ry+>q}Vr+oyX41s}6eh+`#uHLrDo=j4h77Qe4B#zvz$0Aj0($&P
zdaONy9_z_;qSxUW`PE5$Lf>ysk-mS~X9^9Id8U%3MQf`2{2KCsK)jG6-x^_xIy=aC
ziDbN0mC-+N`P#MPeK*nOEYYe~GYK9K_3flb8Pdj<yt_c&Ig!@9q_xBsA({dmGFumH
zZGkm;hpk&hoKojP@X;pn5vxBAMx9K={08PaNHpxPX_t};`q_$poM{eOO{?2P;Y?$n
zU_YeJ;Y<rot~PyvrYvNc=DcXiOSa<gLo$XRdC(QdQ*dYX!N+(oeDy)<U>=W!UkCB{
z_r2ut<va<E`IPcxf!(j0Z76C42fq&_heS&n4%O*>G1Vp8&_(vOo(~&fN2K7dqVNF^
zQ3ZU_wE9KJlP{)U$n^J<FKIB@dUi7T>9aT!ell`A4@MraB7)LHFr>aDv+u7!v8(vr
zn5;g2JDN``knuGyVtlLB@gI`MiD(qq;_DY=dK_u!{p8aCb6T(UAIsvG-hLYXTc$76
z&B)6~PXNQys(Ud7dRD^;PYOr(HQusl+2FvLqzdpb8^_yF<-ijEgpbipMr}&&IIO}l
z)>pN_u4*=^a(k#ZaUyf;+Nm?`W7es91?E8U<u4q!U}`L5#qYu$R|aPLeK=k8gEPzo
zmQNL!{<}#($^KD841mO!fxmNeb8z^pfW5!J-`wM6)%e7OnVDI|tmKgfi~dSLl-v%}
zd$qR$Pg=lqeG3hhCBTK7J0v>#u*=2k*A2F+?~i%%PQ)aVU>e@*VfdVzo7)8-;XB^a
zyBXHS?wb<O)btS@3o0NpnN`#*at&10y5>9o>9aByeP3nR73vRPzkXG&jy5=HH5@>&
z|He_#Yw$FY13Lhi_GGIlD=S~S_Eq<ZA(MOFg9i^5n_F(><>k5YT>O61oq>r-MMXtD
z+t)3>!;PYS<;v~VIr;eb2n!3tbMbWb^l-DO<}d=RdCKewTcLlEzxk`;3*H4iYECWH
zWZl`igqVQ8!<m;VHKn!H)!+Rxd9+qchkB%YErnyDQ_1OwRr4SnbcOXP5m4Vt%v9#`
z<eckdAO<&ynCN))_A5UQ_W7keCHfwQyj*;nR0C1D0-Bexu|Ln#p8H)SO)V^LrxGA0
z)H}caBs~!`Ub?U_e%tZ3-c(^umwJ9tpRM9Y+1$U_9r({ac#XA>ElP^BYA)X-^HjC0
zKQZ?6qj~8+f2)b~3I>)RcP2K%tL;G#)Q6f_2@Y>)hT1687gGu+kxw07yu)td-5$K#
zy82lF#Qt;%#C5gx%|w0=o<oW=@{~Sq(%bd%5GANmv@n_CmRF)Z5$QErG|(og(`3<a
z+b6!5p5yUY7fLK1$6iBaW-G%~?zIKH<l8=2h#RhCdgbzHafOWc*#{H-N_In*8*@+~
z4$SM=g$|0pBI!XKJ1gW)c9t2d(B~(^CVmy7dA}OM*1tw$U(#6rG@7UpbM=0RJV?BY
z^RbUYH2--QBQnERWMnn6F`cX08#&uuj8?D6!v6Yy3;)SJg46R#hZ-URSzz$Y54PuB
zG&_&I@DkD^gYeB&ggmxtrY5~@TGjAPm~$90QeL)GavgiLms0;%_n%c^&qp6W!e?i(
z1NyXN^veffbhCu`1sBQIA0&p+OYo&X@m-PH%uEj+REY185?%dx&zSXcaL^0>mBVB`
z>&gCwy_9rN8&0p|B%d&lPY%Ki|L3H*!W|7X@|-{U{8|&gfh_)WrGQ6xzm-(irTnjU
zdz`0<_kZ8T{Q8%2fEYwmif47(K19fMw8vn@>;D?BZVDW-bK<JirH652fnk1MMZOSH
z&nZt9C%vB}y?2wHVkA>2rqhHs;}9py^FM^;&AOnA$yIwuk45&acJ>c8x|~?T8~HfV
z&w)76Y#J=1;2+Y52kDKI^pGfrsdbaawa3tS*BR1;1x<9A$*>Wmi5O|JwJYm8hV3Ay
zV6QkFkTibh5WYK3zFV1w_1}xR>d=u!O2WJzETMS#5qX$+E{+beBo|o)i?+|Mq;o+$
z_ddrxe`00Q+%PYeJo1P<!b+s&JJAGleXJn%G3Nc(G|w02x_}u&Nl`Iq#(f(-kEvcK
zV4(LY>5X-@H?k%VtjU=M4nxY4;MtRm<Ut~(y;9^s0rKD{B41Uj4-&abB(9-e5)Td$
zBIj=j#?ogeUt7XAN!bzMh}Xkp>WM3T-vm(cH;623$QyIDBmhgFtSMgd`u%m~<KyFQ
z+quCLnB_d#x-b}5ikQN_-XH{_=2!bc7L)X`$qFR_^+ZDia%sb<_p%(9N=^=Z544)D
z7NEQrhpPYO*R^%lPPkk3=E!YE0j=HAz2#quR;i1?h3N(p&)p|tL1TD04LZy;i5QfB
zrju4wR7^5%l2{zg*(c|_IAr3)a`euh<%N+J#tKaR1X-E#S8ZvhBEBSEt@PRKb+qX{
zL3V7QHyTW`GOR|L&AR{Wl6181V4rQa8xf!D^o~&-94``Z>U?=_IDMi73v!$*(vA!b
zP44nlzkBy?{T)XrJA-GBBn-@Z2gJjYPRomRae7w$l?Iiemzq*7HggUljH=PJ(&S})
zcg(2%w+|1?TrI*X2{zAQi|Xt7>`wN^#>Uhil9Ok%Ru;0{pFKld`AN6PH=(QtNc73L
zc98EvBzAcE;HPzq&vzY_rgmj_v|bADvYYHJ%W_(=7L&ld&q_=OuPo|Ox4nNPVU<(`
zYD_D`_Awof@BPx8m4OB|Ik@2Gj;AX>o^eMPY_|H3hVcKrI3fB{<I>T#9Pkd)wYNRz
zS-_*6V|M`YVGt5{S)3QR-XT)RIQz(dUC@m3T>DPV$;BTOB@m-KH{Ru8Ke+AO?y?(l
z;HSiA8sk8};Ln8fIvUx&&oI73mG>6(QnG9&y6JvaTfG}0zV&92FQ0;ap!V{(ADF4-
zPH!b_D9wpZuWftJ$|P2Oj<p_Z7adN%Jup<GVWx2D(xr*BN-qNflvb!17#X{JdP3lK
z{k*-myUAGAUeBK9%2fS+Q*h5qSw~05?L`6x9L!6T5n4bI7M=#axvB8R`5j~CrLTPY
z9U1i|nT_k{gY1P~`#q-n$>#8&bkWjS(KuCaK53Ql5VXij_;u^ZJ!&0u=csSrHgWD?
zWIR?I^t1wzHl_@WVpJc0(6na`93sDVru{(NJg1ArKy}9G&&}#4>?a=R)aRJBOW4P&
z*00beuB$%0D@Ik`S?QIq{AExAg!_4$_M}Q#nYKwK?T#`v=jBwh;S{EG&qTl2eKi_2
z<2qlaUr9KjV|&L`eu1<ESJBh_Zds%Nyq<i(@v)cZ*iMY3Px$g{&J&TD=q{EU3Z+mC
z=sZ=#moE;L=W)7Ry!zVi$IRZqfY0molLY6t*c%x1fi=R!D{t7{lB<c-d^dIX@#Du(
zPW6-rT{w^s`M2eerVXN#12rQLsn}}g*M*0N*F>n-e82tW!l~64^-YZBHn2@U67mJg
zVk^rr!oF60k@J(~qW8?NA;jtRSmpEQ&+F>yV)F_6nv#+NP{ir-?Af!3l_0q*T-En6
zYC=K+0YNwy?#M?)e}w<cHB6+Klx*U3QT^UtdGz;#32A>v0|OaTQ&TSbx_Ob4{(I=%
zz+Ucixu~fb_)FNFO22?^7i;VGmV?#buUFhyK;VCxT6846NvpzxMWiH1JS{qU)ax{j
z!Dyw^V!o4n;7MUqyYvm1MWW&qrwh0!yhN;KUxzN70nXd62~IheSuKOn9eB;E<$8`z
zWp%Y$wr`kSOWG<G2DX^1`NCPVs*Dlwbv1{rN*Yq46qUsI5dl<XOjGZg(|x6STKgy8
zl^LVMY@X_;IX($|P<ox6F;%U;H8E>xx`ERrW4hir!QlG!y4slAuz<fJOpbY5_t;~*
zp6RUzl!(!ln^+l_Z;yex)9m@Q&i>;jW2|m!L?sD~FYHQBR~M_@pyk4)RI3rK%a<We
zK~LwWmO9v1s6C!I0T}rm@NkE#>b)U&GTW(bvvzMyR10>-X`!KIKfhXdXy{H+OGB`{
zNAh=jTdU~&#K8wU&i8?x*}r_0-Yww%&yAceP~hS~QIVLK*w}HH?Enb}+f%20oaYX#
z><->ak=i7KC_mWx_+>3d>CnEqwDwNUo9au+lz7QqOKh`rKYdiRY^JJIl!7IerfRfX
zz-vj&wmt%X)MVANv^Y^3mDK<~N&WhDo8gomaKo8ye0qTv)p#RnHQe}e?>R;W28-cF
zo37tJkP^fKd(S?vwW!go7^|aDxb|JXqItaU!~aI7C-$-xO=|a0WPRjUUSu^Gr#0CQ
zM9^Z$fvBuZFkt27<)2=h?EfBDa^dO^kv$?JGxGycPSbUI+O~aP7+}k^vm8e-A-!OS
z5>V1czVLH8sn%jEDIvFcd>_-b&#l+ZDS5`kFf#gxqAvnM-IWe3rPJsmcA?DD4o}I<
zsb8wR$MpZcKlNxsLWu^2VsvPv%OGW>goX(7U}eR<w4qnSZWR41*dIU*dN59bk`rCI
z9pfEY*W*G&$Jd$@_#pLl!QP;;Oqq(+AK~L_n9lLZNSS%Y$6V<jDO=)d-wv34F-;rR
zGA$2-F+NFRV~*1VWh_Tn!+^Q2<uD=+`#DUmb1CuXf-lRDQ<w>?5k!DXNhBeq4gmv(
z-th5Wr{J%n#mU&e4m1F?Iu8{h?*fDZI;AlOL)5Yn{3(b?lVVAtb6J892E9nzJNR0E
z6PZCEdig6udZ~x(z3(Ga5CRI3G7EjVjv~7bO@#c8V){mrrwke|k{;d=gi(#C)dI13
z_6q`ej6gY1_(&if&x%LlS06cxuD!MpyrF=uIZoW|{0~V}UiECjcs#%fQisUSL;%12
zhil_WNU_Ss{8b~ZrOCv?fpw(rl3^mG@L<^2vosWrJTh<QfQMLCSgFsiE`$#p%4G6a
z=W!6bu)bBkheMY<XG^}QB;AA)M0J^T6G4EbtXiG#&+}w6G|B8<jbrQ7$O8Du*1ZCx
zb50~-etZBnvHm$(xHNfME`p&I7FNB_2OEsooKmm)d+lQR7rAquY)e&9jtO=%!!fG7
zA|fI#7uD58Z6|xrxHEp~N7~~p8QVxec+E1hD1$9oD_cbxM+q5R^-UQX<2E#Az4(O1
zxYi_V>8vMP;H!0OD4&BdKso37HOov{vw5MHK0zI&sgccTY5nnNK(YZ@AerY`i{nM!
zl8h87+cS5+xs$!4DcmqSH_^jg5;+5}kjJEL4^0WO?GH=UyB$*tM0XfsPxmv~w4)g)
zjEk0v;-yHoEm&vWB41J?Ttb?8-#Btd4pZUUO}4H`6^Jj-b}Wo!IT;xZA>$O#*@mDD
z^cBV6wc-Edf3lVbvOq@1WjFvt@~3pZqWcNE?!Eft89dqNW9Y1%Zf#247M(+L&ps*B
zVc`iK09(fD_>5(=ev4^B!vz93&i6z6SXw8>$EB$UYNAZ+lbn_(f;nBDvc#{YTtARG
z?!Cy(wrpk4=$AB>c#5KLgw?`M3%HE7y9*r4Z20r`=xJ&XjXMzbx9#f3RGO?p$S(5N
zQ6%kWN9lg%&`OHw{(e0)G*K=mAU^r`G^dM+<HWo3Y%~y8WfkQXD?R(F-$Ve+<&I@*
zTh{4j*iLQY+`=xg_$@qAEz{m+Ne@^H$q`{BHo;>6vS~U@Rjqa~l-n^@=vl?Dxtkos
znpay#;YF`}mG6T&x3HL6S!!7U1%A&{d|Fn_qVm$YA@J>B*RRLjQQ)Fa8OdBowH)%F
z%!XcZ!5Xy<i4dMRe5b81SprWTg<bl1r`PH3(qPG3xlePq4iVU#JFTMaFy*Zzv5|8(
z5*0wt$8lP#BGC}4&Fp8^Z;;?HtvDh^@$!D9Qm5;u=?KTNJnAaBSP{HD38*cgkvxj_
zb@!q&=BEa%YHzx3WfQAi>RIE=ORt_7A1TCd*(h>Ln|Wt-q{4_S#mk=(99^SjrwLPh
z(Eb!cY$g)c6rS7u#k6fRRhNB7Cu|>u^K&P?EeNnrhH{9T1}=2$$tk%CO?)C-ylWP4
z3|U<hq(3419wRp0Z}q-*K1GPi1Rf(`^1k_y|LAoCjh@DI1e=uL3#XjfMT{E@UhWkE
z7<wJJ-Wnz}uVb!MoK-!Bml&D7kQt;ePhwwQKVM7nI%^}KRODb~NzIviOE)A4OrW#8
zx1sY2bd*8*>Df;$3(YemcERbL`P8gJOVN2d|HZT`KUg)cN2@Q8ijwFH(Ux>_q`rxi
zWY%1<$>k&971YDEnN>_X{laNZ@@7m>vXAYvOSDHc#OZ<*Cm%x&xUvNFqr5~Dx;U;<
zA-;`bs2;l}6Jg3KF?Mwe$b&=)^l^21mkH&6Y61Rp3<au?WFUsebjaq!tUhVEN4V-t
zuz-hqPv=*=1C-iGXei78^vn#w2@n%Mn`em5e$d<qWt<JY2Neb{T)3cRKbH4!9o37S
zN7{qzhEEJ9cT7W>&l<}(;qdp_$WWX+5hZdAaIv(WrnW+k7vaoqY3m7SHnO9?tQcF@
zIL8wkeeA-q>=vjfQfId>oCTu+E}2qr8PjZ8S)So^X>4p9h)g3Hq5;61y)~ulD4+cx
zsj(`tYaD<4yPng9ePuTP(>YJYJjB@%2y7{JykV3q#7Q{+$8?$4)vx-9xvy=*sT0{a
zCfEO92J9h#skADRanHIj>iRM3A7-WGki+*ew8~w$@Iq|7fbJ*K=x;vSfF%)ts+<_f
z-2N{aS~XTAAPx2aN#MzR$>lLv&NB)M3K#CIrO0v;15aDnoI4C7usG)Fw5kC05c`$o
z1<w=RZ`j0bg<@9_y1o+ILy@Y$F1>G(J(eB(;P*ibPbOYfujGRcd*M>v7t7RqNBTf2
z?WLhJ35+guBi#`b%lFuu4H5S0$1TaGw^u)29aR#e!g+OI74OC2ikzpr0SX=wl3Qwl
zFa|>x6d#o*OW1{2u>llyXHYpV&W^cJ0pD2Oym@npPr<zHG1IYofQw(<`P9{jeX{&!
zvu63DJO2f-<6W<>(67lfmfH?zDaAv)_(C_7Z<R=vva3lyIvRd0;)2)#K0XDp+IWL2
z&$!o8^!H;s%>Q|b<nr(?K}}$PE(NEbo_V*xCA(a_a^(u>%m1>@;orWMi_0edE?$}x
zxd8(A#EBDQ*6D!X|7C2M-SecWp~ltGx(MYE&s$hn*e$!E2!P$`!1D9}0-K9;8P+nr
zG55Y{r48$uifZRn?{Lq<!*~+W#x=L#JaEk(_f%&VC|+88DQ->K57*7KIIUE_<=&o$
z5O675aq$<ObZ7+Juljywk6#_8;Wf9=W{isW-kU8e)ZQ<tN0Abzm%MQI-n|C1S^Ke4
z&X>)%WA1lb-AfyxzdRGQZ%p8D{DQEdLX7Gkzx(c0foEf%Na@!p3>pLG(=fEcu)L0B
zuSJ3vuT^D*w~hbfHH6?})22<wj2qveyhPc+zyRs${0<BTJc;N#(Xj3nsXIZC*dDKh
zl8o-+cNOA_*Dic-!@GuoeOpeLHS;evY<9~h)+M1RlJiST6x0-cQkr_mdaJrRLJ2PV
zF|*IQAItRHGr~bN8)%RM0&`a8&{Bm#inAdyDb6&6-HN?z-c<jy5PH}v1wrK4>|=D$
ze$n%#AUR78fF>`ayLX=j*=j)7SUmt+bbPGSM*)i2`8RUrPD~U%31SdfWoZ90`dcy0
zyTErOynp{*H(wOhBI?r*DHKkZbd_b_iD6Mr7dm&Jtr8iD3;caS)RcNzxwVTf8@KP5
zfAcq%P-0T<T_wTq{s$8)eea92vW3e?Bvzk8G}o7}*vaz2)BH!!%9Edc10G7QEKYbz
zF3-O9TuX@-#|qx>TeB>uJzsvJC%}9B64*R2R+MBHHcEjxzU6QsYO#xaW>07;L1I}#
z3DVp}FlQy<Vc#A|91JEsWp+0`UPe?x!?|Vm@o0-DYbexh73In2;Dx0eiZ_okX<6Uo
zDPBVvJ&1jZ=Hvf>{?3DOeZ@~&zaabce8U&hoJ?#qr615#>MKAa#RC}J`8jlr!c2xE
zj5Nhk7LzF-g!Vct0IWD05p9mYdy&~{oQ`y^Kn_8IbX)k~@{}~Y{pb(m(L2VhJJJFX
zsD+)}L-A@N^D_I)kzM-cNC$lJbbFrryVwbu?Glu~r(hzS@0>rDEdaJFx#I=GKiNVd
zaEEYV6n|iVm5JaLv6s!RTUji}@|Z?adV<PAXR6k%(xMbTFGFXip@bS~YQPGcb~kR~
zwN_a{XW#SvhBjhs+VT3AsM~;*6w3gy)4+~X)DFQYgXe+3(@BGjyDWq@j<9~tTwDwU
ziJ>xK$Et(4LcUQX-*m!zlp^w?`$_71P$IO$;~z>q;ng|L7d9VR_Ek4NfdH}xUfz1N
z6`g;9pL!n%(?ltho93T%T3Kj8CDVxqodsy@Fy^i;)*t$5etIYkF6C^Z*=c|9CR}h<
z9BE`jUjN@_7^N{^5v4lJPoAA-YF+dGH0d(zm7%jzBP{GH`zzIt$pJ3Bd3EBy1RUk_
z)mV|3b(C71u@n!gpQJw<pth$%-0(_-R>sLA6SHF-s6g}WYtFFaqMs{Na)e6PE5?#o
zfaDKu+yn2W4(O+-SaQ=fC6Jr%Wh}`KP5eDgn=$^I2jw$r*j7+!C@iT-GHKcCDK?(g
zvSh6IB5MZ{mLyGV1p08<FLUzCSxXe~7#)+ZXfn-bt<iRJz>Z&w)Ob65z2HJvVHFBZ
z*o5z1eZ)44FgJa&T#k5X0jIg5cgCrGl~<G$3S%hF@|cwVI@s4mqCCR~FMo=UpF~tk
z!-?XW&vuN0nw!=x5?IR)oadI0j9We83g8BBZ|^v8bTiF%Cg5{O1ntFnI>>&!kn{8t
zE~n!Zj@vj~{YY7^a=2}CrBjyJcc9G`K;KQAuYG+<&5WPMgY!$bQowNm;hUk-1r<V>
zy&v6Q9B+RLwm`ga@7h%z{X2{zJiI{tosAu0QJYp&-1{nH^U{dpOd5PVsHlTizG~QX
z`E0%0hB7MwL^2`4jwb~KOn{M59HTIPpCa`H{!}MONn(!6^wA7=koBzry<mhgxp8uS
z{1FYc2!Uj|1d&|?pwI;FAu|H|1m**5m^EA`|Mgg%cT~108X`dnYS55vyzi%Jc>=6y
zz6XoS_3PJ@a%QqtmSnjefcbExyB4A5b9@*=%`9D;ZezCB>Vq~+{L^iFQP<W2#;qh!
z6mOgW3d!zO+a(%~cdGEZM5DY15^LYZqumFZ%G7iW6F82D1H6i^-MPASn~5mLsqi)H
zHVlH44ivSL=zaevb7g5nKsEY?-GU^_>bQgL#{s9*E?y+)da1@~spgo<`@OQ$TNbrf
z=4<GFu4mFXe7`k&Xz1vTY-e}=)Lv3(Hc=`Wj3~=l9`u2}_A!%Y`^mC%S>UhSR=$7#
zZa;$)L6iq1<XQ1mf#vWGsinaD7px>QMX_40ldqp&lt#+<2=oyFAt|pT7KM+kBP=8+
zwe{c4V6v_x=`JrKpe)Vmcp-{4Xlpvob)8oXqK`rm68MwqNKN&meWtnRz4vS4G=U<`
z9xb(i&jGc<r=IxTxCo~*Q6f6h2!R)Ayq^OYKQN-9sYxiX<Yjl>>#?2q*1q*2)$(k<
z<kv;BLi@s?hf3cUXjWKvBTTXo1wN8*<`IBw)_a3%%Ea=P7G_p**q0B3M#rh_A1>6q
zbg8Vh+Xnt=C&*j{g<|ZbEhsHXoHW@35rBBQpZ~7kta!Z51@cGt6J=kFp81H}S*3fc
zBh<y%Jo7sm<XbJi%D)xVDeyv>D(ciU9~AC`x{P2GtqS+~;S-3orfD7n6jkxg`0Lr(
z4o}$PM9ZQ2_V)Ib<&hP~-Vd%M+(y16Mz1u8(}kQKY^D6tl1ik84R&GQzin^zf^70x
zt$<|rs%Tm$`|)eGkIe$xGixL)EsZ!yzHCmj972HQr}*Y@8wvk`h>CvyE`h!i0(bG~
z@B7#zOjw~M!VUX#oy%&uaq(@%c28uojBeM*8+3JbHP}rfUqZ(4wpk+V8jtqe@7D*f
zkkokXgYxs-TiGS-%RiqdTt3QDr?E^rX)wThydmN%kb?&Bfb^o;P^3?1JyOShu>b^*
z;z?+5i_tIzNQLnH)F0H7j1lFAiQcvaJ4n7{6|)}o7rqw>@)S}aeBjT$_Rg?zWR#W%
zv{yE9$_I(f-P2xByVU_Eqx5*k^W~)lm21}&{e}8oUr`KeAoX+iY9fmZ3vE!AIAHs`
z1kcC`WK8C6SJH^S@uL)+fQiF^x?2%rcS)dKsrZZw9YWWcKSxUk2{*YEmD26c6Pes@
zm9?O`N%E!P%fVf<x!TvX5UBsMk{kS+W;H^H2c6dl)_g8>I424MZ@VXsfTce_J_VYD
zC`9U(fctQ<+-$rW$%n^Up>3zzb;gPLN}UWEg&R)C964P;u*9ri2aewqYQpJ)m^x%@
zq!ruo`>hEzS}fAm5FB<Cy80>v9df%j1$X)aC}_Tc2o*~B=AjzlD>MHh`$hDF)1^ot
zEB#cFHC!Bt(Ou}^m4{OYB6EKc_8=l6Vk;-=cR%oFxYB(r>xG@N&0a(GdD^US6XLQc
zRJC%tAY@B=xn%``&A#FMi!<E2DDPSIFEXnTHzWX7H8f)4W>Z?-KqRkXW9I-Q#E?$=
zG&N9tQj=u^SZ;O3f9b898<0riPapjm5gdU^wfVEA3ISrumT*RN(m=T|D^0Couy~Ca
z&5dUVeGqJGuE!Gg{8@x{KY$C;b4$`0O8s4a+BBT@Lpbhozs$-Tu-BJsbeliGMLR8a
zuv-n)Cw%;PYqa%8$1jh2_gP6kitIh4M)~U1&hBo6v3m-0)==I`qt-@ZJGbI<lNPX4
zuQ~^2h8nB}Jj(QQCJBk*N=NcgMwtafvx<dl>_KuE|Dd>C=_16{&6{o2rIsi{d1=AK
zr&ie-wqG&m5WI~ooCS5;D>?qT(v`(h7USeL5aD5=p|#y5Ze9I|YNNC=5>rmME78Kv
zY~(a-OdeZxvxXy0bosl@<PS+a;`JW1YX$A+04dzUdeR(x9=PvWfQVRxl{0-6+%D0`
zfdB%bu9`olhqwWw<=bm1u}TUGWjlaK(tM7`G$z9&$-Gm$mwx5=w$|-G?!JS!p6GsO
zV$9n5PS9o(NIyqdoaN@3%q0D?k6?_7{e^j4yn=(BVDD+cPC>c&4;~C_I5BpfPdhVx
z?CyN7=FC9kbW8KQX_j=o507MjXy1CWWtX7Mcqg3IV>*~T5v&S{NtCa$JHk|M@5=zM
z4ri;=k)-+$CHCw`q2XXT#MSQu?~2{rkMh#}!-s=|gIlI=cXoEV&?RPiYhT%~6cObx
zbdOZ018>5Rr{@OKqVQh}R)d@*40Z4^BekX=xVQ6Hw4xIe6H`-Djg^yTb}{VOv4eqO
zpSYv%`ROh8KjCt0e;*7H8_iw^7E3kx{>^sJeG4SnurMqNmp5YHBkJa)MUfH(B-dx}
zp9xQrg3%`rpw2tjX11~(hF<c&RI;IoE6U(q>FAltlim(Rs5&|>qZl0}sf^OrsJkY4
z1URgvJ|!o$m|J<f^a73`3yjiY)Q}VJ3)i&s@QrMwZY{09!_kY9>ZP%WC_!0Oq~3rA
zU(FiEUjF^Sf|v1syj%L!3=WFdLkSgTtHzBmDc_@u-zD*>MspPUp}u?M83PafNL_P1
zXy3y)0YWwON}D_*ZUx2~DLNhj0f7St^hXVmYJyHNB~HAUd9C06Y~QkV8DI`FBDuaG
zHlS70X+T@QK>vX3QVzdH`~^FTps6`ctENHk(s|}5*=S`TK6jyWlR-=gG-RAMl!!gT
z%8cG^-*4>PQWjrbyanhu<fcL64_DM?I`2RKI-?4BXs-8@=l;1^w||p}eyRswMmbgn
zKsf}iDz+$ndsI3IVVPi+bym$bJ=jlaYW0>Nab^GHTkEK31}j2J;WyHzcq*G5vxXD(
zU0E$vsoTIVz7^E{UA=&74c}liVQWZs{Z>&-J{v6L@?6hK%gTb2`iN77f{5@^g@FiR
zS<ZhPsAg1jM1`E!8RX7_Le<|LH@(%q)W31e+8%jyxA4E-W0ETqlA=TLSE$=!i})8X
z_kKx>oUI~UC5lcREG_z@m)s`)cr27yWqKdiUWyA|`hLnj@Pt`6NFHfwhlS~;I}a4k
znyCx|$#3O^wcD(&{p9e?pDwwNQX2|r*~Df^|EH;$H17X^|GDba19QBYKlui%)l1n<
zB*v5s7a>Ed90;jQ$Npb*S|M2dF_V^T|CM6J;xw1c;ZjM$<a{i>@y?0vZtk&rTh=(s
z&miaE3rQ|uE7PAu4j(AJAL%bgxPc{NjR<*%OAUy)#Soj9>BlD~9(!9Oz{j@)+cN%%
zeTi8AlY`MY#7gJTh#CUmL3cYppZAug##wxK|I`*IS_-Q(T*w_f*pOs`d@>P)VwIA@
z!d;@4gMUAqCWvG(CSS|E?2Mb)IMgl~3b$R{NTkX9%I^{3>PCjN&R23lNH>D7Al!Kw
zS51c93~pmcj^%E5ny(0!rY87?ur;|Q*pg~>6IQrBIRDIL6pat4w=e^Ajx=ZBfhAHK
zu<9f)eL6c7S1RGQv;e{xa^%4->DZiZa}EqPaw0=}UmU-MOk%0T!XT+N02cxkw8H~o
znh&?sq_4CvX_{YZ|8X!R;1%_#g}Ogta_2!578R%w7k1N?HSs~$@4ku<1d18UR}f2N
z%=d?T+b{7%r3*?*uAIZ28x%{J1Mt^4^#jiY5xLs^K)T)DLX1tRabyT!QDLv9CW2^O
zr%{hg?f-`5BgG1(iyRJvF5*@Ri9^oQWGNJ@#avN5(dB<HGNte9J6JvffP}5XNDv75
z4qvgH0qhq8{ei-j;qk7bkt{fdk*uuYFnu9gaI(UxIq2G9ypWqj)VK|0g8Xyp(E9c3
z$*$r!PpUavZ4{^c4K4Ni<wFb3P{$jQDv8U;N={L+Ue#FaON!#xOa(M(!%Y|2LQej)
z!C0vK5TO@9D8Gc&#&&F15*s7Y&iCta%sWzOnT2{u&Ixw}CMG6uqoEGM@=JT}8BgS!
zqk1J@{Iu#3%mUWp1(&hZN+oFxA%_!r^{OoivcUo&A|eY|y8A%^=a(;EAjI?SOj*ax
zd;bEKXJXsQZK3Fuv3zz(3_t*nyY}KOP+mPHfwaiKkiY<XGl$Bbh8sO7QGi#T$F_;~
zefCE3#SdX(khmS$(m(+6zL4N%=3vJs?uHY?|2D+oszL4}Vi{X5A@XVb-chx*xnK|C
zQRC`EUoN8lr2Q)s%Q3^y@89K%w$z~pDD?e%j=~pSFTTge3lSP|ELFhCKRiA@9>c7w
z4)`ns8V_@${(#siMbcDOcJ*??yK&zlqpY7rPDM^rBZL+++BF^V!m?rR)vH%%NGC0L
zul9$}>(?@SUQP|vgxGm~-{Em~yU0xBw{I6Zn1>JR-FXSl8&tzQ>T3#$E%sgVR*iXH
zo-WtBeac}<_cjL=#h+AzmrUaDRid=|g7&<_1Fr_fY3WB!>vsmSY-D$sEc@gsQI7*M
zD>=bUVr-++;K6iWBPYP1qOr}NuC?a-b+bGM6?B3+Zs+fWw-QufY05QT<>GyKR1%93
zD1Ih<%AbnCqod^Jh#PlM?Ty>G9>y}~g@6zRF^}mIRUS}NbO@W}_0pe}<U3U&)T(ZD
zRI4ZU0V){`qnBfZDHBwVTW4PyB#sbDb0(N@8Pp$^eT~@nbN9fWvI^;6Gy}+v|B|kv
z_1k79kWTt(-tP^;S`|iqBtc~Vt!_;wR*PFRVGE@)&QV$GkzF7RlJ`D4ruh+v5xawL
zk%|@|6KO*(4Z5vPDD~CUqfRCkj>PM`gcWGj+s<ukoOd8;L8;@YqYz2IGXw(K8$7-j
zD)%WIax&!hqnE^(bsK7H*GO5Cu)4F*J87m8riu|^4O|DmR{C*f4;Y**qOIOyyD47M
zHc{!XkE)QJcEx~zDD|^6k_&aENYXAjt;{%)=nc2tGA9B@R_UsJgxj+;n3C&-N|~hg
zeci5joYUnnTs3lq)e?Is(W@lcKzyEE>)w|IaF#66swkubK~UfNlp?j4v_E!Xntp!&
z$yVoe^t6^RkBKMH&PZ2%6T8zzctg!qHhP;x6>x#Mr&NB`S+EqbIuP6B<Ya`LMrJZ#
z!X)@S)LBVyi3aWzZB0!FWO!WZK(RG!a9#Hw+o}_%Bk;iqu#MT~7@H?5`vwVZg3D6U
z)TGAk6`oTYocLlILMXpMkZh8;x#*v)2YCT`_m=PuzIES(jDvC2gd(J(t|3h*5_tPL
z6p_GF89f4W-zp=i@eR@vs|i94p-%_va0Oh5e!R!9FBL_IPCMgUYvDPPy(xHr7zW7b
zvksE6%CC*WQdfqfb<GGBd%eWXl%k(?_wIvW>jRnn+9-ac=wwy?GYi-ZTe^ZMihzTq
zp^$(FzX_T{JO_Dxy_=e%{|yv)snR(=ei}l{YRY17>3+W2C#NVjmb-p4BSOcS5$%;Q
zf_p)CAURtLR3>8Do=Xbj)<<b;X-VK8Z;$F3*GAtAc9<r_rV=oRDsOde-F~GYF&H)|
z&JgB+%w0aJ@>sTPqzqkde(vP=OPNOe3Ylw~&N_Q!7Bp=~kPAj|*DomwjsZ#ut38Pr
ziS49$FRZe^ok$umoyV@DcgEULfQ4`WfGxuHPXheFoRIgk`I*T26QZTPbA=dsU^7Ck
zlXM#if2Qm7MDA^&6!51^^mJ#<<CaZHYn%G~;C79z*h&66xn4rZ$k?(A->uAomp)o5
z7PlNnyannE9jD`hAJdVGk^r`GDN+9im6yqI4!Dph0zLq4b&_CO{7(|R18~zR>NZ5x
zHt-w%q8lkX?zOQ_)u?W#J-*+v(~o3Gz1HH?(N4(aJ}8N%`#D=roM<>6%<i5C1}F8y
zY1b!o1b^Jgm<0<3>289E0Y6~J`COubR<yy8#;6(sMf0dc8g||W&pHQ`21-F2bd=lj
z914A~)g)uj=1GXQVt>RoXI2v_o@a57C48>GW_2RWm<Zyu$8=9<*_&rcW<IPTrSHz4
zH`}C-BdQqC=4$~hijGm12yr+z9<!N=e+wre{R~Ep=TNcqZTLDx-{g1V4vFO(pAF@R
z<PIq!8wzN)dX_?&>U$oNh^^Rm43!~|>1f~rQ7Wj8a!w^=!@vc@r&ZmsBju$(G{@HH
zyCR?SRg6tD?Ii4WZ!NO9FJ!sK!F)xpXIEEOCvlOlgJE1F&=ur_&^T%fQW4+2e(i>@
zNp5?>Qr~?a$wy;iJY#hViQ|{tD0#2?JDoJlmj6`y|GO}>x9n3=$`Z&6)WHIzsY@mR
zo)~HEjYc45Jkhkq=>72c!+?N*|GAdR%jhn%$4#>F@7kB8B~90J$Ul)}Tbn4~l+>ax
zba$&22V1F~@rB7?Ice&##l`-kS>6@8A)|H7-QC^CW+Y;{Djit<6c)xj;qz7%SuMsB
z_&aj+V6?7WXvD@dc|U3%d?K(Fs2#-ImjqgK2$u<pIU8;(@-mWgVUS{G^L-&81d<YH
z{S5g@xk|n#yAxPHQltQ5)7&kAbtVHuo~F)cvonm-vPa0(78f@&MXsi}M8iLKE0Pjr
z0&6*4a_z=h5P-G;0;Nt(6CDvv&wZbOTGB68&W3)|E05~<>D^RwbUJ&;E5z>E;GL~j
zq4RTk_wHmdNK?nCHe2!Y*3Ot^o#t{hfAyTTy5;Y2$#bt>dwb(biGIbeS{{MC?&6K_
zjy<7k7xde-{(sd1*w~CYEL3d~JJx(V$y~|T4Vi%xcPm2)DFluo!s<2HzeTXLupqVr
ztRMOCPeX*WD?ex*NBy!Eza)7246+BGeZF&>d!7Wg0p6#LutL?Mpwv20u)7RDhkPcv
zKzb*`8eAszhv?|&AJa1a5uu^_^KxKah%8AWGDhB)@QnhFWA4TNohxxxJ*JnGm)Ntj
ztfqGzJH>cJrZHQ1N7B{CnXOEJ85NEyTsfK$BwuWJq=?bXM&6EXUm35ksY3ZbJZE;~
zUA?+vo5m@bQ^#&jr({V+X4&kQ`e!&h@L}mf6?@}Rz2UhB(qp1MjT6|1Xx*>5X1_jV
znQ6;OOP}FW+r&akarU5VlJhgt;8r}Ir#m8(+iU)j$M3VGA)V%N9z(;f!#lTV9MbVI
z_8DKAX<=eyR77=2YLf5*Xbte(Z9_vD2nX&m|10UMo08DE>DJCWFcxKHWo~Y6RZI-7
zBb)4R+-Qx}Ekeb*J+u#KJswG^)2=ig%)+mVx`2T*?J+~?wKWrnzqfAL^1ZS+FdzUS
zLFHWwH%sH8jCOXH{Jgx!uC97|dOSOx!9vlni6#Fr9v&V>^ugC;EcjN4SwVMg60@7k
zPa}uQfhlX_th>l0tMW!x=8NAr&+m71OHM3rcl7Vh2Pw$F@7y9yb;}QcQhxv5J&%4k
zpRoVdAo5l_(ezA5ZENQIbl@aoR$}sr%DV~EnddGY5PbW*Va<yZ-vx)2ViZHRpQKC6
zJ-j7pr9ek;<8jwbkEKID*c{~d9juks9c~b2-j_90lUAkY@UBvxm&78VPI1TEa$(v?
zNvRaHP$?i$XTi&E8bp!L-P}@~R+dSg7<=;XziZjqgCjN5GC>CW&dIG$yIzltLkYz(
zARPKLKadq7ackIBe;=Q&!9jT$+YCprF5-66lM@pi^PfXPlF$w6<_GKIeCryCX2A!N
zH~}1TM5UNUs`<fv`_6p19Cr0vK)~*0rzi;|Q4nKDJm{2FQ}h4y=~GtL3Qj@Xa!_@m
zgS*I20LZ5h>`JrKibIhI>8T;nNEAs&_+htRL=*nUP1}L1H8nKw$PyJTTXISYoQF~O
zpSL-#v{`tI6(~ZX2C5^Fvd=<9;=NxXHobum&WKK6GP}W#wa&5nly3606i|x~1(77@
z%X?l)NePMk3l|QR{iv#%LUfF}{5!fMgWcUotK<iI>$%-CwXksLW+ep#hP8A|Oy^{N
z{P+=-G3_4`(o=C#<Zh)ty!@92vj-0@<X=2`^$>$!c6im@#rkk&p9^0u9&nkLRNah2
z9FCIwbJM3#M=(%P<#C9k^HUwcT{B8f8Cu$5r9TH~E@?&HU0@X4`BZkK<-&yY;iU;B
z9uZfRp3QP3-wg;1OktNs^An<?N4Gp`*{i&l-Yxl`W__Ve$6YP`v)wzCUI%8F$?q==
zE}Ji`xX1B|(4N{^K~KfR#g`Trhnv!@X0ey9v{5)B#E8D)Aqww)DW^}5kN^Jjr;%9K
z3_BbHsoOnd-r?fnBKxJhx3?D-SY6^0!>(NhLd^NkcjZ8FCvkAPckVcj{-ElXzza6{
z4Whcqz6zzllX(vBZQ6c5Mg(MMV}oERHTFu(U9hR)*G#qTtgUtQ^i-y4d<6vsk-qq2
zG&4OlRa{ggO6p`f^0Q{!-33G1-y#cx>hzxW_Hs-MxllxWP-AlM9>LQgetw!QTg**O
z&&$gPh}pzm_<|(x(X6tjCPI6HAJI-S)=ZFCHVm9#0jwi+U64(hU4oxMw`!Ekp^9p%
zuR;$ehAI$c(VS=$kI8_C`b0m3dU=_vK=I=(%gAT|mjBAjo132}r0Qly-UHlr_;r<N
z$pe4`7WCf1+$a4;2UqA<9Dg|sH*d=ew(fR#Gbw8SQ9)J~p+LsvFfsGqi`TAIV=EvT
z$gj!CV5bh0&7p6E)92;n(wtTtwX~WM|0wLQ0BB1-{e!2$qocnc0lRj@HM`!woJG^D
zqLy@JsMz?!*SNP4a_8GbP+e<lD`I=pE<SuJzrO-okKbScZHk{f_@#GeRvk5(ckXm!
zPvc@^q36CnJ_#_4d(;rc_c=5&Y-3`AfwcsmgHQt%e3k>%#X5qdUVeFb8R!KZRwZK8
z<cZ3kIXQCaNpM=s9$!=02b1dC`WyESH(nXu@aR-X#n?*gkD8hpTn%F4*XXT_3(*yw
z(Ld$Azstz#AU{8GQn;JBVarbB_vCliZ=_HXwmw5BqnW3pd%>|m&qFlk)tl3$gF1ut
zIlUd~T}SA*Z)as?MImGrGTHa}=-qU1jR{$MIVB}r!;iNYs;#OLK6%n#o!-jad=Y*y
zHZBe~A6CGOsTw;+-4KcU`}gnOb1w!V?TyK%8gjO_wi+55sQr9G!+Lj^ar^d;*4Dp?
zM(RZ~f?MKX!m_io!7qCPry~H~xN#%6A)c%)*Na;(JO}BUo$hu6i8U|~;1Fe%luXJ#
zp3ybw?dnRfnYhF6ZDDR6fJ1<r;G!=~G-^oByAgIZ75W3I#r0KR>yga@=~6O?s~zg&
zwp_c!$FPHoeq&PI^AN-XbY)ObKK4w$0|%1L`(DZK=UI_oc=P7X^XJb|0GR=fxS$!A
z;1h3N?;6Q^wvweEuzynbS{H}8JcV-c38s#O`~&sg;^CGo?TZ(sWcGU(#ya#@hIMPa
zj=S+=?Ve{ZUL+1)|6zOwSQi(Cv4-ya2TpTcOuo7~KY#u?qlJ6DS1w;xOEfg#*EM3K
zqZ@^vW!$kNeY*Z0%3V%DZEyoc#qD6|u}0q^2!k41cM{SAC;*DvR42M^Jr$L@gl=m`
z$7va)hk@I}QeN_8UYplv8-7GPeDrbu?F5EdXroDE@)3T1LYM)l3ci)+E%KNDid!p+
z0D*r}2TCikan8=ppp%JygF|Y2u}3}T)?=+G_QN-vALXnpf1zh!&=q`(pVFy9DnIgZ
z!}1Z}0YRsK`&Q&$;fG^t_tq8Oe)ZX1M&?$`Nv=P=FZ6OL?`du}iF+&sWfUBgV36zA
z3pjOqu*@|``D<n6Xv&3uW_Iuv?ccXg6G;G&fw+?6{^AAygi9-;YCnAgNDfvHorTfN
z@A76CxzpX(w`=!qdl)IiXO9}FvxkD6x_ngrIAHgcWB?Cze`VeH@vpYy+&)O>Q^Yyk
zEH=mHkX~bzbX+92K<2vqAkvfYmw+K%BbEHxZf4cV0cAx+;pVpu8Y@jJR2{y1XFDNN
z33=yPdb6e|7G4nqMk$kk)?%GruyR`oiT9BDzIFRH+`Wmmb_<J$aj$tQ=zHvDUTg#W
zp2f)~@F6M0F?<aP*<OYK`FDCmQql~mzdn?U-{j#qZt&pMVx{-FL7}A$EKA?U52C(i
z=~h!eOWKUBW!A_1_6pOkbmMWyl3J_xRX=T0VZX>v`0((!r<q$=3{{2K0G5Lmk()Hc
z#f!eAh1pr;kzfPz1Kz4czxx6e!~-Kx!2&N%*jT6)l4=26U+?fw*w>%BeL2h=>1$LX
zs2uX~O&fkJY}}X(c;x`>kD3Per{u-)5#?+so!PQmKVc<KKCi^@WueZC7cZ&;v(;}h
z_!pI?3}ztm;^gG?jXHVagsH(I#t*-5W@0h|Pqc+uNH**$RIoZ+#RR4uzZ7FMC)~*f
z5EB2`3YRR>_A6%XT`^);5mKxXR&B}j{<{sgPY>Kq3i`g-*VA+KdDzH%X6g>`oc@wd
znQ9SQ@%m+DpUxrAJ_|>IgoDrZt3?5w>g^Ati!53<S;plmUu(#6$|}^UPch@(mVvn5
z*S%0<?Nq#uu&C$)syn#%B&aPRS1dl27GVl|^vuo8X0-K`jt<#4SYStR1IU)LirdC7
zid;MQ<$jE+Mw7D1Tx2#b&zpreu357NF+y)clEC#vZQOZBVU#t9Zzuc*=u`aW$S3^q
zVf|TLJcNA0p)4gGozGpmk2Y+j*?s(C9SQ;|h5-izR8mHme(7`V$#CK^D$SHW_YdVp
z&hEf19s6Z1jc#8sxu&<X!x=ZSitlkeL(Z>7X2-5w*Ka8#|B)X-ZJTdHy<0Cp0*Y<u
z-996j^clKSn4cd4qk!LuI)UmkzihwcbxGf}Q~%jXK0!VcK(G(=H#-{}Azx7MB{pRj
zx$R`&FjT8CX|TVadG$4q#Aqf;um~-J!q%1LrQ$$2rU1C&REv%yJGbB_lCqt3di!i7
z`AMy3y4RVQ`~YRl>g+Z``Ch(!2@@prRvyej=ZW38$ldh#qbY^rlRGagoPH#~GIAZy
z(#p}qGkdOhch8?U*_k#S`gSLzK~^U4<kIb&t7>X$XV0GPE5YqT+r`-!-6G)RGcrKC
zop&u%zNv3)Oc^6|1Uv;SUEPP%R7EN&|486B4<=bqVvz;mg|xz^p>8ZmMhHx(U~pU1
zc}yC0bF|dd(He40ZZeoJ_Z~Cr#<%vUBTuj&K`jo71NED%lS?Ij+o%3`HoJcmKTpK<
z5zc6KZcaSHQ-1$D35S#0t~-z)q3hqY!($yKls+*i{N~j+%i402am{6=Z^E5MPUpBB
zH2f6n7Z!dR^}$c|;lsLp;Ym<E{@~tEU>e^**fP-lKXu;$O@jK`rE<ivE>>5$_u5*&
z>=lH|lcnP$yKyN`QZfswOi#NPx7~^ZPF&$-TtS|J!95T4IK3@dP8Tlxh9NI@Ky;CG
zr<j?CEnDj)i><Y_(7kH;@G2a+w`h6r)5}+`fRnzrr@v~T@Zd%q4cAB1xRnKXKqD0c
zj>k;;iz$II-Ul7$7-&xXCv)Q>Hp(p&``BeQF*C#c@%8H+LEs!Im1hXp%Jx?MlUk)G
zF1C`iL;DKi;oE!2k8Z77yOvaQq2LKCJ%($@5JVXv+=?-3NPH{m`uH&x3-osEZe>AU
zUKemZ*GEr@`OtHoPJ>LLH<IWHW4v=`frI9)6*0YR=QYUG6c!fxdcZQ3{`mvX{^CTR
zI)I!8{|as*qrU2ALOQ4iFb<;jv-iK;`MR4Sy$!Z^=M%T5Pw!h<&CJYz&j>?c15rTU
z|ES%xVPq*P^I-o~R8>)uk&m1!Mu8*i$7STkMfFSe_F~(%?mV2Gn#A|eWNm4coq&V=
zt&Pxv51ZIq$PfPcUIgcw?T8SzAG9IrB{4;KtqX{bQ0r`t_zo3!u~$MAiBL7lNg~zw
zh>G?>UbMKF7=8uL0ybuNFGY@F@Y<cjGQe~qML)amXpHY-n#uaK00rH>ZChJM$L*BF
zk&%%s{jxoQiXsm4=HXGdReqyZ3809Dk+B+Ej8dDs3H3moqobn$Sh8P~ro#^Icm^>8
zg*HDwzn_aMK+;L;kWZv`)-rA@#Ty^SMUGdpT)tEbDpXH;o})x$3b%R?jFrR7u2@}p
z^1&%+J4vELmNsnIpp|_8Vd?#U3knOj_4iyRk>jB~hYlS0+h_i7?v0VFhkT?zSl1iu
z9+d~Dw&&$j-8w<`r~f{?k6$Ly7s|^k4L!YC_f?@zNlA&!=WF4(%I!jPZfAKXSKbft
z#+V8JhdGDr2E{~Np$%t0#n`U>-bDhcu1)_dpqltI#dg3O)q`ecd-r5$3AFB(h&X1g
zaEO<8$4M9N-%+m0y8lJdCrsOA=E8d!2B(Ff6SH$Y!KnUo^0SdoR*Nvky>l%3TeR;o
zbjr@ojs%&TE>->6>Lwdf(Uj?s2EubwUPaONtLJat{NU$TgCk8@w>F%%g~59G#4{lM
z*Jk|-kdx%eX3I?*2)0>Iw4QSB;DH0RX;x7(d&njEx2&%(bfGS7+xG3=Z{J?8d#9(P
zqdG&e3~%v5NO>J~tlxt&wqJq=P>YCE>zlmL_AVPI*^r;_`c(S#X>u8a{@lS>UW6aZ
z7_5KF7Hj}0^wHfk6w3Qysj}k;b2fb{3$5rwNi4S0m4>yyZ?5uO3>fNtJ{R`x*>lKZ
z1s7KlISF%dxtP4KvdrmcWMgZ_t!x-}l^H4f?bcLOuGq5V)KumzPfRlBe&y%q*VOpg
zg&~$`Zf*uDb)4+oW)<%o9?mo33|;Tvu)0*9T_?>$<Kia3X6&QK%_6zqeRZg)Lt)i-
zjg)fxNV_xfxcbe{U-FEn;ab+s1mN5ko}P&O9JW5s6M4w}k%sjo7dLn5+oP^_(tk8=
zl8YViSTdDRN|fH<XWfEVRH0ZFYY~Z1DvzKf4n>Yh=@}Uj<q<B<&XlmL^@x}N;Vtp&
zii;%({6?pzmr=uW$Z-W(d{wKM>VA>$*HCR2B<bX6!Q|C3DxO<XPE0-Oh5#HZD=U$A
z+^^vO&p-c|57r*x;>wLI4bJDZwA|`iJLYnFa5jh2(!R3n^q}fB8+}QuK`|b`aP7B0
zdunZV-^$-k6VpM9=N_IO{Bw7S)AHCCl#YKYtBlgt?v)yJF4VD|9leQB_R*i;x<#+}
zh`*45;Js>Io@P3n9gQ?OJ38M-tFOP8YN?RRN3Ek-&Y1lQZp&Q!5p*}f7U8)Vazd2Q
zrq6}v$FIaR*7!R<@^HK+;mm}M8Dwht#}R=9_#P5=Wi=4njU9oVdDU|t4<zg)J4cZ^
z@%{eOCJ_7eiAHJQ+Xc0Yi;DwMBLweVQC>dl2^R38`UidZW6En!leM5|uE9v63IHj_
zjLFOPIyyS=R|qiO3J^o$!>(1|J>)`sH)WL9e)}xB69>`<^<tl$`12s<&zXhq0-obQ
zZ^g_aSb!<~jjG=7bb^==-=o!*7uQ^?l^z)$UPg`A|Hao^fK|19>%&{cK#^_*6eOgi
zyHrBDL!?1cL8Kd%PL&dn7Non8P(%=<Q=~&cQY1z2Ki1xyU)=lMkLNi@Wv{)~nsbgh
z;vMf8^9WM(iU~Z^u`?cUBPhcMk;F$fDIp;tfU?1+I53j-rqcFFACk!$Dam}jLTs;i
z&1rrH<gMSI#}=^w9~WSeWDet#1TCLFL4*_}nGRVTJ~6Qd040VfVVJ40H<5WqWkYS#
z21Aq;SYT+vSoYv6ND*v_d|s^ocsu(c*{K`)wMcvovj$kSn}>&tv@}BBA`HI9&Rz(2
zg((LBu&iAIZd!2$Gu@dRic`Flov<T>oL5|-T?VZ?ln6lX8n(3f;Z{FHqOdcR-u5FD
zgeOm)l$4YVog;^r<HMg{z$}Bt`AIA>4vt6AIl{eu0wx^*Atd^Da+){z(UaB;5osLj
zU8v}J85ITF^bmx86BlmLbd(bsQ|FBqE>t#g1GcA}-1%k>V6OpeMk(U@C|LVG7dR9M
z{x)8yFC2RxDQ`fllcu%fM`KT7HGs#CIIIXxvADQbE(V<qAQN;5*$#+{yQg8WE0-{X
z)%@9+O+-W#BR9`o_t~rzC?pR6`9@1^u?^xApir8hTuwL8eyzI;L4zAE{Klvev6axg
zuUMDL75-#8i@q01IYu5ysv}%GpqpztzX}REm#KC)XWkup<?5c2nuTk98{6qOz_P<^
zOni=Fly4-artadv`4!D9Af2HKZ83QgazCDfXAEPWk$>XeB1ldy5&99nF;}3H8I{e4
zd9$HsSkO2vwYE)(S_Ul_#iZAu{24WPE8+Duai`Y@l$d9VqklWJ+nsJNioUJ&1SIo?
zAh?B?SV%y?)z!7!Zqge-hn$?8TdmtT3{g1%-hqEoWaN|4I|AYi3?6M84Zf4uP@3l(
z6T?h7th@*Xn7%=6O{KR2Rc=Gs)6-|qZr{3vw6i???SzMim!uT;>Qyz+$dFDbw6wPt
zQ5|J~ghH?h?r8PX!xodDu#&)yi+%Z0t>DSmlbE;sm_3M$&Pe!tSXL2NnL+hSwe|H-
z+XeQD2FLC3x?S64$6KwM^#-uOfGNRD(ulaK6lhJ>I<Wv(2e|SuyXtv^kV0cA<EdtI
z1uzm~3ZRcHPfx4<bP1;tu7sxoXGUxLv^m}LCeTkhk-ga>#TSstdhk)RBI_W=CS|;%
zHhl<IkV*gT<ldD7|2NU6YW8nQL3>i|g5B<}t|Lgok_YTV1|blShMiDU_BUvZasVJ9
z??fK{3Fy<l-T15>(C6ST-N?%Zh@+uHCZH*rKlE{$6%zT_goK1RIZK4nOR4Ta-3Tfl
z3;lT6v|b**?eA1c4O{Y34E@VhYp-u;4NH(uPfz1CBlLW5_JI_W8x_?bXXtI^y$xf7
ze_Ml_1}O%p2Nr_t^1<$Uq9JgqIhr}3CYryBJX{(Q4FSeb>M7Ns3J6s@2ad~vpC1;j
zHK0x~Q}V6c#via~g?x_Q)gbL&zVGTHTzHCr;E<3Rb{{|)0*oQoO+-M@n|kw<j7%#S
z%}8?2r^}B31P53O#Av9uw*=x<5F0l8Dr8D@ic{g=P~Qhe7}OK|$A8crNV(;H2(GOA
z&I*LTtrY<9*Hu>o08ALa58(cd0Any`X0~p?WPhW!1dg+uPe%ndA+pPs(e6TGDf_aq
z^xdz*G8QJ{?QQubMvg+`z|@2{V2F_u6;1i`6r%dY(TZjG4Rp^C5fK3m4Beh-$jM#7
zMiw>$C>p#7;?Fg?0CNZZQ6z~kU9zzSe$QB}YhmN9v@7XbTU!8lL`NWw{@1Tx5%~5%
z8n9yC`|ZR9THoqDTwdG(CMpRb;U*w20AX5z7KV|6w*WaJF#Zu?VLXj2baZqa97r;v
z!Fg2|tQ=skfIR~!Cg*9rfq}LoXv8PRW@cy0Oj^!Ad(QzLaa8a(sha?D>eRbefa3;6
zsZ(@(d>n{xRWO~vtjhJWalWbMzY8rpMq$n+XJP)UZvYbXw$F~U5&Da(uppXHBK-5f
z(kx4{&q3>JnRyohgS|cYOGLT_t2JL-$G4p5n4lt7$m^;y5&!)86Vf^Jx)3z*ntsBm
zH$doDjUch-{b_I9ScgJifDRS_?Ex6ijO#1Za0D#^;`{voIP6b@BxzO6&CMaQ1-%!<
z1R|Z9+SJj(!ol%82x9V?)zzf<_^)7`kPgaeX=$MO_$O~(BOHXBLHpz#08Jrc0VSZp
zJP2UU*^88p;1U2$9ZB%Otp3{Cg6!9ghP@dOpl0~_@S)m;&EVaujxT^rEFvi*xG+h?
zZ4<z4*yMQ{V(}9YH{h#2dQ?0g1peUnZt=gKGarK?_4x&SQ!_3_upO2>-msUz!;ck(
zfCHh|wSM~an8s{CPE&RYR5Q?PzQ3&(0w0|EC&=Cijr}KmbXd%}5Hc8LH#Y4<LnL2j
zmnmNM?ViEfA0h;RWW09G@yAx9AwDpdgO%PU0=5k_g^v{jfDOS&Af5u@fXE-DS_`&o
zjfRMzU=64uNn#|?15*Pf9PmV}xF!dRBba3JfUC6N+rrX-kd8O{_<%bF1OgzH+fPD2
z1PE^4D$^Jm8j4K@fuMjQ)MO?h5EK&XisxPcUlcHHc*<DU$$vq`x<h%wUcY$=0C4v0
zm<YD~ABgkmVe)^&)&vv*UjZP2L?>(#NXL|c{e)BvfHD370kJ`=Kv%tz4!-r!3>G!o
zLb(#mFI?YUlSJ$glZ=YW^ZPqWl0JWao+XIMcJAW%0`jgX3D^eX@CUXJiTr~p0`}<;
zrHrNLxRHs*HQh5W5Ov=|m?0nwpCYIr9zKLQm64Q$;*T15EKCCgHXbpJFQ%r9I#x#@
z9V9sbUAsFLRxj~dOG=)3_I^yQ$Uu~KwALxl;BR9frSsYtL`^;kWr!;X$sM?HNXrc2
zvoly&p5OyPd<>bjQG|gD&q6@0i13a-HFy`S^V$oVzPY}a0P{u}fOvYmMe%xkjql64
zDX@KBsT=XcPFf@c)<x}6geHtr6+EJ~3jQ@$uKUp4y;hQTdT5Xhq&*_&OJorcCN06h
zf!~k_r4K+i;O62ofNCgUD01O6({HeLeERf>3y1sBltjmYfRGRrv;evw%i3W831m=F
zNXQWpfwcYtChE)Q&o=<Q(we>mNy3VEa}m|Mo16aeIUfIT0hq@vm;Nsx7+lkmmFA?0
zg|#(<@6`}$?1A}8&&WV$Jh@WnjDkIi=CAu!D9;n_fm@=jjyTghXAzU+*Q1uRH&=6}
zdqvBV$JYIox>6vYTuW8M#R4*IH@R!C0U6<sS4d7y&eTq<udfeYL=d-mi0LT;rnRPO
z&Q3vyps%n0t-{6yQehCV02O3qWd(pNE(cv&6Y?xrS*Yqr<hD=((mjYcqInQb6ODNC
zH-|2oU-@`NP<rRGGE$-p?kAud1Ox<QA0DZw^gtT-ON#KL;NAY@?l`$js40pE6jd$f
zGSJ)Oz0A1cGcrt9m8MGk5idy&4RBO;+DzpT2M2js2TDyCD0_8|%CXAvTlPp?BfM{6
zaY=ra08+sxhM&}JZEP%prwWz}P$n2f4fh04ZKRN9Rhhfp5aoIRBm#8k>r?@up&ek-
zAkC;iTDO%VozbhYmdFaYQ@x3a3AKnTJ_Us!04=aVQ-ob?!OMp{&AiZ~z%SyfWxg;I
zpr2j`4OnkOw7grc`q89he&}%YEaGTMXr_WD_YMCQBkN;1!{XD9yMKDZ;A$Z|@8y^A
zdh9+>J%w}THe|kn<Df^Gf@JK%r$m}(;`TFsWo6vEZvd&O4`$-yt8_f+15ObF*Z9x3
ze=HchCl;Y3NiSDQ)C>I6yv_QU^!_EF#oA6bC`uL<7J@25YBZf!DYe@XDv}=m8h5Bn
zWP2eM2cWHpZHK6$zn@z~WURD2IsUy-TX*+DCC7jk+L2+th)B)KhYj~GpEEp{=4P%{
z>I{1s<c#!wv6F31xaWOC9n@EFTj@(d3n)sF8~`oma3QW~9U$lH*MsvKln_Al&0pmy
z$k2guwq5IfDEAkbMePa!&=|k@VcWvcMsU!#Pj(aK-GWes0ch#Ica)@LG;o$dZ$o6=
zLGShA4i8{ww0+hUdT{TBx#*t*&A`M0DIy>yE_Yl~TkAap{PWeM<tMMk@0}Ma2oWd4
zw;df{3HNDbhCP^=(=u9q72An-mZxsCa4>l2sQHKB1u8Nxs4?S;XB82Fl!^H8lB1-C
ziV9K#0L}tr7+gPvQG8DZU8a(w0l6^nq+z-BGQe$=kZ1xk3JGa&o)}|}*70Kj(W<oU
zodHb_G1?IrUG(?{psr&CkUR))f#xpa{m1o}1=iS2cgP532%16_P!cc~mw5Mce{O78
z4&}#d<w1pl`OTY8+=4)7B(RFfJc8yikTy|L`o7bb2!(zBH^M59<GB4plNJaErE(*c
z*l3}nM+=?kV}j5+f$UmgArZ#1n11O*9;;ICL-!iJBWM>U0J(;PgSwg;q}P{$fdQD{
zn2!1H$z1XIz@gU@>J&wFFOApn59_3_)Z9u73J$gcTLdX;V5d{^+vaO9Ycj{cwgE2!
zI!7SB4dgi;{zBpkBFjP`7y&{%4viJhp9hC}5VGpkGV=u{;6mXvOG-*s*VVz(k=8Dg
zlYhbN16Ua^>QPzq&G*D97>S&S76*vMmoHzy6$f18h0XN$RbWK1{(S8NFsKWWjsL@E
zXun@W<8!)wGdd#5wFS2W93jBkEW!GNM*~<C1adL2URfGM^UXT<%3pg4a{ua5=eXRp
z4avFXWe>2&;LSjj%D&cC8CKFfja;adZGakVFmnr##ep%_aK~-H0Tlhi!~9pTP6Htk
za#Db)jB)?a;)L2<u&M$6NM33Wl%{S0pMZb>h!uc!u^wlu#}b9}A{`)nfiMzqGw>AR
zFC88pLP6S_{v{}Cx2;i}Lfd&iVXD+hll+Ps>7*TiYFftvF=SUJFE!QooA#*~Ksbp3
z3&d^qwE)l&dMUDY#O{j_YWO84CQ4^O_M8;xKi8r@15O!yr$CF0garF?cyD&dpI~9<
ziMbWEKR<Zt1e}u-o1uS8-&v~LQn2_PW?Zx%E^EbunrY^Mh~$`jl$dylU}@X^#KMME
zW{r&+K#x6FMW*NGt~s6?{10H#KmEImcHSTYH4o(U1673OVB*7vTQ_bLL9s5Rr^<v(
z$_RAOfi4Phx_pHs7@{gHdv85oNecRJL~<9>DY=6`U|(s;7*(D^<hQ8|4FpJB1gZ$k
zM0ik8_R1x29Kp(1yHWZS_#rX#6a{+KQ_XpRQv={1By@OEv<(S><pn|jK~qsf(*wVc
zy8N~Fv(nbqKe$s!Zg@2SPU_2c8}V>_6UreC3pgq;B>15M!RquR(##T4EyHUkA75s<
zhGba4J)QvpVd<l=fW|bwRA}AeO9^qx?CdNrrQ+#>M1?HS$`GA2>SmM@01gGPz|Qip
za^9dWJtbJ+Z%@!b)6+-DPxhj53GItENaX6ymQcaPU=E`@CcwKmfF5BywqrWiJ~(Ux
zbnrJ!dj5%i`q*V!F2TC&+lJ<B=Lg}{Q3E270D&@>U`t3<f@ZChhUr4_p%8tIYTDV*
zWwW@;$Ho>OG~qgXk)`DnI2Iyks1$YSBg>N;%*=hD?B>?iY%`H~3=8cVp^Oq;={LCI
zzx4LL2<l1^w?lvChk@3N${^JQzB?^J0g$2~_R%^I)C>u740=%+wjTTwfH~{QhC@jG
z0(MYnP;VXc3hr%9je$GDXFbjz!uLStAu*bcPy!G?unwI9=}A5DE81s^^<HJz0zeU{
zp+ID^DxnHkb`c-t*UJBdhSS7v-zKA%{1WnWPss;C7>pdyOOIMPF^QEd3{4;)>w$re
zMot6Z1VXDh19!v6!n{KLlt&O1EchV?lq#m<jcx);O9U7X&=b7e;3ze5AAdiNQf_K?
z2GQqL;ISr9ASk!zUna_h`Ec|)8XfG3ol=&R&aelE2LY&{J}7wzInP4!T3p9I6hEMv
zw1d;_jf)T!3}J^@1NOVmW|YwO0^m>PzJ17dp!1F>L(CNJ1Owa&+#Yt1-Cw_!p_mSG
ze;MiNL5skrkNa$P=Gipx03eur<mfmClm<)7Y?51NK>AN$mR#wl;uP$R;I_$0@vB$g
zfBhPd!1l=TH_-3{my#zaNGCQ4;6A|7T!6_1UDb~WKq@pSFjI3l;}N89ZOu#o-4w!+
zfC9>Lkzl4(p9=?R55@CvbSu%Wp~iEc3yrzuuFy!$)*cUy0_NJ?U5Eho^@p*(Tul~Q
zIayiGDkkC(*vyYD#($!71lDq-;mhb8E6T_Mgb5dy9`6WzWNZTcM(^>`5|9Q-Vj`ku
z2V_q&qY(mB_G!2foeYGk9S3=^8YJgi0F-wWN~-DY><ka$;^s!<cq&$yS*&>KM7Ren
zsv05AQeI;sJdK3~WQWB?ONS~rZH9Jp>!WNVgw<sjxv{<gy9g3hK}_`Ya94KZi#4Ah
zD4|7U6t;5^6AQ~Sl8H|TR|K71)HsixFo}ZqAd^rg3dTr565>Gsok@lTfU;Dg;|Xm~
zAt?%=#qa%ndzKBrr%2J0#*B+zt5iP7tq}D>45O8=(+ka_zW>F;vhL67gfuClwi)T^
zfj38+Amn6y=gzsm2*k;J67$X27aNWK8q8jh$%OhauL3O_TTl2qw7e<@n*vw|c&3n`
z^Ms_*pH|Ql+2_!k-gmV1JoY<qGppV~|3M${MhaYgV3tFUXE`Rn81bu{0?5PLPrW#v
zNj$`h%9ctd4zH6v{*}^vWlvyWpo1Y$h*{(^`${vPW+^4<<{ZH*lI%97+tU(YW{0i0
znDt)p#NWR-(@+Fl|Lz^@J?3n+=C47(47|Wwt^mX%Gs!o-mN;jgsThy&#w&m#Md#l>
z8VCvqa2)etFJHdA<z9~=wO#|5=KN`bCC6Het~92lbk}vIwGp@3V%4k&eH&WI#MVw@
zd;5~lW`shdb#GLd02?MS#?4bmQ{lkL(8}}2{m9OV13WGX*s-cjf{u=k#MYLmVn3{-
zuR5^>D^;M&c)vI<2YAm4A!N!i4_`C#^KmeP5)ajD*tP&U8Y2UUVT|l#q0;cHu`#_g
zF&FR($1=`?py@Oqq)qy`4mk4wo@7u{TN@%t8+)Hd)cu+x>rVUNpl046DH+)Xsvv#H
zHv+nZ<QqHUZ>xGIKVFvpWORl0w#0I)eyzHaQm2&Fw`##Jot-T|><0yarBXKyWaBXD
zjGpNcdZ2X#7=v5n2g3=Lq=n==ZQt_t$u1<f<4;M6V%21Sa^p}Y?pvRFT?s3T#z{MB
z?I!XSDt^0HqDLHoo>P&JV(xZ*>;L{-HIJn60XJEAKMqyPVM$T^xrZt9MTEw5B`$l<
zz7UxgSu<}AUqKB}2_SN@5--&&FqDZ=<PvOJH6(kG7ka+n3IXw~8)!OT#-qDDN}}KC
zG18`+{}}rZRK0$35J78cwoUcFtdt8F1E`V+FX`gqa-*ys;>M2zHq0^Yo8@WHeC70+
zGg?>0;BH*9QBh`$kq(A7zwtPK^AhpFJD_}jXw_Iu(ZtS+|J{$gbB#6~_yoWc`w1G?
zx}r8adlAlng36?&$wqO4@oIZI9v(-)RR%W|VGXJ1x~f@#y5pjjR49+<b)bZ>SU?S|
z?opc{Z3rMm2Vgz8n0nCZ2MFRoXgu%#Gzb94Gq{%lN%;lJ+OlF4X>(HZxP_7vj$KJ&
zXzx~E8~;omE@`%5m$^EK#?Zmsot(rv0!<0%KkzOg_A)2W`XUBc&t+hy=@uM8Gh4L6
za=M(M#o}_f1)HW>x?kP<0Nw{Rzy+=VC@We@P7|!E&?75|MWrP<InhZBUm<if1_`^o
zv8&FRfFh*?K?bs89f)SKl0Y_`LQ-J^RNySgjyl~oVDCsEXSA@V6Tdqnh*=k*P`i`s
zD=U8>g9Da8A%Q37wJp@kF+h`Denh!R=r@svZ;D6Wx$&AlmiIB6B=fHh%+7ZroMhVq
z7wK9$a9RX;dAF^+Ajm{0YYrB$G@HfwGgG!p`&etw_xsB0&tmMtz)ctpq)-S{ZEe?(
zto8jJLZEwaL*pliqX>EdP3=OIefIX7>)Q0!KhHZFo!_)!B}ZoB+k2M_hwvRZovsn$
z1|XOqjl6(xapT>_Er??&0)Pq|8X3t?qen(a$OVk>i4(9YY>f4;YZO#32`lQ)KG)IX
zMgTw5$hih=5dz?cY6DILUKwA!0!(RUWCBp45ZDM(BIv|pMKTjz{U6*gX}YdX&e~bz
zxG7Zjy*=ABM%rM>Qo#EmeH#!E0Ih?Y#=1BZz!)J_MsMD{0h@W9`23M6%pjWu50|^X
zV3D{Rt3g5G3rpnr4}wUvrUan}d2m?RwRA`w3=C-0=uH45cWo7fP|$>vC5~Hnwco%3
z;BHN38&+E6TbGP0uqL23CI?b`z?-ElIf5=`sYBd@c{jRhwUy3l<?z(h4C^3nAByk8
z`Fo3kP^@48rGOBT1CIl+2MuO=+NjUvW}jbz@w5pCWmZ9M%Fz}{%`Wc2NULuoLwI*}
z7s=lG))#k)`vNW-3cC<GNc@4=v%RYe=1ET&%_sy8YYXHM^SK~jJfJmmu&Y7v_eGMh
zx;MY+vw+`Z?n4T9(67;Q9{Uf#bPcXQO)7L1CBf4>hmBq0IS6p6{45~49X_lurTz*G
zUzY?1yA!2q1SO%)q4M61)eG+*1mCj9$*}t>Eoys`CFVRKl1%k2-V@78e*OAOLE%{}
zEaWsvAVNxh<QSv4eT_5}+5NDc{=$Edx&T;C8vbyLI0xgzj!Jp^=V6M#A+C$nV~61|
za?`C%23ju_vX)K_x0}=yif0%XOj15ExZvG<2Dw-iJ|X0<fpE^+QCnSoS<G{!ySvEo
zEM%VB+S-8b3*rZXYd~P&uge6m#V;;Zux!$ur3eqG<0*GD;C@u?_9V&a5ziT?7XrdW
zFANI!An+n(LI!Ygs?Pri;C2kdtIQ?hn4hKp&h-V77*`SI&<&@xE}D_J7FS=!e0F;z
zZlGD()B{=8rv456C(xaZfxrm>vp=?=l_aBIzWjjRGp9dR*)ij-TAaU7E&t4ryR%p|
z|5-+{DrBEfBLd)R0ymhMxxKp^a`8|c);9AMG3JTn8LFhaaVuYBP43c~&=cwx9o43?
z`x(NrqEOu*TrPM44<u+2Qs7LD_Yj&<Oj#%BUTaa+o!j%3ilzPbdONlcdWu~)ar9dY
z5w0<~CQt27(oj;$NkTz5RR1^i|Go@8%R%)@@&J_`Q7gXsCU>o71P3D%KjM`J1(8dW
z4P@@cTxY2IeM<&BSd<YzIC&60@BI9EH64hcz?Zr~N7uZm8G9c9NpaZQ^#<#ZR{@4H
z`gY^SA)(P8l`IjL&2ZdlP4s_|H1f!P^Ens8^b1MlqUL83n}h`fE(<(r8XO!n;y~g$
zfS4N^8ph<2cw=V=3Qe~F@Wzm^AdyYQ0kcLV+pgJ4!*|E_&EvS=zC%(`6CA&JNcotN
zTY7x_FP4Z5Lli+T+^S{=NL0{1lRy=^I4LEId<A45Q3Jm9MmX;Kz0K{1b$9d?fFO`Y
zEv8FE_=!3CRv>!BaeK+Px_@qM=|Ln5NW%X~0>N5<&Z+~~XTIP_4H&4d(XI6@F2AJ~
z3UP8Kz`R(Uc@kD~KR_{{XYc+DfXVbxV7U9?y4#qc84n>raLZ`n``WeoM+0a*p**Aj
z`N`A>B9b{4+%*CV5x3ZOjY)lOll<PYCm7)wz=g~=OnbM(JAm|mzJnNHgyHkNq|}$I
z(nWND@gDx7DnC3^MwruMk1YQHAzTmYywgD>_P!e;5f3hA^tNSW08M0!KJRf44uzHF
z_Zts1kZG*IsqBFp-k7<^A04>M?e9O9nI`Md*@HjyUUeiJxxzYynPKC+KhhNlD3K7q
zm)J_cpl*<7+$42TaQ=R+gq5I#Ft||L2L$j|E3$W0gu$Um0kj<$ec&qbP<8@1h<90_
z2Y@(`Cinrmwzhf@+~%3&S3KggyCF2@_Hm!iT;!$u^h$5}!X*;8iv$}(uEsiq8qmH9
zsUL!V|FGZBc9cT!kK}5h90pL~KLE7@AyG>`74F8OqqVbZH8>QmTnb6V;R>0bEb7ne
zVSFeJmaO)-?G)jkMn*<JOn}zR4+1?#O-@b@)KwtIL5#$5`g$>xIi+<?r~_>+CL{>_
zItXj7aE=J03xGgIuQ^`XY47ZW2GJ#~iBVB>h%X8M%;<%{417r^4yj0zK-@dFV*AxY
zb1Uf-VxOQOLa`Z&g&;*ttClwiRmX!oa;u*@jcsiUdo78<H=bKrVU7A6bk0HNKJeaf
znxTpjaZBM4QE+c1-wZFGetjFfv$TLSD0xlgsZO2P@W{wXfX)FWx1D%rWorj&Qb$L}
z($dn%=t}ohKYy%=!mCdf0B`iExc($AOHcE1Fzt5Kx)e>1uCOImo{Gj~(4DFM+b6q;
z2wjh!H~o47pbhKYt52#N0UttUXQCcC4dVMJh+Q0PHld9~>5LBV8SHoZ$KSD*IoWMl
zv1!`d_t61FO2zUl&R~_b_V_IAlLb6%5oOHX&qBCTY&MTI!3xN8#~ebFwMCjXffVGL
zR$Kq)YhHOmHZHF%y-GAMy6<52U`$ffk$8Dd@6>I&|8+x+n%gVPg<#+zI#g6pcn3)>
zz@_1|SOSEamk<DmmbEe0=a~q>2a!{yE?TTZJ3kCXY)v}N)<F(XMmZr)Q!KRH+zyas
zfW00P9uC3IAU`NvjkDnW&rDJOtan`Nn_Ekx5F1)tqQ~qLr`T<xy|6GGcDu;XQ0@gJ
z9}SJA&Rw`r_8YMoRM+B&NH_z@>m!1KjJLX7dUh^-k$ZEz8(i+PJ!u3Y4xo^bw1yIO
zD{qsruU`ePUAxB0`T+J-$PbY$@Uim!RmoUz|1!&ZJtn=5>{F`ymZN%@UFd6f(g(~n
z7(e*g+jDi8Fn|=u^XJb23b_c}aQ3)gxNyR)b<m!9!sf<_@Ag++tiTDT=6M~>Y@7j(
z2C!TnpY(w;ZFW0xuixcJto{&UQlyItxWO7B4`U%E<;O{B=F`&F#faZQNWc<hb%4`%
z_6#-^;*p?4RBNz~;A*{z<mb<gY#m)(AeeFlxr5l1{-w1D2^_tZ8@xDuN@{@mXY<*y
z%$#Y5AJe7^+hcy=$hq7|C^QL&am4T2Kw%?f?EL)w;jmq8R?>UIpzN~=Sb;Y#?adlc
zeR7!gMWF-#!$>i|#t}rLJy0Ma`!fI74ZO@KU`oAw35|84LK-<C`C<*dH+%;<Pw%%+
z?u4#S%&;r6Z}T*_lBd{Qvs6Kzs!|E+3&3A7v5DF=P%4Tq4eDqkBTNRY*ZO+CyLC1h
z=W8%v)_8*l6em)7rU*!vM7VLeZcFcNs7L_`*RVbl$y0859dn^d1JI1X4=&0Q01V1J
zh--~GV<%+uSG9bGq5Af7)%>U^jOuRUp6-D|4}_Y{UM7Xaw6?zdEGe0<lfK-0Z#Yw!
zGpy$YGP<U+tANI46G=vvl7J>G<EoL53Vdt6r;`!*7IPbk7W5T|Xd!FlbQdKcXkHX@
z9NMO6OILvi1MK<DlXxhVM9vnekmg3YKDBE3p|4V+j?ult$~~!Y0V<Ul!Q=;YU*59{
zncU%)imZ``>Vef%f_$}q2E~gcn5?gUGI~xG`XWjBN-RCC7W?oHbv7Ojp%6QAlWZX*
zc&rHDI}!!s*PC&9t#6=5M$|JbCpi!ErYP9@a1(;qm?U93anjq;bObN%yzxqVTX;{I
zFxW2w&MyRJ*Dd5_<vI4<V9|p6*B`|(#JMLBn1)~UIu@1u(||QgJiJoxEaiPIq&K@p
z{&wO5@2S8NwRp^Pd`XZQDYgsKqKg!~R{uD-&Mbzx^ZAQ_?<YP;BPIx?6$nbLDI$1%
zOyj6B<vQyWd$?jkPM}|LgwpX9gD2o2zFu6E_kS|Zy^RpRPR#gX(S3XPwA0l90wG``
z!hlJU;jBb5G2XL(6uI}_3#|QG4wc~0;@Us6R)Eaw2uW;=fyj-&vo4=f8Nnj!2X>8$
zoswlGJ-&?r`Zj0O_1*|LN<5wwa``M&mmL_+$7juYEK%I(+nkk^E1hX83i~aaCg3Gu
zvzj`cf9szInWC={I$j>(X8qE~HCcaEU}F3+`^%affrk<qy1uALu9D1)!CvU*Rl}n(
zeVoI1{FGnldoV}I#y*mx-X)_5M~bZ~ZHoVMb+CMqi6ad@ze%F%(X!-HEwAq7(z(GY
zeZ%fpk(@x!z2M7^;E;?!@}z&MJgP5pnHZIag7R!Ul+kUVMp6Gt-1>2U?WLxmWk47Y
z&o45JmsETjnUV+odpnBrX<iD(psp<Y6u3d_PcdO$)(FiWI~NHm$-dRhLWvMjN&m4x
z$brtNViM#wSPbpnsX?PG-p^a~DKGI1KydN6;Ml=2vO&gtF7SRPS$D_Zy(G@XU?A1#
zFW8P!d)@hGfr<r@+xWdF7S*~ebfqS~UB8rl8@zrBFG`Frf*J|7`4o>rT|vo8-qc`4
zw!z8g7{lts>5_wISwO|}T9fH`B_2?~O88Q?EIMvBM%*FlwvPjUlY(QQr#9(Wd{=KH
zTfjkbmn&M;L&_}@G>0A0Hbe|jWye?OOCVR!{eQWRDokjnn)w{ENwzz@-i+4kzqXI7
z<Ertn42a}(=fmeiA{|um4d>@gHarPXNYT>TN=8f!^&R;@ZGj&4Gjc;ZzkyhbFz0|z
z!Ds2~!+A;~oGVEle?9VAQtbb>VkqZNLqvO!PzO|V{GD4=5O9moy7(njdqM^Q0y2%C
z2a+-}@IdPeM5t7puv(p;2RCQ8%`l4}K%P*Ef4unWmB~m2&=2r0AP9E@XAVeM;FDE(
z?r#Ev5dh8ZME)XL7bbT0i4s#8V467vo<|wUda@|GEuwx^Kd<B2s~W(e{LHM1nTy=E
zEFz?Q4SWLN=s@eBSNH7fMvA8)e~Fx<0I3a!4}RFx)YS9uL>U;q1B?MQB&gTtC4qg;
zDgBIf?G+{jqs{p;VE6p@{*|;@w9!0ZAQbZ5gaYN4LBKdNtGPD;86x0M0J)^KHSRCM
z$ODH-nLSPz3c3XrC{WgZU;^hO6V2YvR<DdN9=@Ji-5nu&>K_58hy!kytXS0aHZzkz
zRYm3VeMlmv!f+jmVv!!PVBKaLkmhH=r**=@Mr~kqw09-Fpb9mvzGeT9MLh3<_kjXY
z`KYEuMaB*2{X<6!K{4T~p1ibl0E+VjD%$^gn>O|6jRuF%Iv&9~OX)CPvEe(%RzRy{
z39BJHJ19slg>QXv@s1=E*9=ju0SoCqbr@0~3@nVhF8SxV*8da@yf{!f*~|eX1~Q@Z
z9xaFXCb%6IwT|zLT^JAl8p_S2c9r(d5_1W<eC7y)I**w>A@68V0BQI_=Y<&tZj|aa
z{i!1dr*`b}%<eg5$dfvpL-8dwEpNNiG9MNh#K=iWYKbRwS@OizbQ%CUd~~jU>coh!
z!WExa&81d60h2ov$|;_Yb2Q{+BFQS&b<)YqyO}w{J&ktunlEqqVb9@VZYV4$2wC`?
z!5@UHW21G0WSMY6Vp*BPl5^DOkB?<!hXsBa_TM@UVLi?>?0l|2__5873$gYKkq?yk
zT&`OaT)|Cjs;-WH#qtMh(~suX(Z=nzv)(87XTReU>D;f}xE<7ui+bP(a}C}&wBwDW
z5oho1s(^wBU6*R4jW`mTk0(PfZ|Ej6Uc!7QQ_cR{MW|(W`oALwmVuO$m+o_YR%&UV
zHT`TxORL4&yS-Sj+?RxNhM5<zIa_Fl_Cxt_fB;%a99U`=XgPFi-n?(`@G36OVlG14
zPwug^b1dNXe$Zs1kXpstyP@aF2t5=DeE%*DO&$!I1X;b9NKofKWD&Wp8JG3tXxrMd
zPs=i+@PigrD`%z*?6nvGhS^0l<<<rvW>;o!Kd8lhS5$O^K)?D?nxuz`2`vGyRiC8q
zD^#j&l%A5t!|}aq9v0;KDseNic2|%pdQfsk{n(xmY6mSMJtP(ep^GRK3(7J?ZLf{r
ztL{Oib{P+}wATyA1nzKy-%^4b@uMfM;xMq-q@*Ve@qQ$dkUIc>0qVdYgAPT!9imQI
zgVx&lI%&hpK=+@A<j)!f2??~tXoNuTJQ*u$z#QBd+{DJXHwLrp+eZb|%5MEVZ)z}_
zQzFTvzZ;S{NW&z^B160fl&5dNk_X>aD5(%Q(!gXYS*Zk01XM{u;nMfcpX*krVut@e
z$H(D;=o3IG41k3RgrYx?>Uo=$RST^v)+Y|3EzL}qIe2hdGb?8|4R?deXdcC71&IBa
z&QSc9^ZSG0p7K<1+kez`n(b3TKFYj42;gI*yI*U#H%&P|YK9XAm*3bLa9Wn7P4wP8
zhir6fPGQ74S(B?jF9HN@Z|xyOCu+*vo?$<0q>p_Wr^|@N@p%rutvF$e<p4zSok+Qq
zPL2X5en25sP~^QOZ^*0Wmn>v8@>)542o6Af1!HYdml2yLDLLb@6!OeNa^Gz%v$rVT
z-SXV>;<e)|%;7CkySUNr($5kzE&>9U*hTlPA*UrJ96Coc<W*6B5zhP%9#y4EK+ptO
zqUw0uT&)8ldEH&lqhorMgcWj^LX*2oEx$I!eJr-zX=wv`ofh+#xu27S`ePi#R2<|Y
zc9W1$;y|C6H<^#(^2<Wqp|F=u{N^*XT4q90Vc8DX`1^@Lun=z+ahIB=|K-i1<`E0)
zbPmwf2ncABrrGb-=za4@72~*ro{*!+df~{(+=gB8Bf*Xb@hg5*v<D4#Awj9E>}^>)
zl?LCe^PmOScu2i_p4LPCZ|;M+`sKQLgUg#yir814o-CD=#6NM@47(KT*3QEr?uFgf
zULJ<4e$VH=$QpFFdX07MdNgqV`aHdSZC$W~3`O}Ag_l~;6G8Kf{cbiXiB6}^@Y<7(
zg^z0eD+NewWC^E*!0Gg)4nDQLW!@F!0YO2z0bPmY)Z(fdgKO<@g>L|yfMz)#{nBKz
zPipi<QGq$DD4lDIUWTcKM;<aIU2sUegw6NSUzrvyg!JU)3BB)|pZ1%-VvnoWMD|%Y
zoL1f=bUiI~;o1E{{ci<X@l$vPLfoTuDaW6~xgI^|tpag@BC6FcS72T)vpzLnrs$K7
z`P||e3_(*Br@0=^RqH>)sYt--B<eXGy^(F7&g2}2U}&%5U}N8*+nwS|ow}*>NyQ1$
zZ`t2dIG8T29*>d`8p#$swW@6x`LVy@MNdgUpc=i=R+9F)U$TFt5Qz;eDdCccw;Ecm
zaco0)D`g4ha@JFuU!|)Yj&Gq0pLb=gkI&zGLH|T9k77xjiN`BoY5VIv1TyNGgX2P8
z%V7tK=4n?~J}gPdYQ6N@-O90t12ahp*e9rc>8xsG>hE3{Ouv<9WwvupkMEm;Qb`p8
zDyHtk86yIUtly6uGb|3RIdKRG6fY!s+}!(f;V#<GD7XaQkIK%ZpP;``o)K7Jp06*x
z5yBFVn{@noCy>`qTOawZU!xH0CnZTt&-K7;aYkNY{Jx7mJV8Pp4pMo9q9@<-d_F%~
z-*26OeKd8yEWQ5sC}sfALKQWW6zkAb<yoSpWzL&%cn0fjH;)%9F_P4)gj0;Kcwd(u
zsP_*4uSHmNSS>^xjN%g@1+Ue9{dy-)YR#yuqnM!{H$~!LYIVE>#TVf7Kb(rYv5ITF
zEF?B*0-6^isb~^7?(%e^<qY)=rr;x4X8Ks&ivHo0SRRGE5_Z=s8^?}v^b|EY!wBs^
zR=C6=VuoLw@Z#v%t^Oj8dIQW1&mb?Lr*i9RcLJW?w@Mww{^g5`^{iZvvN2-x6U3;=
z`$q{~L%x)?ecB*T*AhkbgUa~m2X>ms{OG|U$;Fzi-zWA;7!OQ?TO>aYY;E|-PTMbp
zrjO`Tz$N&GFhvhCVu?NCSbHL<N)k%Ka6!t#aj90(dnrg6-f*YK<-xB)y)4NOw6r0F
z6ntEcKk2j9O6!m?_KFgmpIy#Fud<|nl-yt!Imkow=Y5B{jK?-c3a)e^pEIpac~5S-
z#B6fqQLOTFaCimm5n>iO$Oj(4Cdrw;Hpnu5Pn@Kmn4o_ZS23BILJXS2!vjzju*jC!
zy57kapHHruB@c!<Em6MW<d03TMGw;V8O$*ccfX_E*SWi8YB;wgB{iE(TLT#sNEOTK
zkifG^7RH$HLQQw(UagMsYg1m+>p9363&$U_A^PIcHP~t2XJE50JYz*x0lnGT`Soov
zt8h64HtW*$wC65^TwV)gTK!x5PvgFX?NAwEULXTT$iun4XlyuitK!XwHKnNgS@y*H
z3=EiUH01~*RAeM|u=_I-by)Jzgx18`-Sxwpjp|=gs+f)X2#)gh%}OKv4y!PrG^g9C
za{#NLO}Q#Of)QkEWQ-@gTKU%7Gv%()(xx*_zc6i%dK-_HWE2k-e7`Im+31aL*C%aw
zJLjUC?w5ssp{^uF;`j1f@CxB?7gl!J_475N?{(g3e3G`L{YvauT@?_5oS8E9W_xtX
z?G7oo@JM~!a=jImB^3~fS>@D0P?!Ryqd!e2^8KxipCF4sie3bBjHE4t!h6J^&|iN)
z%A!*a+3T7YllVDBywkID$KgatI3(>dJmdCt`j&>w&~iG%d!0IR*2r#WA3xshbg)X4
zA1fB*yF*y%$&Xjz-zxXv*%Q=FOIW9RrT%u;IhDauV855QkUrR|szTzk)B=$8>`1q%
zyp3-~iS#X6{fl3PA1$$RbAF!4h0lHYf>3x<#5m1K_3X9wtNa&QmfZI~T%T5*l{{Vq
z%TRcMAF1C{w#?ZEHY4sSW~#@vcbHR9P#wQu2YlYp!J}{ZVB2%Afe_ewOd-xk{-g6{
z;WH+6EC?_@$^#4Gqsq+ku47^l&+|Sk1o7d4c;CSmw_^Xq2e<+SE9rTAvhG{mj41s^
z=Pi5mmbZB0v}BE*NfawARk~ybD3tMge0LT5XD=$%^9WhK!tD5LIMJ!xNq0@v4rDF6
z$C^^V_H<D-SM1B<Z(GA=d8qThbgt_kNmWhEE2<Ekptj5y=p%$XthNnnaSyT6378i4
zmXu+hAOLo!hjdA}(oLI9{ev+9fmATgC5qZb!v1-TETG;>`koyn1+AN}_hYo88Lf3Q
zxijhyF!Ju(53=$w=s&cssK?nGi4T7hVN{C5ZTzVxd^cgoZ`+I|@&#Fi1m!GGUwG>$
z6HX*mQinz&F5dDzt1Hec-kn&X(jjcj8?L1PRQpSKW<?oVZ=x*xU>dBmmVTS$Ei~`O
zSOU}be{6b`MF3o)D<@Us^0n^wdoQhi1=)*wV&4{>BQWE%ybR+Ms@YlFENpuzkvSN)
zMMIq6?Z32*u^qnBa3EJfI3ccwR?(w}-%NNFBl!ikOBYkqcC?T!HN{FW^+Zcb)OjGT
zLWPaJJVDyLyJ^0x^NkrsX*}9Fw;^Fe5LzQXzIeA$7u1r1;sxbbm#JOEF0`Qc!zoyY
zivurBkabuz&ZFqCq<PIbyx4=0R}|0HlOj{Dzat(_#B@a8(UFu?rEQiDU~l+1pzl14
zyg}-9&^u?<PsU8luP+_!=qPf(r+N24t6bX<^9DVa&u#dsTL~U`bxdfscjTLwPc}_$
zu_&Z;qc;SKhX@`Z5<=iqWH>ufep5kRQem8=VPj_FAM+={4GW2zk^JSCIa`m?_q&Q2
z4#QJnH6`Va#}62T2ROXS^5|gLO^69G2KK78>&jk_-{*8IXKloXN&O11XrUq}@7Y0W
zW4HjG+G9Cn&;O%&B|2c!;!cg$b(EO|6Q0xQT%0m~ZxxrwUUj@ivRJNH3PJN)be=7I
zcrQO|X*(dxfSN+!BW4KYA}u_Mbv8jRSys#;CN3nFp(cU2(v!d6-hFPzJJ@izy?&@7
zD~mBisLCr%DR2UNjlSKgbKSUYRS?#*_t3*rn<C=7bDciBwv?3rxc}=k9fiX^x`*R3
zgrUJfi-jX^u7BDK?wgsU-#_%0QdQ`8oDd7U<lQI^_-nIq-39w=`or}GMM3b79phP@
zgW}RsE_1KsDIB+Y1N~tsKd%u>gCcF!=%y;Nr_*6fAv3!NQqaHf(yp<INxHt>#Z)1s
zH1T>BCbQw;%elF_Cz2cWE-At(Jj2i4r^4HusNk5aGfQ)(9k%|&C4^P<%BFmsDNUt=
zEDEv)Nl#|-y~wqpW-z^%Nx0a4%(jYT>SD7I4PQhB-tDyK#%xAUxXlh}NeOn0Git_d
zrBwHntL-&3D^7Z{l9t)i5KcW2xaoU#CbYpc(PdtZp77X%oc`ReKpKKpXJD7`Z$wLE
zy6k)WZV&~_$9U#OBFXC4R?Pf=#D>qwwq74jWQN9s*x6R&FY$8-{M)!o`~h{ur=G3~
z^08Ks5&~O&>8BrqVkiy`#)Cu6fD2U|<^1OrzfGF1juEdZ=ri*#!I@d#(1y=f0Cdd*
zlY}+4L=hXU5GQ5R|7li&WUySD&d6B>+bz+qZd?TqOZFn-VXb1k1Jn!!Y%{AuQA2kv
zv+M7D2g7iwNnbnHEIt0`0WzRzA_&jBYqJkG{`p{?(BL-(%FOFG=5v$`jDQVt{~znj
zp6IGQ;0;f%KNZ)-MHGcG3cl)~%)VQYudO&GuSC+55)r8PH%Db1>oPR#9RYKk5U2Zm
zy;C<ksjVWRWQXc0cOlQHOEE^5McjtZ!!dsRqSwDx;XnL%-(T<$^R)jq?_hP-L?56X
z7~f~0i?ZI*jIHZe>f|t0Vcbabf0_43V3m5Jqb&#O%`;)M=2^Lmq{;TvHp?6<Y%O0D
z9uLLp9_RW?e`{65xUW)Idn&Q=b{=mtTrAhr>mN<ll=<C<ib$LB2~-Dr+=!ZI8AEbM
z(Qc%#7Q|$)*)rdZR5`wIp0j~v=ssrv+HB<O@h!NpT&MD@cRZKwmPlfsDKK~3D4r^(
z$m+O{v3j~R@QQcjm=^bq;%$bV^Q;*5p2c;A?W|cGf6fA*laHo7Yq%d8Q>q4DWUsL6
z{)7G$%j3^EPl`M;CvJXJWVm{)%=GaL1?WYrUo(Y_XCeyMvh&<V6NSnzW%n{tOE<>+
ztlk{?N1w)Y+=Y+T87a(i0H@+G**#I3h9XJm=n+ClEy%~?mSCBH@nI~9;1)~QV@M|M
zhi;!b$j^#@Nm=b#*}HoA_!e1ki~gS7watQm+Z3_pt=|}7k+=s-tt9(@22?$Ku#UC=
z6;R>m@pb$YP@x`@o}kXA|IfHK%fGAsisz9(#tm$ehDjC%)4u-l^k{E0+O)g2#7GA!
zVQdx(J~rH<`u2_`?)r=3prCZgfN&gr&)}T86?D9V0(}ln&3lMyS%=q>z+`TK;U-xf
z9L6kSDjctl&$i%ew{Mb}|HWGR;Nm%@6szAlNHjOn&C%Je^To!60)<2^4--rz$Crnz
z`5)qgN3TBIOs0e#DK?CF^b+^TI6Jk^hfcDxk5iA35Xlad2YXL;bN6-2Pm!PMaVD7R
zb<Wv$gP^b?3bRTZAz*Zz(PL1tBp<#kRGP`9GpKG$-XYf!7Z*F(D|VP1vHCFoqHp-Q
z4a+FB7Yd$ii!PJWe|rTzL-DlY`@eOzw+y`c)fq3;RqGAEeOu{ijCvsBE0dU^5$%CQ
z%4<~#C3DeW>ZGd_&<k6f&%`pRqI$R6w{>&#uP9qf0TY`Rp)Y7pnQt@ywqAGs-bB1=
zgS88wJZD;7_HF(?ODrAkL=r>B`1@WKH45Uj$jV#T{vtMF7@wkupJ#ist?4RyY@-pW
zR-`iFCe2L2Eg=^(1b(lcYG&g6Vn_Ya<m@eDPN^TF4GrSNH3B7>KhEIu!$^q+Ol3Up
zh7Qa??ESi*`_b3S+}F|GgKs}CM--V%_&dvi2YbuOYb$%%Zf9<sNgce>lu;4d5*E{O
zNK^1!M*(~iJmpu!)GiH=zCy}8JClo)h0lb3GHdO}=XX3iNfHJR$kWX4^cqMJNIfm-
zfA=^mQGE7EWzgTv-h>?bmeIse(?GntojTHi$YcdZpogKW){OU(j6%Aaf~!VGRgITT
zb68D-SF<1fCZ*osmYSRMy9$qoO>tNxWK&D^Gn-6gXj_ibn_cF1e{Xj1ENM@Q=(3?h
zo%@8?p+QV?W+QDqEpe}EI}qt<`YQBHftvfnTj7N9a~HekmTH}|szMvNc=sds{SEy;
zycK$Jv_P>Wn=pzceg^gSYEuS@+CMImXTO3%?mIK{a%~#)vTj?tGTh(^3o2;++|S!t
zz&Bgr$}Y8UVLmJ<7#Jy&UQuQ6;^o#P)hO>|-Kii8EZ1_r(f5bYsE~R7zE=Nb0m;VS
z36>mYhyp6Vsl_0!M#01vmT`Z6+ide>vt>kGDX&OWRX?w`$T-d0s@D2i<;cmTw!c0t
zne!#D2EX>N*3#1*$}HvBYcTo7O8Lqxr;)4f{`i-jbPi-tb_s8k_~{NX1xiBQa?gI#
zvzg;T=Nv_CR;xOLxcVy3>aM<!*hGbT^02loIh6*lAbD{e6N-<nDI=z5^Owjv8~i`t
zJYRme3EZ{;M|R*cr+Rl*;OjnvS99sv{x0GqO1%`q^)clNG&p&KIMr@gt^z4mwN(!`
z^Y31Vs_yjM>5|#oL=2%>UlC2umo5JMtZ7NAmDc@DqF$lGQZbJ~ZnuEesH|n|63e9c
zcNy|_rY|m|C?ipL7V6;m$&=XxI^>(P((@ugMnzqL)CpA2M@sMK@8U@r?WS%%PRVsA
z&Wu_X_AzE)WMm0}>{Q>J<_Sbq20p6+d$E5-9^2DF=q}n|bSTDsLv)d4&`>=-B`Q@A
ze`*@6#+-hI!sNdHAdgqW6g?R4vq;#S8g6<OPM0mCZKCtGVticj`q0{~Ca2DXN>~ay
z&_opG$>1H7lIGI43~^-0tV6i`<>R$`p{G&hN>6lH#fjNNq#FyfS;HiC&Rs+q(VY~T
zd7-BD<K9tnY+0lG+ofOr?5}g9W86!{WB+p>ZS0;$rM^eLGNpp$`jIO0-Iu?g#G~!-
zibI25(S|TR@!dV#&HZh%DvV{JXH~v#bUPU3E6~58;LSxD&U^a=Nc5_cB3;xe6s12Z
zvU+ykvc;vZE@)s+SVr{&U26cNGB*OA>E?_+5iv^`nz3MeC*W55EoSX;=2ERVsl1kd
z%jXj)J{WTYQw&LAeccZlBJjz0=KErE&Z2rWW}L6`6s5arR!B1x>7HW~6#TmC9C}Ap
z^@TCHq5*(@7eNnt_5u)8Xjc@dJG&=>YGSY`#!oQzdszQU1orIfYFyMK0ho<Y!fcUM
z@`@9tfQB~D%%M;p+2FSdr;dn2-QKaOoI{mUKG9imy{3zg+xmBAj@9oBN<#G9Mi80Y
zeSZ&!N}CSGw^S6%eFdj~obYx1bjuRiL|CpP%52x2A9guE{FYAM+aIYF-PS|UX-JGJ
z=OMiXN_CsAsaXl2Ab{(<h9i8&tCdvKMLm{VBkpkQb^czb9i+bF`o#^9(w*oFsrOfL
z14_?_>`;l3^3wNb>d$TKO(_@GWpm^6bQ-aehQ(j}Jl9!;ubWW#woYG`R%=I0lfeQ+
z1~o%-f_kS%X|7i6?)qq*e6G^UK>4EY`?Q}#0R5H!5n1=RF!1M{T_DSl?}i(h-zP&$
z`S&PYwH}_o3nRgQS(tN@tx^aCCsQ`!Y?JD$9~{h+XkSR%*BG=;W{z@kTnY*LhA^JY
zvEfm=CqUaJIByjnL~Z{*s@D|F>Y$<!q7|`y8Xs-8O-@0@Tv#VN*{Izum>M0EVC#~9
zyS2^ts`=v(Ma@%|f>~`B4QcMFY8iO+2ZXfOoW;Ab9uuEsqqyyiOgHNFQ&eQENy6_Z
zMIv8m@>(N~^15e(!^nxKJsOsN4PGwF?QJ-f?sc}@@&#(9>FG?LP=DRLgutJJcYj)I
z9rG7R5HUuT%VD|li1$0Lj?YgNOLvMxOBXC3*N?B-UO6ZZ`JLL(FCvQzy0e{VII!W<
zLVvq{;xV@p*)92D9)4|*VqbF!Ul|839Is(q<(*b*Rbx#<RzZ5M@9)|+db7H=PK}nU
z8%8qL|HB2KV(c!=CVruw=|Yeb7GFgp?7efkSU?akI#NR71b;&8IzYRR9G|IaX$?|z
zEaFoNiwRj~XXl1joUmlzR9CVea?k%XJ}zfcqL#5Ve6yc~UP@D)hfPyE@BU+6sebF$
zO|x)3T8Q2`U{Dnj;0?cBD@cr%NX_kD?BNvoEs`!gB@kfceW}fK=-JEI&bhab<)z)l
zix}@7cAfYoDP|1rOP;i^OcX#odiVJ2#1AB~Pv;3iK-;{a6a?O?sWHmha!+((^J25*
z&*jgCtgOF9@N*ZIKvy?JP*5(E!<6s^pduMkV~OkdEb1yBs)jA{>xU4o4)Qy$`9UV5
zGWJY4iGA{Cn;S|7kvMTj&tB7Q<5Y`@#%zI?H43SX9~tBU8quMJiP%R%XR{f9iKN)}
zxU4=Ose0SRELW~Q2>`EjYZsd~Qx7F*`IIo}#M!xRhA(p#W^(C&Wvs6MWo6EBSYE`M
zTwVRfJ6;E_vA~%V)K2{o)Xo+@`Bd#CxVec(AHgY^{>9GWg*}PWCcM^euz+j?+<7Zq
z$EGRp{Y{ttJiIv2c{Bc9R7aN~zRzM`N#3_@M}wQiqn}Uaw(a(*QLhfXZ>4<*Ab=um
zQwHvu;w2FqZP|*(#F&uByWbxS7g9Z0*JqjjNAT|{#<6fYnz@eco2!*lZJ%@&ddN^D
z@cJG;E}2yNLl_5vqUzd%$+h2G12H5|4mr846ULn$zHA-rSgk~3M&AU!M+%mcALflG
z^Rg_Dz0rtf{PRnSGc1?7e>NBgdabUX>`xg{JAJaVYhrfpBHo-;?RTrexTjV9<uv!-
zK;=>Boh$gzRK2(k<O3y4Fr^x`FpIbR?axDZexg6q1g5kZ-MsY^u}l{Pgh5NY^;$+v
zPIu<zYV9z9oTE~E&u#4CRuoIHwTX9LQM_QlBP#UZO%A7z5u*pUTXSU8u64@Q=|@1O
zR*!@GBguz=brty#CGHQz{i_jSm%q`@r*_)QF^Ff(&KYQ4<a5eFVwPuc*FtyN7uVB+
zn6@KMSX;R=h!ceK?bq>(M7>>3NOTy6^}}z4s{#nm*EoAmkG}gkK-aK8sbj6%5$(5U
zc6#%<)fp2`F4#k;9+Pt$_!Zer6^9PPmEF-s&$@o`{@$`%d)f2hhu|IRHWpUCuyjiv
zapKZXeqXZ0&puN?s4=fePc&TQR$T4533(3CCAl~DZ5bG@&)<LR`BI!yB;>*rRq8Zj
zVHDUM6cR31Ny%|6j@EQm2;O63+t)72mG;=dj-<5MTAcrEuwbzJV&J@!`BU*ngS=UR
zPO~f1ZxjlxY-7-m;-Nq3`7J%E&FSHR!eotpuyn3aCzg-dV$enqC+NIEBBAxCmT|Vp
z%0pI$3wKx<XlLYuaL6xHiuZ0Yj;guYko=8*=V5IvTqUg8AKj;h9yFM4&u>QLg<c+d
zh7Wol<9fcWGhG1f(|fhqx@ghXY$yB6N}6Pb+}+@uB@n&x71=I-)!7uO78_>fNU3ye
z@Q#-Ey>;8Md~V4P^$5VyO9c=_9L&xwUH_b$n~-%^!X3%oKaldZn~x58JRI>5kPhbi
znVLyiSyBgEc#&fKLlqu}%KT2_3O2)Ec!!Md-)-rR{7xzro}0Y-zUNX)ri0qc<kg|C
z-)4s8t4h%m(G+}c1J{fkaLdtNIyojIccL)c@eEP-(A**NOqd^^o!|zKx8#3A-kY1K
zxsGJFW!q<U4>!5wg|+B~%vitw^2h4_Mr~2#5XCdHm}n`*U-MrIE&j6Ow$H6P7SQjC
z!wdWF&H18eaLyRQx#vJcf^sM5S_F-<W`5)q4e=Dzq+CiVzkSxLV^TAB(sPXnAeqxQ
zQcN9J*=a)v2Ax+O?Yf4PlCzH_Gi-L-79Hrk&JINUb{Zu)FYZxlr0FJQJb?e=$e=ss
zx-5T%y!?Za{_amjkCg|AUd2Vn`e+!qshUtN^bN;<W(kpNmsHZgY6A}fg}Tf(amo`~
ztK7yDb%ZXh1aM&8SE#XtPH9ySzwP819x}r@u=~lE3~_EB{wedLO37<=h*6=EugGhW
zYN#@}XdoujA-r~El842d#}{ww;yCquqs<-=7s{gm(0ltSV?@&<jkIgYMvEmzi{a@n
ztK1!z=bckJmDL;T(>5xbtYR~KruKs8y?s{msUHR?B<EUvo;0)DPSiO{@+{QqUbyxc
zdCyM9iBo^z*d=Hs`yT4@CgBVy3JMA+Z}*scA>+I<^0QC=#yFgL+_^g@pjzzo>quX?
z&c~~X?EIj(8Yj-=%Y8MrrcZr+bhl-s10uQVFN)rq*i^mSawC#Da<Jk$<K=ztt{0I9
zJ0gjBJ<H-QUb{=)Y6M?NaMsTUT+){SWn!_I&}EN)lb=@$Pay|cKyg&^t5>ejR9vU^
zOP%A=V<?;byWn%Q0DYYtp(kP`R2B&)U&7n*DdpVmzbJZ=#NyG|nB01Z2}F(})_Pz#
zK8^Dqr4cbME3BJ@Fuf#-c&Df0=^(AE>C(m%dNJc;Pz-qqVRLc1>r&P9J05eb_3PX}
zyjRQ8f^@7zGR@CvtML}xep+OD>#44)s#DXu&uEI4FDd@;3FMd*S}r(j>W9TU=xSm=
z)$|04Y1+EFlA#%hgUGP6a9rcCOLyGrhqvOS`p<f6Y#t2_me|EP*!jJFWahC!ybe92
zPbepUF;vqv*H<twA%AZ!U_bu5bV}p5>saIRSlSOJce(m_5hx5_-X{GN=Kcv;B#A$v
z%P6GeoPdJYS|j`j>OEFRN}5MUFLzfACyyI4(1;ji?sDlXhe#J3rRKUR$K86i%B2)D
zamlrR``w+bauU)@C5kamxc`($LWibW`DhkgPb$?P2Wm!2a(({6N7BT|V^Ccr!zYfQ
z*AMN>7cW#ZCVvRJZ&`AS4jZ`d)E<#c=6|Kmb0!zczdm98HbvV~x1>XT<uOU!SW9i5
zjtg1cSe?31c8wc-iJkKOzdJZPAv*mj)$FvvxxdF9hGU|wWm*c<hYHM+)2%KxJ~*n~
ziYB`dy6o1!+^?Rv$cTP?YtS;vP*wFRZN)ly%BxpQv@tO;P9h7e$sSNuh3@ju2*(!?
z5&qCA&lkjcc9V^GAU7%6pm1pyhuk2&E>zE!rI4&fkDlE{N3i9w?P;lSZx=;XLf(pN
z?g_JhPI*q?o!OQj<SB5o-Po}#auCA{<c}NvY4_%O`=tBsV$9<+!{-)j3z<SDy9o}Y
z)fx`9&1v`&(wW%fLa9asG%xs+;vtAO!Owk9a|gYj17|c8Zvh@{mGEhOMmq1eRK6AE
z0ltLZ**&Ovr4=*mOw7J#FQl=)+J5Qh81y(rDzPgLkn?(ny_3uzmlhZ69Sn=NbJ~3n
zj3K#Uwqvz*fx*%`xA5&!?e^b(M^Vk;K0SBqLDTvlCCT|VS$B$7ItAa>;uuIu7OwR)
z+;b>wULd;5{dlhk_xrm?E0V>dHK&@14!d_&c{%B8IE%ij{e4;xlg_l#TsHl<Mvpyi
zTvH&iaJiOxxUp>J_3&~<9agI;O)|ZA<ZF6O&98x|nctCVP$CJ<?j%yb&kqg#{PlbI
zv3TOb=$(T#!sSbSqZe6k{QPKgb=$D@n>yW#fjj@Nx-*ZeIgk5zWHbg54JO3gBuN=V
z`%0xTR8qH8Duh&8iPjsM(I&|f?FucoRg3m+ltL=0Zu?%+y{+2!=lwPFoHKLIbIx;~
zKc4fP#~-@Kxo*GvT|V3Q`+2>kMENxQPA+ad;&=Go(|}BwYuoD6dfl%K>WJRA<R9nd
zVrd@aaY|A9Jf6#tsl9&X5zQsD!ffVpSllx8dD95vHKF&X?T++y9Dkc^F0g(c&%USc
z+fhro51S(*Jlq&JNkO^6GDHJuJo|{ByER%fz)l@$oWFJE!NZ4PT0MS&?e^KGFCOor
zKc;-ux1U*)uB;!TJL~tQ_nkR_y|Vj<zC>i&Sa0Ljb=OTQy;hy_B6q!Z_-{w^3!nRX
zZPMBo)p7U0*^Duf`2tPZb92Ib857YeRJZ(I>G0Djdp0HHXS~Z+5u3;=?pnvQ&$P#H
zM$eX<x8zP7Q{oidxf3Tf!8>PDs&Lb_sKZvO*=upj&pAnkYN4+?Z0CiHJ9HA<PHJiK
zSccC=raLP{RLabH;1oq6LR0I(%wAe?+%BK9MoTO<-SV6L#oC{?#7Z4&2OXz8!h7~4
zRGH}L?4Y3I*|M>hfYrlfGOoJg2I$d++LO$JBECm7G^KcbsE~D@_ZmsIX=s-)RW9?1
z7-$&E=Qh$as<E>k3>Z<JHm|b1QI__zys1{H^+D?9j?p^~*}>0hxMb5eciAaAr9QDZ
zBApra;Wv%EOw-xR%{gfcx0naY52%~fZqb{gF&xIX;``ynnN9tLCHvH?e;DpBI-E9f
zMrw=1stq-lR{58j%Hx<OxU@@)_s_*YHml7~TzY?U&xJ099fh2BogXTat<MAn1yQ~N
z3Rm}~?WI+S7g5Ow3b#xayC0*<JCIc>LU|PWVSIDdyI#$Csr-EF;AGc=ed6olbjIs)
zC$i4JeQc=~>fY2|Wph<QDAoe$$KDp1c=c;O(?1Fy**4Ml!c=MKdH%18tZLJ?!h*e5
zhV1Wz_&SmZu8`kuYdaujLKL(TCoMRbLRuM=kQVDnbm0Z}Wlyj|0|RB8SoGUs5zn9b
zzFpZQ;U_AVb*@=8Uo3c>y--y8fOp4F0eqT45LSCKDeWdVm}5MYaHVGa#fpNTjs25X
zFgAa)TcO!m3V+fdhXH~@8`>iw)wtiLRZhM7P~=BkrsnrnA?yXWpZ&mLgU%sZFCmo>
z+IP9@V<Er1tizY#DxuAKKR!CQjC13L&tdChr=BR+;rJ+H)bVppi-(+b1pP<8s)^vM
zNs5tNTm;U{@gzgL(^a(^a3H8J#)Fpusf_*mN5OyVu+73{XZNY1ZFCc#%jl%`Eb4$T
z_#iz#Nv66tTo=SDBNt(~7lX;9^z;X^r^Pi69lBY^PIQx8cHlT*dUSfUIWWleM!wq5
zV`8{G4i2`~NLgp%Hm$zO(i=+TY_Uh`Ud9?gAhKw=c3=sU6UdY`n^A(%*Fx9pI&zjs
zrB+v0t3o@iK5Gs5d;QLq9JO!8g~uqL5CkbMjQB(Md2-qn&8J6!)<Tav^hg`-E~80Q
zk^^Cm?McOWGr?pST4z6B*avA8ji(Xc{4jjsqyX+D;*S!pF_1q}OP3Jt00##bj=0Fr
zcAO88pB!6o9z=d#=fq_N^1_y5>9_t$4sP6gBmYpi<Uc=z<Nv{jC}rkya2=<LZ}4sS
z%8Wd|Yn`_jniaHp-vsg*EU7-w;NquiU_jsX!|d>-$I-$R>{;)_PXz@9U%S6G^E~1x
z0paVVH?IG(qH^he@{V_<+pq8T84vK88tVnDVaR<k*J}=B6{6HNAjV}?1?8!gD_7zQ
zv*O&aa<1h;!*oysLFx2<>kmiPWSv_+X#a5o;VQ#YQBrZQ_QQu`xn&1Kv`?Ot0#}v9
zD>aZwI(<!=<5G$67#VR6ssG7H2L||B0)AIfUjAbMb`8+AO-9^}QE2~Ed8%bpr{>3Z
z{+j4TEJh!w%8Bsw)x@nAaeTD_SJp)V*$Tk5UYuW(Cjx`(rFF}03IcIZ!qaoNkm~Z^
zt0nZ%f4fSq<gQd|x%sNyGC4mmEX>Hjpb`C$JOYPLfmRTVit@%z0}YJ|sC*$jh39Gp
zQXMR`_@o?q$yUoYFz5*uy&`6@XRmu7j5;;##{Cp?&sSBQu2=W1FwESxFzg9s?(tQQ
zz<bY*^$FN1YIUe=12)_Q!tIcYjYj`+MrO&V5P$VL+bsjY!KLReuK|5;k&Sm?_r%4<
z4$sbGs9J$3MVbM_*E{0apTuov0jS#=xnURwuH7Oc*o|bcfVV)qLQu3cM8N510ECip
z96lYEYzu8_@bwM-H%qMuxw<J-EZl74{ghh6l+y=y@zTu-Z~wF(WJbXPq1{_pppoL@
zGQr>hx#*`pGUr{IzAV&BV0h#UB$)Ae?V$b=Q6q?6nU>y>htw>BWk*nkA&|<Q8?P*f
z9LNa>-PJ{p0qt5X*?c!V+<H?9K{7&;g8{uVsJs3oap|%6S1{`+Um1(D40Kf_k5Fpe
z4JB>YNeuiVWh7TAw)0H<B-lyJhD9{}DnR%bzipuEsb}r_%Jo0qfj+=AQ0Hi29o6CN
zoL=v5zCwzpuCO!=AB9vxe&4<ks3E?4`*v*rP;*3AGdLPT-fCzkPR~^;ehR4c_AT6K
zQ?><qs;{q)5@p^<Gz1~x9Q+xq1^QtF_iTu*wHW$F?P5M&uJJm0>n+{w)@0yBW98X-
z#>vs<;BjG)I^E~%+^K%Io4kK8(+Aw%2&%X(4t1a}e9<xe`_XupC=7z1a8LzE!}|{&
z=+rhq!~s(m932!EW@TdTWdftQJ^;MXx-9DfkN}e#cT-^X!HE(VuVs%jM=e%|{-G|r
zN@HVXn(vYN3;7Tm18!y7e{pIJJ4D@k2)`T{B<ZXDk4w9bXlNwOuwrgGE)<&@Eag_c
z?k@|D^+rVhhikuNIwvG5tp*=OK;S7FTEu6J#2$HVx^+feGgRtK@P)9<`naUOmT_OF
z6P9rkk>kvVG$T|Oa=t|Q(q+1wqF(D;=Z%!q6;S0l^)^;UI37D%7@4(PVG!&A^xIz!
zqNh3E;g`2iR`+s2zmG}SJOTN7jSaRmSr`c2LFfa%nLFtILK>lHafJ$9FICToh*r_f
z#Q(Vjn#NwQ+zXZloL!Q$nVhChy24*85|<vgU4!9Mc{ZV!&2&>25fQP9T~U#a_9x^N
zzHfn9PUL6rtwZUczyFd^9jqPbf<T`H0A4)TEu<I(q@8%ezrN@b-1aBV3t>H?8ME;P
z$MD#TZvU7;(sztJL40QYx^*hg7R2p2j_%?mN)%8rUklM>z)a}pyAv3vfWfzcZZ7&>
zzvLKeX^mi|qD^TErg#=K;sn)u|JsPF2_T$G&Ho&1X)#ths-ntX|M!r7KWrz2ib7VF
z)Y|WBzGN2sAYZGhtMm|EIt;K8K?^iap1if$I;;EMQM!aWTDL)?LtsXY`7k}dfPasK
zuq<pi5wM1$w!rG}e=H#;WE}wtwLz;?#?Y#D*i?y{a_*h`M3;DW?!0POnxjoRVEkzD
zLZYJOw}RW(!jds&TZU+pxothYy(YG=zC$*DYJ5zL0?6nVq_6`kTNBQ3ac@Zp{?}S%
z?lH>IQ2|i)gp2mSW>5g%PF$R-g8muHF~AbNe?yD*rPvi3cON|>GS+BfJ9ptit9_X$
z=kiq-r!K?J?4p48L~c`kDm+r3Ui8u6O;P{ZbuWeSr06O+ij>05Wx%<(=pSP!wRGDx
zV9Hdx!4XTYVpugv@Wk2e=Px_~3sT6rhL@N3SeVQ>Y#V)lzMigb0q_@)s)#mc*250(
zrC>OjzrXzJ3*Fh!OEzD0f<ean1LeRG#aWaq+IPA}MU8v+YRW|m;!tfWYrl1~6i)*f
zapO&%4*~;$f^8YNsu_IXUBRcd3<y{_VUn;iny)rlF&l?6rd<m%A$EpPx(y6wKXu4>
z2Y8JQwY6DD{IwGdLWwa9Nrl8CkMv~oWlI~MQEwlnP`KjA<G}06JndZ-5G*1}>N+}O
zANE%g*mdObFJPeht5AULiEJCnBqk5iY+i3+<Icuxr}#ucnn76w>fYw;SqGjK%RRuH
znE?bCU4C@yy5?`A+kVkbA@}hgR%!<iz5-fOT%C>OGLUGepj9<f{Fl$#^?XAAT#69@
zL$G?(6DL4cz8e(egbE5WCm`Jv=Q`Fj^YUi=bjz-ms;4iUuK^kg4NT9(0HF3+KoqO)
z5Eg#dI|}Uhrvv54R1t^x@yqA1g#R_<hY4mHkj5+EONTplddqMXPpkZ?JU{HIxJOjf
zt>e#)XxP&+)&6S%3w`2;8e*=ck~-o7FyxaG5<oRShgK_ic=X;4!hIDNtK=S4mU~}T
zMl2Njr@Idx8~{F!M~mtJBLNPoFigsxJ*$>2h4LF86am45HtMp-{N@RktND&LyxDO{
z09rc`mfjDYH1IZ=(%UQGLM>k1{4VnNMTp#HEFSFo+cl~EXQY}$V4VM}5vtPBO^UR<
zs9io$a}$}2ME7o}_6VqIp@M1<o#H&#r|F(*A+f4x(4c||CqzT(0t|?jp-IKDkILg~
zXei7*)UI~O2c;a*Nj|m%f<?n<8Xg+AA(d9Tz>6{vmF31?Gz3Ev4N)_d5Gq8MB?>!c
zK{>07pa-Q7y)5Xyl5mOu=;K)6eZ@vtU&G=0%Rdf90?_5u^Zw!0jh}8MfIA4AOR+~0
zkcgxLbz$AUJ~@X)M~`IfmpvjPQ0^KAqy<ux2B%KB7Oj><xDwrw3}oELi^vErnj_9}
z1_2T*DTCaV<|4=j&Ey?faWPP2Z5tMXF2I7B;;epH2tTR}NK{B0EEEUu#fI=44a@#q
z3XF=c_Ln73df=dB1q75;(5kAZqjT<>w3X@sz6}0QE$8Hj%3c=)IlQVo0YF+v)RvL|
zh5iN(QP{@;Z=vwD0eN|Oi_;N{<}W5iMMPFE_Yj@9YF6~i?YulWjMZJljzHo1yhR{)
zDDrMg5Aqi#ZXs}pvU1o-ArQs&nj5fjZ%R3%tQ>CNS-gV1&`VT9@*RNsW-{S}sSc38
z6!y6XU0C$>f|-WhGUZx;zcGyqUE3B@N?>jlSb7W=AcKbv9YS4tFFZE3Ayc~u*o>Kp
z0ru!LsL5h^!>>1*xQ^&}08)e)Rq^&V#RAL&b_=MN3<OpHS8BUFE9*JI+e3uC1K?9p
zdw<Oy7dw@e@^fwvW*hU$M)V0+2TTJvZI1ARq|iI_SdHIFMe6}_yf+pGvaF4aDOF`-
z#?Oc_6Ca^Y$J0|Dv|mk&m(jM3+cSQWsAp2I<?R&}O)Tj97^%#R4zbROqo+N4)*b1|
zA1vo&kEYu>8boV7$fAJjldSc}+;{%Dg5x5sH|$!t7uG|`fpR=QE|Ky;it6Hyb#eQ?
zdMeIEdOZUyW_*hxt$C)Hyw`&96~TN$qO`O$H4G906hyE7C)ZeH^@h7ZHoP8CqQ;PL
zw}}S&Mkf%#6Dp7dT?4y(q8;l71i`vkGcc|YdB@A2)Ovu@b$ZEZT%$b)oe%1eAcg2M
zXyz)SLHYqIU9(1~#OkUw6^+FEA=8>9|M`yWVC>=>2oqWRxpsOBv3}a%`zE2K345gE
z;UrSE1{RS)9*<eX7NtAChoTGJhJ*%*3mASwdHiPC+9(g;$Kw};kWg2=<{_&ezA<Zd
zFjw^)c9Ew1xXx&P6{gI@*qBx>Ubja-^&G0(&;v;(VV>MXy(!vQX-I1PBwME-)&TYK
ziI+{*X=1$v7z0>C%RDgRE(7FKfrtU(+f%iACiw#|n_ypeP{7e6LGpM2!stAtlL)CO
zlr~^H(Hc8V76b|AA#HK2IuD;l_FJO^ltM=!tv2-Oc^NFi3|>|Si^nfSuwhV9BhUaP
zwrQZCFFJ(a{7E5IvgOMCrhz*ulxxj)YPFB}uu+rs2SUDQOPXj8jAGqZHR>r&ztiN?
zYZZKS>a6?}B#Y^c*TBfp`SEpL&O}2_)XOW0jn+?|04QM)OdL==G{z#nsY{oiB-G~T
z>P`I}1{$d<;jUlbL8$D&yEQ08blwbW<L6Kc(VG80&ph~s(i0%KDe!5>=ig>GJX32o
ztDA}D;8^(+;R2|3lUY@m!YZZD%(;IKGNaqfOjo*Fk3&)L9V?EeGl#J6&f%w24-C_<
zD7sr}>+0%K7m3y<B`1jp&CU2iaVG^pP)g_MG`-Gz>y6CLivu3h4nBvUVqVt<Ajv|!
zs?(MZ=g-cXMDR%0GZZL{(4ALnN3?^}iV>VTz3(0Mep~Flp7Vy%)3UVaR}X}>>Hht#
z`MF+1T}ox=vB$^V2q`{MG9wsbe)p&>BGuql4^UHLnjG!x{fGm@F{sjx_@f4Mnlri1
zGEBh;L6F0rSMLMhY~p9wV}y*D>4DU^#<6YNTss%jBQ!LWFBt`&2#_XU&YW+>{&U0F
zCWPa&Q&`@JrSrU+p{|M2S<=D}AbzFLT9vjBXENp!OW37nUyyo{u(0sz=}2Q7@gZ&V
zeapE<?iZ+vvqJF1w#XNK-a?pXnh=@0Ygv<JCJQIx1f`{&P?PhQpE$XTHwVci+L0lK
zx7nx_j%YUpdaJQHs}^8KkTY)xkR8lYR#AY^meTgqR$w>H<M2?WJGD$K%~=7I03N&C
z7Uq)a-SF_R!MRBYfgq0;Hcxz=9c~PBOgxsbY80<KB|w4bygt(9%Xx25X0&KNRW6fe
zKw{dnVf#&<vtEy%+@ss&K-|XcPAgdFi~?>*J;i5eGyzJG$ZH{3$tQ+nk^RMPAS7P>
z@h@j^Ft>UxG9d<?#jUnQt1~iNuT-lu8@NPmS4#SqRu{Oru<3sseq<J=y6yXAAmFTz
zFO_jGKxRvJo9d1&+<I1YnXp!$c>vhT)^dAVwrLMzB|1a1cqn(MAZ_6wXoz{JoL#rV
z3O~Y58MVr6!SZ8vC%3^}@X6C>%1alddqXeN)ia2YQ^f+}V-uCAs#ph3S9ZHCdb8Pi
z-eF!umsbTxhFVV-Z{Jymgh5jU;^3#lUKuN+MPOX~6&BIvh2R)_#CP)<fhcMaJ4ZBU
zx3A5Dr?ouA;U3*%bfrp_nLUUE@1=7B2V@nB;Rc~9qF<0dn9n+hlS#GPo6d~84!N|A
z&dyWptJNj(P!j=a{Q3-<yk#Te1_ShJ6iOh+aqfRns?rTjUt-k^zb3N1*%`G?4ClBF
zd5KaM=j&j^nF<5Vsh}W!E4q)!Fa_IIpUW?c`<>ocvttQX*ob0$57W@b4M74P!rU!+
z1dB{(@%PSSW%%^~*Wb06e37<v?z@erwo{ls9#pQ@M6O4t?2ptotjFvcrC-znHcRy!
zY%zkzHoWcby?eicZb2SNJn#uBP#U<n&mlYCZ>A`Vm|e9yyEnWk^}KIOPfcRUy9p}2
zf<KdcDN%0uIhnWt?ZJ2*#;v(MN7wPNOB`RF9zJF=dV7bUp!{$lC$kGho!WZm!#(=F
zMMRsM_Xor;1I^jmDmyhocTw)|R8N&qr365m1%iJk=tdkiK`_@S7Xe#D*w}_<HJ*ne
zflwp?7V2aPcJ7cmX(!QB%6&SWY}IJ<TW99*u<>s5)8@4bPg=Sy+L#q*w_0uBVW$rR
zKIZcl)f!C~PO3<;#2tsG3GIhFC|qm!6L!bx`vKeUe0)B{_wWNVqvbfWPy43uZWDX~
z^O{)qVyWx)we^TUO1HUR3XZ1J!M=ze3Fi0jeJ_&iiUkX7FVGgVungrtOyj@IX1dR2
z-n@A;din-fek~~Fo=!vlfh1{wl8XF5alTJ;{PqF$w5z>%u;BPq8+dFphY3STy`#VX
z4pBT0uahg^Y?ZQoQXxaSS&Dw)8xd4>&`oWD6hv5QJ?D)=EKk*E5TG<06<e+4;Cln-
zV`;)R5C~ufjG_#ohat4$11f7=3?d2{V%N9HUQNiz3+*9>g?lS~K~Mwc6pXqO1Xl{P
z{b`hY5fLV)f=(FKQ)E=zY-ppK*)qd$%tKhIF??5%RPf1n*iJ!r#=5oho7-uOa{*3^
zX>gIs^+)2v4%A;p@OCDROeBz@Ms-7g8Keo97=$E3p3P##7|YnP<2&|i8}{ozE>Oqi
zRh^us^#h6aYutwR#Fg!$%!5<lyBJ94u@*)H5g;Pb?1Ev7pEo9TMt+?|8p`ETd?<<c
zLr?DFRs>z5iEdT)h#V3#N{Ja}>G_95JGHq@>c^auj7MHS*SV1q8{UB_O3wZqeDMIC
z&a!czHH^Ou4Kt=&3vH%4LbpY#WgWMtJ6;Apd?@U5qU+^=8B0xZCcq(fQIauT9<3fi
zw*k<U_9fC&g^aqxg~kgf=8f(AM3Py*P9erRnjtWZD+e}BFV!uKI20{vcJZK+c7DKR
zmVxtCrU$A|L;W-bTqWAe=l&yw<H?G0%sd{!fj?K5>Hwi1cq@n`Hvkko2S4#O^l}9e
z%7WHUs_V2pm0J1r+1M>d_-JTBLwYj-fhv@#rWY567QvxV?IxpS@Y+c00r5)M|GW0=
zN!QJWW{4|RCbFG)5;fr^#h%<?$UJ6)*&&CP2n)jzqT&H0TOFS|d~`vug{ojw#?3}j
z>`BZ1P|)!ZrPoJ+fx3(?x32d@w2Dq=GK-6gv845^Qms{Nwsvc3Y9_yU)}}7mGKQ?=
zJC9eW51bTvmtoVkfYpOEz~tqnpz-&l4D!+q&dGpjZUcczDHo>JN4&LW|7Q|84>YUp
z5EsvwiAFA;12g47yA5Z-$mA!u9>2^x^=+Wfi?>Fx&J7_tODc(`&nD#~*w?MRTq!R^
zf!(AHzv5zlD1yrd^H$v03^^LydZPSEYOSG7en!b@_IXmht{z|nhdZ<4WRCu>scDjB
zd&ch-bIqq=6DRO(GD{Rkir0_(Z$X6@4X8wx5N-^J|J&jhAh0YhL7lH&Z0AypoiB!;
zn$?laS41`0$JclBlr}COPL<+3VeMWNY}`jf+G?_i^eh780;~XdbmJetBDyq2JQuMr
z7E^8jr4;A(8(b1l{TRYrB8a$_+X6+`tp3!pR$O|4c%_qEIA2%Hv7l582#(b(5zV7c
zicE);Gf^)mry`^hA4kvWx`PT+{Rm_*^RP&1jJBakkT@3u4Xvk9h(<nNtuthAz#LzU
zHebLukShC17qd>LEy;T$<7<p{YvTkt8%0<H94C<dj<Pax3>Xa5Jgnh*!6d=cS9~=?
zGCZQZPtz{>G>1~NjD7cSD6o*wPWLQo14{A4bKp<ml5g~Hy!t;q0UKUyO6FWg9_Zkr
oNv6QZjY~NSuah&D|Ke|@+lzj)qn=*hPUDWKYN#X~I(_p$0C;OzW&i*H

literal 50778
zcmce;by$>Z+dev&C@8R`6a*ZXfutfOsSXy6bSg?qcVnO<D2NgwDM}5(NDLvVf~b^8
zNh2W=LwD}$epuiBe*f;{*n4}t@3F>-`?=$a^E|KX_OY7E<$Ze(?Zx46`>tLgXy9-=
z<#D*}6|~#n9q%245ctm?hbwx{INae!=>IgKL(^tB91rd);k>3t?0C1;jRQM3zD)NH
zst6^9U$h%yRaF~QV^UM!!QZzxXm?N!SIy|CSHQao!K8N+8;vZkGCvgbB*QgS)R_r`
z|Acew-L|u+k}L0r*yR_sn^p2unq=jlD#w@ojQ72``lpTNx$L*Wg9=(&G%ybwF0AhD
zBK88;*F`&py<*4FaA7aD^Hgj{FK*y=cwleP2jCdcivSwNKJ4ehxNXwdC%fbWcB2;+
zxc>)(=+1Ff;GCSCG__jH+8nVzc&Gl#J><5G8v09YY;4005>C#}^IVl_&tJSyKK_D4
zl9ZJ6YB1IO_U)TpL2+8zmC5t;-^_E59Xs}zSTl5rpTF6g^Wx8t-_rxt>we`ux$GJm
z8qe#@4={z?{PFzxb7qBa4GnGHj2F*DE5Q(W4MW3s1E-a)UTw5#!(AIV@c+pw|6kdv
z|Gx)$|5pYbXMl*hR#Vu;jVHGK^Q1j{t~qTNO%Lf2q4{wDGZCW1pNqRW71?5Pa(z<U
z{yllkag|P{ZBZvqMdrlzL~hJB81?w&%p`{WkKdJ#rjV@be%WmdSvC?^n+rYSE^SXd
zQMMhK6oIj7$caDNM7A`+{Cf?>f8J9dbO+A`Y}o;{rrr5h9gF!oZ!D?|5&Saiw(~gs
z&nFv_vz!KLq~*T5`q4Y(Zy&)dmWja2FFh=l^68%b%N_D>x84k^G?yDbPFnBk#1UmN
zgXX!n#36QWSnTe@r#3hL!G%4<tee1SlB!5hKh@6j>}N%9;nn~1asBqcB%Wz5z7vnB
zo8YiZm;Sw-sLfi6S^gBqY7kjzxw^sQ#i+}NRfYv|8}*u_G(^U&M``Hh+vaDQ9K|P{
zjJH{RWvH+)ZlV=>S^3J`SzBYD#$;0W>%}KHmy4Ji;qNU^ov)6fqnOz@824n!C>_6k
z!6}Ex%*N(DJL?*^KVV~NKk~aoxu(ZH7AtqECfUOJ-44riXEl1`aZHb}b^KMD_B!lr
zeo`S2-*4{{KOCuPJ&MyLU=MIfCWQoA{TVk=AShpP*~gWo|IVvn2hTrDm|M%Vm+M)#
z7aI^XgR4Y6UcP#nwn)I?nXx1P@zk}`yz?Oij}PiETGsX-IPRsogU4y>hrTZvxbrr(
zrTmQ?AE~U=Z|C{6_5R`T^f-I}*QuQ?=*vH`Hv>Af%iPK~Gn?zTOPs(gMXfk3ZZ{+I
z*6R%Pnse*5H+p?`>$T$d=6aLvm<~@|%LQyH*7rvUWzG}tdaZ}aqv<eF`mJMZmZ%zY
zO{r5eRJcHu4!E)P%tI+?-6{Om!!!>bV7t`Jzd9q74&$HO8vp*gPO)}B-4p0XYFj^g
zWh(3QV|oa08x2oAg2yaO^iz1kaIBOfn%VKK6JBE`w@JDmr_AC$5wW2k81Tr$DDRJA
zzqo0>Y0is24B7hddUVI~MmnsXmJNF{qu8H|a#79o-dPhyT^kpChiMp<u~;ha@p5K1
zb|hEPdbkvQop0;wyUeVIImD;uRFmDf<UO`#`Ya+NPW5n#K0i9ljx85aQ<q{$WZioG
zObuSUZN2V`NRKOj2rClZn(dBs7itN)=~z!jRx21$;}dL}J1$OpT(chDt{hXmy001D
zSKpexrpZXh@;TRF4SStB^HPg#j6+*<rL;%+TeUB~Af{=@*h~qF3K~y7#u4PU_Ex4f
zGv<z*pAh)K1kL^ReTPP5m;zoVrjbaC^kMbGv%En`HCYhm^T*_QDta{?@^1ZITt(5p
zHnm~bAp5SmJ7O86L$o{QxGJA9E1Zs2!hbqV|H_SiS%@?5d7CN3_`J=k#yJz+?(zn?
zd4bHr`eyrpbe-Nq-Ej;0?<SiPqwRO&!agP~;&%1TF3)vPs!X$W`SJaWtIMTFBv^y1
z?ubFs8IvKZTa_1)NZ(=b$T%0D<EPKQ_;x$Do$%lm`&lacijQm$?f#HD^QLXm)xwtA
z8CvS$e?1RN#%Cxckdobt@^Z_oDI2fW8krS%_e`DRy0}f6@d)kIu(qylS{t34p<yg*
zpOIix+gOX)t~bHK!Ojw#J9qB%VldS_R$k;pf8RW}y~NmFT|KPlK0D;Drfoiu8E@Wv
z-7~=c+>JgsD9Gcegv8aWS3Be`{7+VU9MZ;AHSl<K@M3b>6~dbGtD=>jcdb(mW@n<u
z-HM4z>wbqrUw{7n1Iuz!D9g*USW;%&Q>}-M?mt2X61inS&bdOHNt)C(2XK&PS?tst
z_G71(#l{Ohi~@lv6BY>4o-F^Mss5?Zku%&OTP8-GkJ~L@yX71fX5Sl}QyN;WhkfG&
z=iR8tD^1z3UrC5m4LGLMX&MSLO^aLonGlx;`>5RVZHMD`18)81ghvzhZ#{Zf$dHNt
z0+k89MZaZ3@vdW+#zK38!IF9IZCUb}+DH2(A_ZF`W!e^yGd$RWE#>KIQoR>p_`a=>
z;;<LI4}TDw1|v4)mdo?t`*jy~{vV%>wLWcqI;MbJy9l#tqAU2Cw~;D1^H(f>$}90*
zm-25+?%^vQlVQTeZVj5PIPDQ@C8E+&FK#<g76z^r7L1%IK$nG9g0=RvnZ@HPJn12P
z^KNK8XR!5{hTm#ui>#HwCw1JJc<qi>wG}fNgWqwqp-@_L!ylXYVQk{3t4sXJ@RN=+
zFySZnu;G81ey?*Byz)X`9;STbBo>V-8LOyHu+Lo(F@(9c9(9SQFLpC}#3KZrQW3DV
zGL{fCla~y)(G#9L#zuR2V|{uRg6VGh*KS+^zp*W5T(4bSJQwMoF5vBDG-=bV((8%i
zp~W6VkVK2wjXfdhAPkeV3!7+TFsxr4i?IP|LrY$J{*RBoEev>#rR>AQ+=%~djdEmm
zqjX`&#Ih!S%q|hmkQqBeo{sRWIH9t?8j~#DbMdsePZ<6TNJ}a-tj{}mj55oG9!|IQ
zaANPI;S?+{3oIKmY98yY$;)uol%2dTHTk3o;ukH)*0&!}t0z&|uiDy&U4PcSsWs)z
zk?6e$>^))W>vkc|A2E9r$A0vL3L(+}*+t9NH_pK0E4K`j{vCW~LyK0qWeu0iR&N?X
zkd)t!<|u*<K>sbasBNQ8!?wL?^%8npW^3H)@EEG0u7;F<bU}?4q)43TR+PQ3ZYXHc
zQG1;-Tc=-WU!nlUoG6AZ>c&aGa_0>!1{<Xp(7|I(w`2Q$35+>#dAJGf!F8~vt!RIP
zy}YG9MhR`WD7NA37a{#@ZS^}09#!CAU0Y}#AJ^|BIgCuP6^jat?Rg!`(rBq$4ugL+
zFm?<=2Nl_I<(3`eqijFxkNx2${O>3cW}C(UiSo=;CpmoC@><-va6h(!K-gDUudc%f
zw~hXqlDOhgaG}3gx(Og&FLT@21;cI~^p$}{sbPP8lA9kptdM8RR1Q_QDb~tOvSqsY
zeRXIVK}m`DpDjL{q0I8aA}e^ez8aK6nO%n&1^nDPW!jWjC(5i2dLV%n+j@LJ#@Fae
zRB4>%7HTM6UU!_8d*jB1J9Qbsi-3Ag6Mq%>Ja!B5dgWxA-0EI?gAUqZKlYTrUvm7(
z>z!nZnfoLMm`}hp?1T~-uB+tV^1Di18>-q`BVjAd%_X+fdqYlL9g|_i5w^C}`Y?>A
z=8KKjEC%cU&%2sQu!4(d&{J4wO<d4Dqu5fvI5J$E__S;m7jPGQmYSlHWP8ZYWb0v}
zat$z&iYwT9(^z0H<Sn;ed;0lzp+n^Pck3w<)olqe3=?eVtsPr$@y{3Aquu&}9lP2K
zzm8<TqALb(VK*+EN4p`fe0u11Ms6w?7VhTOJ9Q2$l1=8s&SU6N5?iy`4k3`6jp?dY
z-ytQ~x(DAO<H8@}g%C1)5-d{hvUsw^w8@hPCoC&_7mfR}z0Ab6+ll06e9*0+***~x
z2ID?#-n1ta|8T8-H|xkLR+Y2r)<%)%Kfe}B8RxIQmj~VvSFC|%v=y6Q76!#STb$x5
zATAkD+!o8!C-35TNZ8C6_roTXs8YZyoqXFWj+plADC1~P#3L@uxGRJ#E6y#=gC~6+
zIdsO*Rbn0@PSQTit6t~$mXqGP^@#q>moHsG9;%L^rhsR<XBL0y2)0Fb``XVGNSmPP
z>2G;W$xq5`PrVMSiCbXL0UG({oWC*^MMd2cBTcDfqvD0i#|JVE3d+WFYV3-sV!7>x
z5vv<Z1EF-R{1PsI)%^VY2EIJ)EU=gJS(;jzmrSJtfq<SlDg&ZSs=Hk2iEROvtRpnC
zQc~WZ9_)#Zj>c=Gs>D&@2fby!vw!MtRfi}mDZN`O+gzXYZ~FO@P-&X|_U&6vX%C0Z
zjrCri*}4aN4xEDraWnzNC!8+tm9!T_HfV!izv?zK_zi}PJLkI6cb`EPrnR{?ELRsJ
z-kWdR*_?izuY6#lyLh-h&PFVT+j~4)FW0<ub)vW_RkeI&G((1~Ves3Fbw7V<EhIn2
zOimL4SuLqamk$rS&Vw~H=Mnc$9Xxo@pukS~$`y8TJHu<&f{16Wehw}VCjhUii4tn~
z{+%>frWhrtrRBOb_1kxKfuf&@|A(7<vs!NBw4mU`-^MqO_8%rCC&0P!9BOaP%&}`f
zi8mE1-qpjJ7WEEU&pQ5G1=+Y%pO979AN13rM%=nN4K{^D$}%d>f4;i0Hrs%A8cB|d
zG#+t~cqM9c9FKe;AW;v_vs}tZsyCGzZHE+oMQX}>ae^hRK29pipts0**tRp@svy4Z
z$LotST&lsW0@uizTJta?yhedtw*vzl6ypJ<cD<0D&MIy7?f^(|*2;w4OGht<5Q%Ox
zK2{yCK0jj6aq5MML>dgZ70WI4-G6aHHUCcg%qKdbcOGoyHt#P@Gl%zIW>K8{T<`K>
z&zIY0Mnvy7@&@7YoTZ6YyVWEfzE)=r+|C~RJoP+6y|*lr^!as<sov-Od_8E(`a_ZR
zFBtjfE*oCHq|UrSxBF@8d0euBOMnnRKXX`+?d+aCd+-{XnwnMB%do9s(b1mczrYUs
zmxrQ44~nD=Ru$%fJ)zmLwE3^kx4Disr}zB*owELwR=?D1eva$oFJ^_`><>SUxSWe~
zkqWH*?I6Lqzcf`?xZC7w60TZ}P&p|SzwXF0Sx7v&@6ea0oQHLWPKX;{zHmERSK=WK
zzfwBb-*3v4zGPqzpqS_(9e+wz<vjXxw^iDIZE5<u+y@8_6$h7pTI>cC$n-ckH)O4J
zs`X!7a~5%9hwSOor-g(XhMXb{$i54H^<x!LFNJU3BNiHFHOOuHbmUq2uaB9Sc#R~7
z9G24VGO9W-sX)-uqGmqmun>!(%%-gQ`SURT02D3~t#tgwx8d0l1pV0@pEcR19ie3n
zKE_BpWg;PQ%!Fs}j%|iCXC;D?bop6fD>DZvzlLYifwxF_X_T+)ulqeE65HR=7gy_~
zW@%D;Biiq1GY7WyCt8uKNTg8@dPP5OTp}3^A5<q#Pfx)6L#;M8B=3L1nnpO!+`gi%
z-I}Bj{HGzo<a9-WRnwo$f*z7bb!=?x+S-~mmCvxhzdx^O-(EVpQe(292)hyN7r^;3
zd*QwRo;VSsbX-eI^5J<!6_o`KX|2sxhHJU)cYq<_`=>qfCbeFF$h)QMbo>yBkCzw3
zvd3S)y*P1enH_+&qaTG};I=rT6Z=z*wD#dyc=+3sd&x&X0<}W&SpV#@i}XVl#dO?X
ze3I(xwh#Tz&dxQNa(-)t#AdR|<{fuw&zZr|mJD&bF5T%AMZqG1O4|VVWM3?DfIvtI
z0UO|lHf(o0^Q^8H&But_vxdz!$c?dFHWF-md-a8}-||pU>IbXd*f!Y)_@Vp|b5u>V
z*Z@$F<xH7Udn~ydF&*xs*uS;^_5QVsN%!LeC-UvOt#jMG=0{!grNk_M@W8UDZZ_)W
zdGlj!ipjsZA=jiXt=G@#rL0La?fJLOeINTNLlp+x+<`{bk8DOR>Z=RXP;R-6N&iiQ
z)z;|hf#|bV8jg<Bls01yZf=vDAvg<Ge%<#L4P4aFprpC6HdR83Qgn{*s+(<c+t~Wx
z3svnbLx>wIi<2f}a{C}|d;|g_#=W9P&`fO-9o!$0A-OFe3;&@~g%I!2|2Q?l&ui#=
zgxuQHMF<!R)Xw2~ub8t|PQAX15k2Fd8RX`?N*8;)?PJ7j^F$#!^k(0xhHq7e9OhtS
zi+cI;WkduE?T&G`)t9Gl9TK;DSFH*``v!YHrB|=Oj+yCnjM05Nx@JYh+lj7WTzw5v
zlO=5?{_8JPWtV?H;$9eUe-Rto_2zQu@6V5Hy2m5l>t_m;E%o<T1?#-ibsYQ}24+Sl
zSeWQK*82J3zBdX7Kw5H&OgJ;%D}XZF%Cf=K{6*XklefD)n69d*Y<%Y4P+JyD@LkZW
ziZIXh?5nuDefR!%(M7QFNiQVFUfxpwkVYbU6dNi)z$Xxi8<<*U9j&4HS1Bf)*;^qa
zu6$w_Br#kpCgN1^2KBN<$r?z83E}S!G9yTmH1&iY_VqfaBI~RD+~JVqKO-1K#J(Ru
zB>Ga_BM%!OxEVmJfFU(qzwhsX%HqPFA|AsyR$Goh!Im&hw7RDL+Ib?q!_$NhjlQK<
znd%gEcrMCgZ%priV`R+Gq$p9>)UXL)S==<-x0|FvnJuSlMv)wU2*Cj!Rz0^UQrx8X
zi*f}%qiPylEC=@XpO^kJe{`i6!FpM`gM_F_kca}DG5M#GHTivoq<e~nA-_J=hdhR`
zh1xMJ<*<X)gkxuGXoKB5eNl_5$oevb^B*S1FK97eu=NsI7T~De@36p82k+`ec$J+F
zd*pFlH;Z4|zHbB&B)fzGAfg|FVp{|XV`n_)Fz}P42atb-cj7#9OZDkFwBFM1E%gis
zvpQgFac?k$Nf21=h(H50V+@RNqGlp4UyCo#%)K`T`xdqw!jV~Wxddx!0x7OM$9p<C
z4y}jr6oS<9dFp%V0D#4j;^fLH!yRP^EzxerKGNhhk&=4`#E|%pbPU2BOj;BST@x?`
zP=4+`S<kB56NrH}I)W|Gmh`}EQKfSf4x!Fq<lKH7Pb~%v8w1cdUKn!yi~zeS4cd{4
zU(Dq~g1)Sfycy>`oFIP)=Ya7%QUDGkg{;LAvle0D)mQd4t@Uk)6Ur-N6o@?ndyqZF
zY;hj#KsdHkb`99JESVr7pu@N%PV5;T;h(kvFGG^!$QyLXhcL1r%w9DdteB7-D}4li
zH+&Zw<4Io;nv|MmOWExW`~OH8?leZ+Z~U9z(yQ-ocL0rSx*r=^_+7DkM;mZ2-02`>
z#y9By+`3E!^0tUjxG9E4D{%1m;V!*Yo8eB6tdUGKfsRYa2Z$~~4a2bpAKd&9KViJ@
zF^b;2L$WqG)9VwC_RAGBIihyH+lD)u-<D!S$O4+NzFucA3J;dQjD3gJOxCTr?5_)T
zO-m+NsK^j#1<q^BM3P`b%bI{lA*0L2s4Al)>@7~Sk`D#(2fSKR<a2RZN06R@W5kaB
zZFZ)yx`x!`hxrBhxDUuCwn!Ti_e-0~%#-sGRY=4rlKk6~22%c_u!R4J8`J9x)HPS^
zNBr1<uWO~nabH6};523jznM!REi?>AWX2VY$sEA(^kW9^Vs1eB^eWrv0W4hJ3^N+9
zoQWPQ6K^muL^5#}m@WKJh7-2aVKYH9=zE8KaYb#f$_A(}=Q&gx#*p;D*oeJ<4rbQ-
zLk8Jih2+#ZSgqO^Se6N3yF+XyXVD&46k(CWI>xp|T%THEozI-ylXbiWDf9o;&8#C_
z+xD+^Oy-#_DsSJW95V*HRB;j;pHl5Jp5{9(o$1!oM78b>jQR^)wL*C7GXk%~NUYSX
zCe^2Ne@})tlAiu!dX(RAPFF!vyE9mJOp94z@-+^32@4%*o@O>k<jc^Gfx69q;<(zt
zonC=dFepYmb~tr86CF2A3a$|d>gX7EU=dX9K9bs7UZCCkflWX(S&8Z+g*NU8<}l5~
z88DmO$O>q8BJ)k$1$s@Wssk!0WP%2b6Bfh9UqYdbRtkIl5>APm-SVdrGN?w3=QFB+
zzE*5REuJz(If1YT`{J)~a@?e69B?OG(rz?5k2IQGbK0xDQ?d!<hg=8Yt$)%<zK-XR
zrMx<jkMoC21f{=4U?=DA?*bv!ni6z{hmjo+OHZlyX9^o+l&`lp?TTRkk<+Q&`gEYr
z1TY8yM3q>H=@|R6Z}s*2cWmqM($YGi68j7I4vhE6>@wIG)7K9+kDj94*hhCUz#q87
zm!}-VPDO;bzzFW|+IPy=#?&;4bpSD<AB#I7x?0tib7l*vpp8~*y08zio{rQp0M#Y{
z_7Q`si_vk)Xus0~tPN7A!tL+jyi#uCw2toRaSCwSN(sqJKKOo{Hl!xX*X<FlP>q+?
z(`q^AI?At?^VWT`heiC#)vL?nxeY+(7-@HG%#-~C3#HIH)nB{GM*PpuFe>D?x!PM`
z-7<Qo^VgmH#y6LPzeWleT<AIsF8dZlH=J5AOcHU)pUH|*-^Fd(vf_y2J*85W-C*Yg
z+EoxOXmFHM%{Y4y!~d0&E1#+J+gP?>QkO`FTus<HmUTEEiE6lJlG=8k&U{;i+wD1K
z5%&3RRDGuO!L#15tT)KD_n`ypV6qP2KIf7~tc{g)-dJ0<=rpeiVi7T~WekIJki;JP
zS5Y!-GFIdbn7j}&jmjXk?Q3sBWjo&Ki7*|6C0|#SK@{<40f}_ZzNh5K#fMNN6Srv{
zA4yHvkL}|*jiA!{ylBY(M5y+qpk>CuGN-4fXNMceI^_6<;niWe&GvVCxvT>Emf}F=
z21;h@6_u6i>gu$ZhsrnC?SbxG;DR_Ke}XisE72C~ssuU)14=sBSa>JkOcu9qKRGCJ
zi7!veV=6J<ukMv7z_$Zo!1A(;OFsZDe?1A*P+n*uqDH{m?OS@s@=XZ9IP2(R{GmVJ
z-~8+khvUD}#Jm{-iFvCRAVGF>Z8pJg#i}7*1_qOYzafGThXdB#C<p0*zOKYJ*S6ao
ztHycWNW6L-+a*pY@0VH-T}Y&vT|02NY<>rx+@)^u%Hab{v-9G5c7MqzXXrA=tM<Jy
zfJ%dLaXv7X3@w5*S7^g<lMVevAP}s<3f1KVRH3-e4C4XM1Qy~Wn*fCiR6&oa+yCey
zAt*Rl^5MtEkell2O!n=&tA9JZR2f{A^)gYMyt_Znr<**_%sMI)6=t~^+-+BYDjG>A
zp`3wgF0RrK;Sjv2hvcn(@0kKhB;`KQRqhXzEX}9p=I4ijW}y_o^<(pjp-Cr?(Zu%}
z-hXp4sQr#xks&IUsDKK!JrF7~sZDx@hP5>Gs3PMxzqstf`MJ-s`p3XDls!r<PXC*6
zXiC*vRX~Clzwz*?k@!B1JxMLMDf%o5zsnsQfcLTXZND23QvD3aKp+b5SmM2&^5n2o
zeqv(cDGrX)P`V3XPbQ6~CVXxDjDQ-?wg-p$*aY;?n%BO3{{&h(7}dx5qON{%V`2(!
z5t&m#F0!&$v;rw1c+)e{TwZ@)1L_8oti8SWbMgpER%Y#GRrz)iL}HSzzx!qN7b&0G
z7;(4{@GCGP;#UCq#@<wlLoFbrM7i0piy=a+vHBC<FGh0s2A3RtFK@%~G^^Ik&dw?f
z=6BlBRSUzQE7ze&=xo8;F-(mU%r^9Yf#N35KVAXg6-8>QU{gtf8so0@ap}w~LB#T#
zx6Hmes56&7+J=+=cMd0a`zs^U@?4YZ=hFyvXKA)%J7h_Y=`>moKcZ)gQJ##(;UtbV
z<K*nR3bix!5m~vOqSQEMiq>=U=B8V;S%V;xdL3(9-%V@N=jho}F=~qA&~pJHrZv~1
z;q<L9K2sle`>xDcK}<NNFc&JjVoCG{>rC4dX&YnP3s7>gOZuz7F=oPvPt>-T5KG`W
z<!dv+#5-Muj-ZD{*efBJ!}Qf>@Yz63U`I|tHKppX&`*$-ai?JW{Fi<|5-MMF(!B%$
z>xTl!<5*Rk+yc25SI^Uo+x!z_A4F`i_T5+>wlAjUrzpp0J8dA;`j+>)RPSHu1_4x(
z#o@1~(fGAx8R2IjvW9TW$#TxZLYtGE5Q0l*%|NNt!TgA?zA*?WYHB=q@m?S2mw(`?
zX)+7;<)C-jqx`}+aslNkRdr@2NxwCZtuGwImP=-3QfF>Z#vW#Lg+`W_Hyso){LHr#
z2@kAbIRPCvWPRSFxQ8q%OsM%&5fb4*w1GIWY3wy}th|>y5{?CD9sWFFSSi7X<hF4L
zd4FHm1%Ye`LVdXnva6sdNiTLebNkLU0=w-@Q-5<UKSB8B1#IVO1N#xFxpC$7t|gl7
zyTbV=J1B~L$>bd<+y1?(bvPI$Ty~8#8rdU9j_^an60Uh0Ncz~jnrSo}p!3~~Mpl}l
z_rKi@G4|M?{I4T_!G5q`0B{vw8YJa{oe~MFh$SVA5d0A6Ld$=GwJRjW_RpXDQ;X8j
z4VRYH{cE5zQUZa;HjTJM8edm<bLlCmbpd8@L|v=;(^+RER^mQq7zg<G-QArR2-CNu
zDZ8!Ag0PpoZ4P<qdyabnf}g^&DzCw2)KNW3{%m^BiK+JX_KpiqPC5B%YHEys0EOT_
z-9;aUryB>5P0Ia>L?JHy=GgT=h=Y)D<J!0vkm%PJAH9l(v3NAmypuG9kGX9wH~1rL
z#B(7cD^3-)QlMf1&p9<y_+C=bdQ39FuLFo31xz7A74Y|FcYb^R&CIW0oL`g?vJ7~S
zz*NA=o6u#jVE{N68=M}U0bx7zcOay1J=|&Xcf^f?9E;H$LEsx6sCcW81k9iu%kp#S
z#&o6TkaqSzI$F5i?LArgv6~5p>#IelRr&D7wzWK~Ca`zbt_^I$CD{A4*yij)jFQJ~
z$LyXp-^K1W$T9CjA+)Ip04^g>D5DQQ>^#s%{xC{nI(q4WFEiLDiO)pJUrQBPT%x(-
zUT7B*=0Y782<j09uqmE}Z28jsdlK>l1N6{}8{;+Qh{!PB$NqZdzh(>66od6b;7l>3
z$NmuOe+kYrn`|i!?lei81Q%^i8KwLv-`rTm|I9rhKnAPJkH%siDlG=+YL(*TK;u31
z3fbd7L2YAONO|Bm+D=o()E}_NL^fLJL{UAAEq51YDsOZBa}jFC`qp@o(Ch=SQV>3B
z`A5TlB}QB}hKTXuymAfEvc*scP(*u@kB#szNlhvNaS@&`*xv_vb)@7xURtO3Q0!^`
zYc*nYgoR5RN!h#Z<4QS>%rf8swQd}gZEAb&Wv!O);$v-OlfngBFq-GCW5+w8!c9CG
z<lV=Jy#L38BS_0$$ts@sUR(R_`<G#*B`~?jXomH1T`wRruX?r$@PPUNR5jGrjyI)-
zRuAl{Ig7pqmbvKqq0+krb>HCRcL0h+Re`pjtIv*H<btRSA^Me)4;c(jGp2u9q6u_s
ztTwNPAUcsBq5}v+kFQu;j{;KrbQ*96@hYeo3uJs!Yx>wBX(iE^`S?BX@PB@$4v|QJ
zY;_1|T;V!63)3*)WwuCPxyM9nO3^Nc>jV9mHxLhKASD`vFFYiPe^THC<ZH9-x^J&K
zmO^4I+)ckAZuu5q3XYKUY+Fg4x#E>8jA2hX#2=#Q$Y}7rwSEzW*nnJE4g-gn62TZm
z>U;O@2|tCLVj75@R0hOP%tCpb0z25{K&-40EWQJU3xMk#Ma~ww=3wR7eNa<DjXb!-
zy%%g>Z!UD&Be3N*{%fcv%EV{7O2%u>5^@3L9~K44frSIdm1rG9VPg-WbCeeiJk+<F
zo?R_xd~sDL%dp64==-(&6X4M}CP)6w9MFPLGX+i}6%v{@iVAA3t(hDkq6%qfNcCb{
zEF1u+7DU<4XW(NZ(^gNNO-;NF!qSN(2<I?b+PUj~@glwp5H}P{jo=ftec;a5Nnb`>
zMrH134Gj)jim1k&)0)(yPXTxsl(aS^<lpH1tmDAw32;4}mgZ|WzY598ZMeoks9ii_
z;Lg|OH~Qex=*`Nm@oQU2?_>hxXp&aO0-T@wq;WxKH?SW!Y6qM>HQsw7Th}8M-2N9U
z^fJ`L&dcj|x%c|&7`bX4<9K;==HQ<=_a4vbDz@ty8aX*RbMZSLGRGUn#mCD6WN>P`
zE9W%$HHd|;2jY_d+5k7XWac}F+=n84Ax9i_;$QugV;>6p$-}v_K$SABq%BRQ4ILaD
z1ahW800V*X!UXAQHXw<uHYIGl6*%$lzj?sBv@*0orxy`%2F`~FTdB7@m}ys$^Y!c3
zFZKen5$62gc8-7<W(BL}w6LD*73^DdOjTw+oCp%S9OOulM+JRvN=DVpPZUq8yc9li
z;>7UX-CTb=C^j#SD_`h33WG(0PT~uDe}ktA`vV{pK$Jx~*MLgMwFey+HotJ#FZsyE
z)QS0TteuGWNa`~C9j7vUD^w=`#>Z_jb&)S#h*<r6%jgaV{UrGU)rSG%C*wshw-lC=
zhG>!0{!A#|HQC8H^vQqwURg;{PzO>12xAHgajj$%Ia%4B-wrOL&9Dsh)x}A<fe@gQ
zXT0YAwq_bg#nnID$FVTlVgo$Vm|SrC{v&QbUG;D_6CZU#PHrw`y=-G;46mW9t9zAh
z9ZcZMlY>p5G(ULopt7<O<`8mF#2ogKrGVaU=11sD;^6SG>Pz9@AI7q5THopU`1nAH
z@!|f%AFNwQxzyzM@26UHOsGJt*H`8b95^6UI`<PE=mD-`(dqa5Q(!@tlOj0H;r-Q6
z*+Zyh<H=<n)(0zl)<ByUz%Tat`!~c}l+r7~Gr2wb?iTcT*^K;5J~FW%U;_~aB6F~a
z3I`~Akc<AjQjQU`@`reVJQHTUxv^Ld4hj^wbnXir&>=ZLu;I3izi!33IXPwP3!Q-E
z!F$;R4D#u{o_Nmw>Gt1TFDfcJW^)Gn)8jcDr(@zfpIyyG=Q#6Tmtx%lBpm7j#yOWd
zZ|UoI98i^>Ci|~vbS;1`e@Mci54P-D>m7ivHIV|7z!9N@3z5RW>sCU;BH+YGzh~>K
zi{(%W<@R5jM)P~W$q&d)wU093SMR?Y;>gK%V3Cy%KtIz%Wa7<1SPbOkb8h2Htp@g)
zM#XN8NecQW3aSTqb>hWsJMv(mE9)B|G{~*bH31bIS)DBRpRE%Q{tCo$WqxdUc9}3p
zE$G$Oa)p8!3Z+RT_+z2ecMPgY39nwgiixqwZNC)CwGM><n9}B98LzxLb3Q>qzqRS=
zrlzKg(bD{&fJ6ax=9Kf3mX+NI1qTHx`=>5uai&HHW}!*gnB3g3ha*icfZFURBo|it
zM__98Qj@i>M90!o<4M<3p|%RA&F9{_00oVA$%{~*k(v2=6oQmDXeEH3Mx2T+Jz;n3
zuYCM|lfWbh`=X_lxni!>1$#oEfgsr&3kO=<J(1hpk$yeZuKPpa%o1eC;)xGqXfZEe
zre%)wmTyAzf{OMT5s^h|r~Aj9Y(!!wwe^fA97Pnr-p?cjI@<wtfQuaB6GKrZJrtX^
z6A3;}4Kj<G?zr;a&03KL;+1RHzOOCMD$1TlOlGoZI36PA(4RlD5C))<!SDXD<dh1R
zU@)7|#(bNJ!oM;y>nmgAHtqd~&i;k;F!?L{P1~5Vq9TWIUqKW3xYA39wI#>ks+o3+
zgb2gbarlb=VAy554<jjXqU(@r%LhWF@|sxl?Rz&q?&c<5fmN{y8vaUA=CED0i1++6
zUCr%3@KpTM(tUP?w1f(q$#)<}q|DUi?RzC|&(%Kxf%gdbLgFRmxO4M&b7Mqpw27hi
zVlk+Y|D67vCon5Ypp5r|C7pfgeB{E-84mjr_2)gJNK_bW$vA-ri={ISfYt+yVbTee
zs7?3`$IE!572|WDFb=N2L)^2_a(J}!PJWs1YNWA0Y^^j{9Tcwl*Zn?jtgrTZ4j*Rn
z9M7))Ze&&hkrzD+Ra;>XRSlNg$~QzAX^CF}W?uVQOnj-7f=W<8K<FnMOlbi`aM%@|
zC&;c%eZF>P?St!Bp6tWt1-BvXGea|7@0b@D0J};km&~a@pP$?3M#bvAi||1;XG0$r
zCfHlpRMl5!c?%ieidM8_8n{CMWU&`0TGFXXDny3`=grDQZd=`0os25Bf~E*=zopl1
z5AT^z_gCp>_YQ>08Bf1nnHy=!w>99MzX&+j9q$BXZ4=iPtsyaO=Hx;BKc|7>Zh|Mb
zX6Q5wOw%7aYYCBbsQ#8?akqWh@*fBfGF+i8)`vyRm59YC!)HOYNFmCLOz#i<<4Os#
zMl!{a#~&>%Q9CFZg+FhfN{9!aJ;7A8HdE8E!~?}O??2HNV9!}RyMD-?n*8zle$&W1
z51R`Y?nSyn7&MqjL2D^r?Vju(7*M!9_qS2U*mDR1jZUWi(6mQ=VYXZ0D=}Npzj8(F
z5=-ab+jfdENdm9|+Kb0qtcq2lB4$WJ#f=N=`+Xpnp!)sC`@1bHETE|WN63F!5jrRI
zDb~mEyZ3Yc0t{qq9G^dwK`!(9;;#gt#`>gP)NIoxRB!A%x_plUw5GqhdZWk*SZwht
zePy5Px^z?e@6Wvf6~MNuFxxbM{C)6dGa!Yzxw%JNUX~D!ImGV#dVlji@fUT$CacK&
z?~m7j3@qR##<wpW3`ni}E~HGROhm;}{qu#WQLFVAp3r`lI-Q=LJ~A=_>L^=SNl6K~
zZ9SmPKef@2dm4XwgU0}FGw`eeCNEqfUr}@;1UeasJD!ggYS@Fq8gO1KURczatYh=C
zpp#0Gdst=46c(iHrxh(-l2yL30^>q{wSZb@N9S&?&|$P<Pzxab7WypDJQoOs7J-P}
z^gFhJv^<y=Pd(uz!3jdukx+Z#*M8f`Mf0r&fbyomQ2~j2fgslra}44$qEv8hlVz)R
z)C7>1iKeKba&htiQ;WGXL+O2B7ED=l^DXxo+1OCQ8=T29;qxOM$SC`{qznBwH*#&C
zMMv}EJ08zb{@l4-YgfQ<@XYP6&yL_{npEX7b+hpT0#qoOBm4w$K0Yi=z1+{2NfNX`
zU<|#b-is|6Itk7uY=#bZlTMOI+R`=hFCrqeG&Pw&souJBv!(?Kw6JP;Fn!y0fV<nX
z?F=`8W(wS2N3C*19t(c^PKu2Vk)ab(3YuSUdzK{xX*i6hheS`g3(S!$n^Nj#hH5~c
z8*wX!#*`uhDBlB29OCcwU<f|Nb(&fi3=m>)@V2co$iDEu)r##5O3`P`YG0lPy4~xy
zO6~i2Ux6TH;|={uBvSmevuk;;5I!jBrPA?7pB*6=AC`2w-TlGUB36yBB2-jV%N#JP
z=_Aig<LVrtJm8@;1RkoXeX?h1PkXM#qowy9J9_{SnM@xYlINL%Aa48X{gJ|^63XTJ
znwqApVXgY&$ue)aOvIF~+rTBLC2b1f@lFe=1tPw|G}E=wMMdkz4=y8kJmifSIe)(@
zYHn^W#7c{7-FW{EPg`X;XToj3Bfb8czExBc&!Gs?riA)^RF%8%e|Fq5-h?=jXk=##
z!KGudMU+GW!N^#OPKMckV>V%|HIokyS{F13T~_Qp>)21AHjO}Z&h9GBYl!A^hcIN5
zptDJ}_vd3X0q7)k^V1F{tG2A}gD;#<lDsR=xli!p0i|eFH8BEGND&Bc{>lJp=sO>o
zIShdT(9Dy?7odLt<_WQfMWG0cF)u=>Wa=JhxM6Vc6P*R*xdnu{0q(H;47FBjkqRsu
zq$Ys6a=keI<H24=QsqIRTR=YPuQ?V&3p0y2_&q-C+imFWXoF@5KkyV;-<35GJM7wX
zhM*NHRXHaA_P2jtd}F5Falvy9ka{`r4QMwQgh|4-o&hQ@X!L=ZnYlJjN?brdnE(Jl
z@1M&<F0+Ql#ytSg!RDf3V|@?`n+_9TtpVrQ#lRIC5drp0-p%D9zpCUnKiV=`zFFoL
zmk&|!*tpTH78z53>BjGG-0h^6^#}3g@qH7u{Lwe{`!hs}W$5Wr7<|ZK2NCCfes<DH
za`^`GRA^eUFZ}&bYNGHBbaFu(+a?qT&~u+Wp=UY>4QOyi6o?P@$j1P9Omr3HLhZx7
z+f@e+1i&P#kntKoi4N)lG)dfmF!Wny8&j5-e=mddR6xG)jS8GVMbx`=td`bzl+mGn
z^9lHU*f{v>1e8*sx#xQXA8<pxm6WS5j>|w!JtDgS=<;SkXM)enH&_ugV+1GLe}HDd
z6x9Sf!THKGU*k*?sdaIo!-lNol$q7s+zf{+H{X2SuD~0T5~#1dMZA)do{%6BZYF+C
zP^X0a3ZU5zXd)VRt7>4nWcmxo5Elt#MN~r=Ap@)H@dblMkOw4bxs|_ANm)r7gx04y
zco`}65Jr^*bp0ui&e_Qc@#*LD`yUO!x2`EF22*eiymOEtuKaJyI9?YrjL4ZY@vsU-
zq0+Om8_i9c&^8D*;I_HG_{fY6Sw7@?l<|PbKt_PP(e2WtjM6EPmG4GheW7M|;Didm
z>f~-adx)slu3Z}nU4vFg`_6o7*kvwo1|#4e@CS0h=F}rO`06oHQBnTSI-1=_#ehyh
zroUN9WQsEJ^Mu=k7cP9NVl$Bj0TjA;G66=Z#GM<;pX@A{1dfCtbg)pF1TQb|>C@TH
z!}SpGx&Y8am1ok4Ly9wGCmpNL%19~zdDL%IXKvGxHzKpxl?w%l-t`H0(dmK95<^d@
z_bgt#c=1L6sH4r~8)jBkZ5#h%o7?E&@2UEid7q$ut)IcLERRxSBXi*og@qE2J~Q=;
zUNF@I;qlvT4K2!Gxg3d4QnH=uR^4Z&^6At6eT$Ja!MFcx$7yL0jJ3*~^aNVD%iVoS
z2~e9gy#HufzfJQ%P8s0*J7O`!=E~Dgn4r-sC<&+<+*>&UeQcRs#f62s)wewAW36}~
z*MI-^t!U~KU8Z&hT!-9#nLkN$lG%$o<C}c&^fQpe7^4>_UHLaP3>FOCFXpAV;~G<;
zofBllChxsRe-jnL0e0q19?_hs+gAr6%%UMa?)mdGrj<m-MTi%3S|8lXmXZ`pbzD#*
z1<v#WsKSsVqfYHu>gwtOR9gb0o!{Y6&?y7bKP2^d3CAQ*6rzmW+HDKE&9n3`*8{Qy
z9s<piucD%K%Kg^CQ0Nbz%L7o_5(nylC9@7(ykbClZ*!!uiEG~-5G9V`_4M>wN^iNj
zEkGk{-q^m@^y?O2AE-W7;yDWh5f~|7`6Ngbi@@O&S3l?j_<~4)+8d!w3>2erU@!Ee
z!(h?*JGa&EL^5GwsP*n8u%_QBF?Od4a)1;>bX83*4i61I{@O0+r280r(8Ao@ZM1m=
zbb7cqDhUw-z*zyHFxOE{>gUZ|xM1o}rL?X2ZHl%Tia->Eq*&fte&gdk2Dvo{pwZBO
z)_5$+P^0@MpasVeaujSZF2I;{xX)MC#LdmED1xOewh_`r28I0O$rI=iKK}2&PdLO?
zR!d4DdO&C6+C>JLJOC{q#cIxc`3MoF)EG%t2_F4>s6O7g?Xei`i;0gfbNGD$)zVu`
zK~-~Y+Qr1A;K;z<uWE(v@!%>WVYT3@kgm!GL!luoi5PC|{HI%&KTgV>OTys>Wr5-~
zZ_QK$f%aO0Kra*zKqRn(><;KQLb?y=17Q#xn{Ux}nM=m&ZEC7b!8TmLsWX$kHYOW^
zSVITyLz-KFee1h+IHhYFX*-vG+I&J75&si9X`RNP<W~ZFM{UiWUiJZ>WD)X6G%9^I
zR=5phvXm~);d_$Uxsaijgeqz8YaK6L4&h`93l9%>8l#AUij_53W+BCU1E2%`Cq$Yc
z@i=KuCy3E3ZO};h#WX!Rzuf7X3io?me(1zJbH6dV&Rjt~Puz}i2mEIzR5jrySxg;g
zMPIn|`YWa5AvfXLnI%PE`S^8*(8NN+{2~#S0t3RfI{l^9RDD-3y=H}PO^H>edsf)L
zNKgx>QGe*>wq@U40S~zP5Y9nU3=9m!Yh>F()#N-OC@6@u8UUS-@gf!>M0x;y{h8hi
z<DNhePi1WatsH|&N4IaFj-g>}*>?803IF?-Qcv!!7G`!UhLY)+OklPn{l&MXF1tQ&
zwbP1UCWL!^3t8oM`m%dfZI^p}O>d%8Nyvk*rr8ekG*F#n;FM6yq#ESk(VKq?)mw#1
zsGt~-ll4NYug^c4@sI}hhImf5W82Y7PYNONWcZnY=SJFZ#8$+2<MD3Au7i^iy2_Kl
z%I`rPKSMi{MDeB6Z&y`Ps;#MUE;5AP)r3-=^`}$70fCT;N=bG8c6f3~lmx{Si!T0S
z$IJk~vb62qMpMvfG@tz$W@ecx%gV+!Te{u{9kkHwo>}0n_Nv=X=fBD_+*0eEgwib(
zcCSD>f+kgB!|(w6uFFynxk698B%5ZgtLUx9fiRqQadN6T^TfGQp3$%7xmFedyFKZT
zB&`bp0zy9yO0pd0-notDfA6~m|DAk)45YlFAr@xla07B?x(>NVlKVHft$q3W0tlXj
z!O;vd7r@WY4=qanC__rCe#)Wa=&PclQgzV!GMo##u*W9k{qZq&ThD8K-zA!ux)k5l
z5GX&#)cu+Fy1&%QK&XWIl&`=ILYv-<3$9kx@6*!KKG^Sw5n24=o2_QBFjv;tnHN|&
zI4cjA(icL7iMME#;Z@Sfl8_zMx6fG7^h_>}dC>2fl7B!eT9KNXni=bIAHoy%?*b%H
zQy`#%oKP1pBPAlzbo<Qu0QY7gU_H`&9|Y7EuGFrTf$GrUE1<^lm9GN!Splg7fLWep
zV?Trc(nZJxBnotvm`3u7R~Nm1>$Y*mJ}LuMpl}a3`+&>0eivWr?A;LaTU$cN6|_-J
zkWnVeRwqbu9?-Xn1S1gPjGXI4liQ%ItVgJaV03N5KvQ!LXg)k@P0rh$i%qLk)?rOD
zH;0cF?g*EK8@HW}>T1(4lsDkF&93@s(^~hl*2K)qw$09}VD6f^RiKUj@w~!Y`hT)$
zCpaeV2@4C0l*@&v&=M;SvPU+ox|)!sX5%JfCq=56<2}oUB>~5C%6OesmZHI(<f$^9
zHvDkKR?0p@U8Ho;cOnE<LnJb<cB|Ez&jP-EW9%mDHm*w%GNPf_c4`wqfr(%5?}WL6
z$Q`4U0vL60w&&NcU)NJrDA0JFu(@`k9olDhn&o<CJ!v&6`xn1(EkfY_WE98ZS?KAR
z(E6u9M2p!B7!}?L;Hk8Io*y4<ifwaRBZyaksgBN*av^Smo^l&9#Y9(oXJ@CRGkg{C
z2;lk7w;_7s&p_MqszBj5bW5v+c&yF5*jQp>a(t;2`H?|Fx+s(v2d(;hM6-)QH)`Hp
zQ0gG_h|SpJCa4-ft^pd!fdmUo`esH8YQ^V_YHCQ3>$RqYa7yO*t*_usbK6zorFrpi
zt)PDPJ=AVdy(f6)Dh_wDT*?2A=my1zKuA@bl&j=&si`O9)II?_?>hX+b!YiO);_0W
z>UY_~0F?$23kuwU1MZ#w^6c3&NaXVq-JMNMB1ewAhAIctP|)>|nN1t$-tn{^$KjZR
zHm)Qs7qHJeUTQPen?LkH97;9ME?(8_0}$N<%V)Cs_5Nh}-W@x3!2TTHlx%ElY){b6
z(6)u<Xa!w5TtLuO_mOS3P%b$+ly`heF_Qn=uolomxQ_z*#th^><}k=KmWH4Z7si@q
zdk7l&lu~^x1v(&ktDUn_?mK$)D4m4J$&-k{9*P2qMlkXXbuIB`9;&3coO*22=j-DG
z`hJ<*#?nz9p2QuI@SUBPJE8paeQzeq5_Br`xT&bAJyX%S&^-g<)W=3tWI98z*~ZPu
zIsRM%-r6I=nGE7B&TLPhBlI2MyOZg-dYjPKaw<jl_yF9$x08+wD(-=FZ06jPpjbkC
z52&dY5chHzp)o-vTUR>j2R81Por`znpuPmGn1M_BBwqQ|*>4+g7(W5ZASD?>2#oSD
zFwn{b)@*4CR=K9)9+(ecDEzcOA8IM9AMft%EeA~0ZgTjn<+E3L_RuUb)%Ve@*l55B
zrh7Crq3v%dHw)9Fb|q|`jG24ZEE0TE%qKxM2Y%VLzSx_v4qCpi=;_nbPsMu}!=U`-
zv(RpV?V$e4{OW`58%Hi+KjwlT7xen^;Sq}f`GfwUwAkR-{0cA<$ZfiWvds<er!_RV
z0K5pN@>z1Q7Pp1VSe#bDU2k+TATUYvZg07NIT7IOpq%ZG#Pjvvpj<oH-RxGTM`x3f
zk2x@NNb8ObD7)Y_(lwITQkPd&g1&-wb5;<lCP6$Foy*2>oB=cA0NTk(Cn4wwee$Nd
zA&Dq>q341gutPE#F`QuU*QRE<?GuFqA)wPl)H>-OSbFx1k*U4R*L$O>dC$j>AF&k%
zn&(DG`M=1duC2qsYD~$=-zdzl$qUq4cq$M`*rrnxfq&}byr~_ck3Oq+(XsHA8p|PZ
z@_Oo*>guQpG%^?_@rtHqGh9%x6CDPj5&Fy6!XUt4=E|(y8f4j>eKaXQj(aBHb5m0{
z6OH_y50OcmRuNWz=$m#oDT!FH_r_T3e$c&YYU9vTpCBg(+y>fSpf!r6_Y_>n*j#BV
z2W?HwM=|^@$n|IKx;#K%IrcuXb%!6+M7&Bj#@llNW-mj%#aZ<i;9#V3{;564zz_=I
z<2*8lM722s@%DVXe~KF2WS@Z@onq`Bd6Ci}*J|I1Z_{XIs<LjOcG|A>a(0^(v=hcJ
z0nUb&z^qr=ClQMT;T&L~p|@@-3MpVF^n3Pr0@Vw)+C$~sf7#mF`r|NP5MuvuYXM*+
ztBK~R%sohmgxb+5Jd~^^LB0Xej2i$Dz}ZsZsG)Kj9snvqidzE3`0>}>+&(`_Je}Dd
zKr7PTFAirLr)pG32cFpZT7m*@dR0+`f8&A%d1x9g$1mLsbXvRexC&`y`#2@u!*0dL
z$KSkp6Ov$!V=#2ydGQ844-cQ~4;Io+*QmOErUg0&&zzZo8yTlQ<&6XHrc+SgMbG~1
zmNeWVxgqS~<#qhTiHxJLWKa~KFyL+7mX#S2EA5%1k$mmao70!AR}i024+r$kydaI*
zBcUc&Q^O}Ltfn6RsZ6E-jvcCIuYr7)m6bv4g&M0~&h4uUw#%qq%RB_CK>hTW%UqrW
z^>ENSGSH<tJ_U8)Tr4asTABKy+}sJLv`BN4y(V!EAONE>BABnNtgL{3p6>e_=R@Ae
zOjT^Z^u`g`M2DRE21H8y87QkjtoZRmjj0`YJ}`K6K7awnU^>0|wlTz|)m48e6F^vR
zEn|g5z|&n--dS(4{wA5yVual8TDYv;o43i`GiJrzQ#MN92s2&I4f9Z9f#4h^WCG&g
z0;m}9%mZ9p8$f>sUrNF;GsrISz(2|(A+ZF`0s*@S;(|r(OZ7Y}($tj0n@08n2i}PB
zxCe!VK#L);>c!dF>z9$9Bw|rd&mmSTdE)F@7brx6nW19r`lZzXn6(Y)zTlnk9scq3
zrK!&h4ZUi~*CJqhA3>rPweMN$cJGz(n7RRZmFl(%g)j*G&Z%QN=g|8LzivTXJbcdO
z$A}AvN!pt)K~~7MY{Z+G^j?cNMFsT<@;6F8z3lAlP%~bH!%7~SO^D?y<bl)&ni&`L
zmO@wr)c`erHgkMx1G5KF2tvU%B_;m5To0kO#y;|FLvmYZ{n%)8?!q@c51N1+#Ud|-
z#o_$UP5WW)0?GQv10M>X7RZD+%@jsjB9;gmq^fEI)CW)V@>;=7Og4~~Kzkwl>;AnI
zayn?I$B!QeVzmH@5vVcQom3DJY8ypM0-7{8Ha39e)O`KA1PnVjRLHnQm++8*dxBH1
z*e#O?MPc66JSZlmzz(4{8Bos&m39S4GZKrT!d0L>w6s5*%=6&l(est37a_+$UEsy@
z=g0?`K&lO@GUfe51LZ#Eg9kl9%LG&dn(SNXQE@Bwg*da1LrfY9&Co0o!X<MK_MRzP
z+}=23(e3Y#BQi3jg+k-vg`@JL$0kO#Cg3!+Gg@kDYP5Dk%nrPf>zVmPyR%;E)G7zp
z!DHGPzwS)zr^6Xuyn61|(#;~ikU`-}hyhpg-45HnrUaFI0fiCrG#A7OC`FNu?z(>y
z;v1|>#&hO2To?7Kv5hD%_g7MSqW{n0^t30G&IAn$g@IOy_jnv#ja=QP;v7UI>K40E
zLEC|nH3@mJ6Zjuk@I)Iq!CC*#op%6k;GNuZ9rbV^qhj_wK0n`Hy}z5m81zdx2>96{
zCMNYk=l)Y5p|OPRIdEJ;oaqVsX;UydAed4xm;1W{%N+P6Bs`#h+4}g{z<>zIXHalv
zr+)}(8?-Jh!jnq_eB_cG5Urrj&N+!L$|?*FC-_f574=I)1Vfu<pR+JK{qYy+5+M-q
z>}m#vx#z0^dwsrB&$;lvpGut|;u8B0iCSwy@Pz=M%K-MFT{_d{H)~6U+huz!1WB*Z
zy+=^sZzt$gSg8Oq(-41Pkg4rD*lAzxD!~~R7G|EScen`HJjiHJrJXOb6YT(tZZq~4
zkd$0*yv*%8-=dSDnLY!i#n#Wv90T?j{1tFaj`-{<G)?+Jr3`<268|&(`U!4s?&0h6
z=cb`5-Ii(4CaIfiuFUl9`}gnD#`D1p^HPER137!wpmL4cm%)ml#c3Es3ZS#U9Y6sD
zq#z9*3Y+NMaslWKcZ#jS6RAyR_0IasfA8JvKFowGe0lUeDRY$?9J?aGQRx&z_uTey
z@%N|n3)<r?Hq7`!1Gs4lQf@bhJ&@)BI6;(o0gLq32b$va<UwGdR=UOzTyINNP0*hG
zc2Cur1SK!XKu|H*tPv_t<Dh9?pJkAld*)mZ=}6&UARP@$CzM`JL2&+3SxNi~U?@5A
zP4@&SiJXvfVSnb}BLHQK&<ZmGLL#V~K|b>S{+r{fuIJ*2J9h6L=BibeaUHedm(9A`
zhoUcZ<A^Sr?b`q77OE4$RN#`cfqm)Fa|(3AWwM7Hf7NDEo?+uv(4*%%mV0!+{PF;8
zn(5e&<%L-d@<_|7WzRjCw+}N7grWjnZ6FlJp1ySn_~J@4z&%hdK^85B0}Y98A!*gD
zFZ}rN1M1aWa(-RW(QFR-(2vZ7ddffr0Bi|j7(^d!pmTsnBu=D74`&{#Ir#KTHJA4t
z=o^5X+3wz5+y<3(IMGEo?<|7?!BgnZH3ICM=&dhLj7tp{P8d|!EdxKVu1+mH;K6q1
zwa%=wsJA`9I1dkx1sz{s-<Gp}fId6gpwLv1RP6Eb*3f+ZY;>KzdA{`+*l|VN9qOdx
zOBuW6oTOjH^MZGUG5vp_+JHY2a1*+fJ=~B`?lSTd%1zG9-dgql%1Mcys5Ra#*q1l{
zkp5uO$?wob1UDg};r1vVlFy9=)0_{B8`|GS!QntW7sRs$4<$z#e;7&?a0CDn^7N^U
z`-Gv3%N%qNkG7FF!S$fSX%+rczxfnuT!IXj$Zf+t`jh&A;h^}h1Q#p_;B9kYCzaLJ
zWZmpf%Oqnm+{1tGRr3*H71&P<HhNx%{P{MRD3GV^MfUDK%!n(D3AJzi@jFD?{VQKk
zriGk$_sTKGDj?I`OpilC;$?lkz-Z&Vn`dk10Ix^f8U&O$y07!|r%!vO1d8DVq16E{
z=~_Xx-g;Yt?`BC?+Nj9~7v6>fC{Hdx#{#5;bbZPwAUjZ+3Jm=Eo|{ti;c(&O<9Nsq
zaOhwtNc)E5C4>hnsgspljr*+y<lo6w8p5&;*?8ZcJ&x3y&OcsZ{nGm%y#ZtcMOI|h
z6MOF7y$gvEis|{CR8)KUG6WQYD~A$SphLQI(4~CP`uq+Y%d<Cb-S19Rcf?R!*zFC~
z-!`?6H8*fhzs)qNe*XA3OcehUgvIguj~_n<ulp|J2{jeTiPjB}T!~*nPJh#rQLNVF
zq_+A+H9^)FEEygMl_j$Uot7ON;I~D+g{D<ORl#g<UBw#89o#-scZsMO!PEG&$#Z~p
z`k)hEdV0u=oIK|80Ya)OA@+B=%rac!30p4(B>8py81d4mJGZBCci#m#Sg860e+7^T
z0t(fj4(>N2urWBV03EVF1id>wYo6`qjE~1)5Zz_7&3AXhqBmc~E|1eitA^J2j8NYF
z4wbb2W0WGe9~c@_7%<T;$$vOFIx}>HD91sI9U6ZD`S@3Xy@_b2lA@ybSY|=9)E9^V
z6k1s30VFD)w<}vO{=P-kp!WW?v3di7EW%@B^Kg|m+L+8O>*E0p5K#Gt^4g!+S%}3A
zuS9DC_x+-Cs(=d#&{1Osw@kQX^le9k@X!hl>Q5HYWx}Ua@d5;Og+VY`7xOpp0n(Z$
zUX@$|yv-H{X{c!7y`-gbUnr*}9uJCgy+2aJ75<f#HAb9(=MksvF!gS7l87)aT+plZ
zXkyUOw9($=rROuCPX>R5VY?t%0KtLC3UNCFDk@MNiz3cVbbCWrQ2jTEy`X`BmZlAr
z13mz+P_}yfb#uLU({<dV$?xtqm73Rp#o-imxQBnqkw+owIyue2wc#8Fke@dIJA&Nr
zdfXTIw^&Ra2y^+f|BOUIn>$b$6UZ;G>y>2pdQ?C|z{#M004D=%15yooF9c@?a4I-g
z--)8(W=Vi6wt!Kghy+zVR1JVh9anxe-dUh3vndFdgHGT9UAb)8f5xJBL$~bdM?fJJ
z?{+Q%Yu$E+^INN1nRmjwe~RJ{;M5!-A{Z9C*@B6}`4+-89<Xc3{q7**z%1Qh)yU?2
zh1Q`)Uk(TkY6k7p1vo5NTb4fM+WKCeVpzQhJChZ}jRlSvsq}BdDa!r@6KGj*d%)7t
za-qWyjsw~kL3d8@8s&{@%ce^PM}_H!LXDD+P72D#mo8lbR{|NK0B%#gX-uo1o|+;)
zeG2VdE}8liRCK>pz5h+41R%_L4Nta?9JAn$H@^1#DwjBUsKEow-VbOR)KzalhnYXO
zruyI?SsEyQ83LgIb_PPUkU&jQ6}VjYd~b3T_6K!F0*WZ8RBMHQw7}EDql?6`+m4Ti
zM?SSlk~2hFh<M>VXgMJKrzUs*k<I%G|4iD<?Cb$(*LnWx)#!x7wZq+2bjBv@s<wtv
z?eEOIdTEM=(}t%Ywu{ANR&M8KfP3f7PCJL+e}xy!3PrZ1yFGu7xX{qh;GaiFk9=|V
z$nd@&^)@`5+5M429GPtI$e9oOlAWCm;i9%Y$hg9);0*i&DM;*;gI0XFIpb7rovmvP
z1qV1MeDhD#pa1Z|1EpArY+Ch*S{j8)YYLZ?+dE)eTwDcZGUvnKUqylB&A=`!Z_B9b
zlc(L+GIkOQz9pb=8Pvh``nXp;QSltRpdj%_@27N>U=m0_i2?BMkwys^m<zK*5i@dl
z2Vi<=GCLzCwg8qKD6<5*0Jur2<QG%vMAQkjYu}+lX>Yc$zke-*t2DM1sI<MRKs~3g
zU%$rx!^AZGQ&DKhGWoBAg!Vh#)6o4<TU!g+01zjp9T17X?ilkWx#|egA#JSSR&M({
z?J*Fr1y~g!4wAG>%?iRn(P_0jdr3ioWpoME`k|`!`SWMWuK$a<_m1cCfB%NxG>poM
z7AdO~sWc=+GD21nS@kI+TVxY`N>*k@%g7!fdymo(8D(XYl^wD|uIKso{ax4Zy6*da
z+>iUe+aGm$JKyJcjpKD3&*$?vURq{U{1)h81TVT+1$>-2(r8NZ^A})yR##UEU=EU8
z+ziWn@=;s(^4#3|SkS`-bT4afLC|PQ2YH^C?0CSR?Xm>!<WBh$4s=vh4_wi0n=6q5
zGW(2R3_*<%QAb_b<fgAsC&BM4e3LaXvtdUnkbr`M0?hdE3aqgXj7R_)#O(O!zc*En
zW2mIE8zBV<O>!P3c#=sJcbSKa<UX=c`%zb@y^cN_O&&spnaBp=)y0cm3JHS^Nvd42
zU%%r<5Mu(_J-zms&}9RRm?~YMy|)6lusBb1jJQ2e20t)d1(y30zqP*?AS5mXyNtGw
zzjzT!Ku8yr;23G;TkK_L*GpKgZ+xGADb_ne=4580{{cI#hW7$Ol|fSV@xJdZb_V^9
zTou&r9uhc%zj__8YybLQY0LBP!lBCt?BeACe>gcgg&tP#+<S5Njr<Ovve)C~SsuJq
zp{yyi_CwaD=<;~Bfarb_sY=PdujJE$Lescd_MmQlZ8tDaN=xz&Qo>IEyj#@U+Z(K{
z)W5t2(I$&k!iSqsplA>nn8XJIxmlgOMW>FX_PKAx3wP`#fa^#B{R9K(jow^9rz|Ni
zL$ZLvHuDh9nF26%r#<Z4vE+U@jN*dS@V_#r@s+K2-q!y`ESd>;0Wd-RY%_}Q1c1kp
z4PeL0O?Yw9H~$n*lz1i@RBfQ6C!?%=SVwm9YclNG<$$FC46msDp>t=?7NMCj%C-k#
z9J#{M5FS!ArQmAW=Qg`;CU)b6#8g0^EI~l_!0ES@yRAVb$}bfDgYc5AdMlu<J;1{=
zGWUnTxnU^*71*DzLPi4jBQBbuZRIHejl~O@9?f?wfxK4e>b&tdckI|NtV}Bqa{!2o
z4f_Ey(dE-#``0>!m!*h+A0`o^n5feT)tSR4-#G?Mh3i?0^rRdiCQ(wAPRm*X6cGp-
zcb{`9@PcNILs<A)Vxuci5-7hA#ocW_j0e~+++JrtHZkdrAt_SGe;@L)D7U{@@@ujK
zkc2w`VN<#mcu(5*hO=7Jdx?v?KQX<>0$jl!kR)*~)EfZMfByW5c&qltF|+oZWDHwa
zt)gZV+s_Ds)EL+^#eXWZZ6})8ofU`<JX4_e(7?{Y3lktKh@iCQ;^gdx<P%gvDdybO
z)H$HFx4TQ0V1;xTkC8}&QO08j`qTGv5_mVst?_w<5W-qC2JkHCo3$`c;hbUKH-~HG
znH6HO#I$AI`UrCiKKI}=qsXsp#AROgUHlhOb>6JlpKrjaNlADjBt`ke5aSn-*95{D
ze+|V0jX*#_+!N`zo|V&3cxw>;M2KZuN1e8IS7=*|UojU#fq0LWjb&IF!MAn5PQt;l
zNfpMb)sE;MLOhdWyU62OI6;#?Qemes-0HhEOSQn-ld5tIo6xVnf0wvw?}bhUD(P*2
znaOw2KK)|Y$I@0KgPLe_3B8=ZhAM&MUCCg>$fK`>7`llVx(~>$A^fZP90;5R`wnQx
z1owZN+}DeUgd~xCrC{SdxD05Uc68Z4;od$zGrxcHtD$m(JME8zyxFd00<0$uBwNuS
zKmq49TMzn6ctHUJ9=WlqA)-CEv)jHw#L(F2%fyBD3tt~@owr4L!U(Ls_v&{^u>2kP
zAc_Ns(Sn^l2>9sVoTz?Qp=g~K5kS~DQWaG>FHUAJY_D@J5aC3wNJ<bgNqRRYq_G-9
zCP1R~*ocP;Hazq4M)%^2KTipJd~*iAJ|tmDn8n@Q-6)>~obc(HJ2yc#755iyEAHw>
zGb4?Cr&U?4A;BnNX}tPcHI#|_7y62~%VU;3)gmXtvdbTrQ@NSNMrZG-B6<1B6%pG}
zW_I>6c;X-rfkIzPYk}ue_w+cJ$3<B+wUFWsU+MufAvKQ3YM`uZ9h-wOMQBXG44Ip+
z?g_-xt5{o@h+I$911CiI*VTpkFIFQ?$k|W%W~_WlcNe$s$#h=xJ*c!VTdQlY#<UYy
zMnZFrg$?fHHU2Po$Z#F!7N`5s^d%o|?gkPPnZNnyw*MU=U(vnrXsdB&!8DuzpQCRs
z|4VKu5GDd^Loyvp5(;jZTN2almU2^Xrh2s#pUhwTHo?hh{q6bEQsEZ+DZlHEODV-M
zMb?Azne1x-P(kLW#qbikd!Xov#nagda2g142*f4kr#moiQw(T4D}&(a_}sAhgXCGH
zz91%1PqIJs-w|+GwRoy>A@rrBsM{Ysx0wPROOZ7ahHSKE{P*z-O#W>7hC}xbHCwo?
z{G8rv9L7X_tGmR}CkC3TgK&0SArx?Nb;Z+|52wP7bHeXX0FRE15qeDk)RS0<>ok^s
zkHqp4@K6L79O(LbWz`U<1I3pQ0}00%6`O&FC>2Na%~(tA-g+qJr1ha0sSCoW5$<`>
z#T6im*cwhtb0XA+p)tIe4uo{!YTEOF0A(y`sCAT~bOrz*glk_>%?T(qggU5_wExpr
zq9;I3;`hZrf8cJxmKb~v=J9`V)THXmmzM}h5%XPv8*Mv)r3n`k+~X%KI7^i0@`(YW
z_D*;PdeL2eA2A6B|5gps&|os~yk*Y+`o=!o%S+FaCK1zvxj5iST~sxnv&OXv|1}ZY
zg!lhuYNZi86ne`~<`!^}=C=UB5n3Ms8VJA~JRpb##1s~=kq8!=Xuysz2B9r$9bY13
zh>wf&s)Fr&`QY>m#V)3dx`kiA7=mDT+0VJ0vFApN(+X69Y5R8H`SYN`ivgGLb8*FD
zHk^&H!0Ol6*9QTg9D)G6`oP$jHux$6b6^t|JbSi%DIJsH{|T#hF$JnEj>%()K&CDT
z4j9208yJWIuC*&J@|6U1p8@)5GZhu4Sfg<1d5NL_(gK_~cI*v|Fj0F>L4mJbmT&u3
zX@`1*@4!s0w<3;~b|q4C2JiXKs)L~1U-@dokl%k+&}PDaW$xN-YN$goZ|UYZefjV9
zmPDQ9nIAk+9lEt@J!Td|gV($kX9P`K_`w^&b3=P3bHOEWg)Z*Ii4zd@&p~R}R$18(
zAF^I__6x9>6A(fzE;>B<+*E9D$#|{&Ug_)^_~^`x5lSGSo&lOf@akQK$O*c+2M->=
z&lBLp0u3wzG2JDMYoc<6w3HO;iuE{dj8M`5+|TJ22sct%M&=&Uc=Yu2;K7b{6)(VD
zBr=y^)WEny7k}!+iCL)riJ}?+MmV{+n(Yy!%hi5?rM3|l7iR<`_!bK2L<o{CI?te>
zOCk}z5m;v8IN!P1FXUGuig|o_HVe)J@mpe~Q?+5XD`p@Cg^(+2H@p*q{X&#Bx^ws7
zWmTrdB{)n`;%iqkh=hhcKPX7iC10bIXEaN7*yvuNi3{FE8HWXV2AGQbcX8D36?5z`
zzV+{h+=IU`=`bd>Uz~PZqjmsr4Nsm(G;&*8a2ve{ss*(L?m`kt7}^Mt1!R;!WuE-p
zXWtpCXs}zHS;mbA7`?r`7C~a4w&+Yc@XO>;587N>RqB05$1<30-^Y(d;nV<ho`Fah
z5j$o)VtxyZ_cZm?urPWiHa1bWzb;3Q9%aOuz8h!>T6F8?O3b3>h)puF)1exA*MJ+&
zFE^Wa6{$-hza`pmijxri5n6LB$C$+1_MB9qx^?STX=y2j|8rMqX;l!U9q0f!f>@#*
zAT-5$vYx$r8(gpdcgRRA7ztvs<_=~34kr+~LBV08g?BT9jadl8yr~6&`gmKh7cwP|
zat8NX3?qYaHVTCQb+^(oONArvM1-2bY-hp<n|y65^wm@B%{EE{11Wm<X9H5)h_DH?
z8Jmb5|KY<X;CH|RqL;OEZmR}~U&N}EW%-G2+cuUUbhJ@<cNwn385|lq$d;^AnB_&-
zu@kdXzI&iceSs0){H*kPsne-*=gQFNPNnBQ#Xtw)1YsWnKO&hP_;4<$BskA^Ccig*
zxxSheU2<4wQ&~vQF$R+RW$8=<$0KW3<;o5du19OK(P_Vot8;mcLlNaq&9Vo@jnq`4
zGya}l6SV&|Eg)r;9~c{@BO~(>&e*oi541Gu(_siXf0VeC!a*fcP+_}>L=s>mFyNy|
zAu564hv%`~!Lu^+pUTbBs!nr7BO*6pl)+097PZd`Yr!y>dS7_*-rh^_5%xhqhbc$+
zx6+IIfflw!r5u^WWVE3-TmV5z2TdfQ2u8pFzttHTQ+F|Plp<aT@Lnli*tB``P~jwG
zd_-#9)2F)_m!O(@>AErNO3#bZEjN}3FAA;xff<hHpkHNQreWx06Ac0C6V8)PY-Zvo
z)>A26c3M8Qrx#W}xE$yl8JRwxHk$6sCoarX2O|xAHO9bD`f1OH551}Y2R9{R-M`E%
zomP>$v%-F~=`+AHGz^Rd;GH6e=^rr82sc8Fl=zTIaFR$;%r-<D;aj7`@JD1oFpZeC
zR~BBHBB8AQ)BlNZ%OBZFK#Z^_byItr-Y;Km9cKOSusFdZ+xP?4n4Uh=LnCU7qy!%i
z$FT#IWoG8Xe1zyjzG=2O#cA~JkAlI6-W+f9U22=Hnd4K{Qxh6TKuX=H$(8K|;W)Z+
z&KUIufb0VCj0g;j)+1%iJB&eU?uBsR&;kllL^@kkxKp6%Yxc^(`oV14wCdW+<>~&(
z4~~q~E1-Xt0h93Q<Htx5dx5iOodC2$$+`xQ6pI0YKc#MOIxw66N1zuo84mhRB5exD
zLYE0ksa}-p*uA9-lyE}FcK@J%`qF&Z`8T0r2`SM!(7!g?B-nG)bLC-Ruhp774_ok)
zHnVi(&_T+AK;60|2bB;RKX(gtJMLb=^jQUJ!&h2NBu_w&R25ON58^qjzN3DFGjT1W
z+1LFAIE5`f3veRFK3r{7^T8w+tR`TWOHwPVtB7F?>s~tk>ODVqbsh+{dSY_VHJzf6
zh0+ax1f0R7fgw^$)3SvZ)MlAB91<F8J2$FsyXAtTH1YhtBlSFvD{saT{+_lsI&Dej
z$NnwK$m+@P8=XrZNUguxy6Tzu>4#J9L5-t@GtQBiJqI}3ru5Sh4u^T&ooH_sjydoP
zu`Uc}hSP#ABf;|z$WTu8_cy+}KsD@Qd*u&^nouJ_7v%NXUMSP5+{tLmT5i)=W&-n#
z@%QZyhcsk^&8g@BOJt(W2bU=&!uBppadUB9*sHpI(Rb%;!RS9o&HyQ6fwU7y_L~$I
zh(%<WHKM70tg}Q#fb84L%a5L_H}DPpa*6SCitf=^1!a3#AnaOKj@JMDW9xsqoGb6n
z>M6g`3CkA!{=M6#(#D4;3kHz4CiLtQ;*S^SToFXF1VPaT%wr7(7Isfrd^N|$5RY0~
zQgSsz&pf{aMm~uSz68qP{)8JX3bis&Q$Lwlo4=46v!1+d1%eeB68(s%p*$}rQWXa;
zh&LMh8v6;dZ8%9oEB+GZaSK2d&YqYDmx)~8GxmyQw~=#?h6oP|Iu7R$YUg1l0+!Za
zW`FxXv2K-xZz8&d)&@{BK%|1YB#mQp;O(X$D0guo>QY$q=eVuiV>E6~2~~I}5OJ{=
z<yD>mly0OVjEB63ypyZ8Wd_~3?o*dOcA(M>q&LsRoS`VV2ig+DanEV9T?`Dr&?;0^
zR7W-6RG6I>F?xO}0UmP7NRAUlNeZe7kKXX_h;(*vc>nPuGU!f%XF<{cS_|wKoC|oR
zxQeEvq$C;!Yx~s89-zlqCEz6HAPz^YjEL3bTij<j9!&~^x;K%Z<`W|&Ee*>^WW3=V
zP~@EmA>pS)IZp}?=YXh%nq;`(^$n10An>r(nvJ!z+UFz>PY^PYI}vVBmpU<Mc<9nj
zYeQZr0C|wO>tzUCNWAV4esBs1SRi2zqf8uL5{w6kf;PAYXMMjZjS4tDc-sJ1VmR7H
z%YUPu6wqfhfLm>pxEt&Uphc%r2vbtA{3B1Lg1|@#Q{m_+d@z+}4ZUvKoC4%54@)RK
zZ@_b@tx3gF$a;cCk{$Yq9XsqGqJxwXK?8t$ODv{s3v@|^`R%F7xo8)%+@t|M)LAN6
zfir-pQ=9)ZT~RjZIq=4yX2*lK(RNh48O$0JzPz6V^h$oFXAcYx#{X@WS5msV&>Z~+
zNq(ZDq880tH{`Q0FhrveE}d*-E7z<2l$4Z2EdA{pP_$?Ig3o{rlgn2pBov9f>^Szx
z{G|CJJjD!UUp9+W^^RH!?<zkV2LLjie*UTqsGNP^iJa6KPwAoI;Y8kQxGY_`PF&*%
zt}De2>8#)IudDe(0&>usPtx_By?*^V;@!va!$4UBIUecrpNa;u3MK<pYY_%Le^(}i
z9j3tYqgj!0t9a-ARe2R6a^GZ~8<8>5Vg)%3EnMH|^BeNPk^dMOSwe}}k1*82HVJR0
z{Gff&<R@~Y{VU|@nH)%@9Ks1F2}~o^bdTRhG~~OoJFs?W;hJv0avcH{kRU@ry@yYk
zwm`ZtgII2AT;n<eN*)87!eS{<Rv}OGAeHVeh6g5NhaO*KaXFwCgH}bviDbMY4UXFz
ztdc~8R4dfO->VHG<RsL+Gp-wiL$@#yq!>!lRzgsJs7l`n)r+d>Ju6FwoO35i9GtE_
zTEEC+nOizKI)F7PhC3$a5t4&ikJpSn%mYXW-y8udB_s1lAqF1s*5l}#$&MqOF|COn
zvh93SWDKKv0~ytvqEu=bxeHOxhds(8uTI@3ve=u)U7-|5)|kGg=2QQ)2f$k@7|MHc
zMug_>I0f7z!!{ojO)1N5$F4H&K6$$#m`6c}5jco}`(K^lPxLhz`4Q!*u7<NSZeLOB
z!AeB2*%}e~)2F<gxM-X&;g;vrwAq!}&=W?Lba<m8jf;s%3SlrSDkOG(*Z}w^EPKM`
z3>iV9`TASLlf2e55(>xyUVuLGnVsgtwd5z^fd~{O1QvX&;Ag<H48;VZ<Kw^FVI}4(
zsZ)9^y&S~b8;FU(yV29r6G}&gn*dRn1>a8znQJAUOCb84*HNpufhaVD=>>Jq3MMTQ
zx?2!qz{)}ms|Y}8m?-7Z?*n{BbhR*>xV_NK+?=gFwfiz~UP4CaVwT-Tm>nc0xePr4
z&Qi_BGxvNr%Q8JbBI5#~)$C;l+Mer;h8YaKSu>;cx3<T8m~2o<1)eP%&++hl;DLZu
zWsfl1(7y84h;uw-iD$2269eS6JhNfjKCTk*dmQo5RtKSurIeSG+kO0|l&0o)LhWTJ
zz4Ke4Won#_afDK0>sS|y%riJWC<;E#@+_N!XIN#+U)r6|W~0vMuSmD->ojVIShpm%
z)%9|*-nEr0YW1&J-NHg`>)bXAdrBwp(u(`~ekonsO37=^43Q;e4O%$?tpwYQ)L@=B
zFHONYX`gDQB86!9Hy%v0G4yFZ%=xFTYumy$YciY`V{?8}H^R2xUu6lT5$mDlsXoB7
zs0{xi_o&7U(P!!LwT9WB^_d8c3f_PEvFekkb$4=4*P^Fp%hKW^ug?#wGlcQz?|%Z&
zG$^QVQlgss!Q0-EkwctUQ*)*2`*NADG#Q!9TV%9Wv?sp^X^T8MDtt*%F}Gu4(m(zZ
zpu>~9wsO7`<`+~}R+d)%wr<DbczxPWwi{*_PP18W5sD50`;O^^vQ-2ZvFVu$#+9i~
z-I3Q<O+SxcsZzuicr-4Ch6%uj6}B2g=w%4)+B%>lJv1?qHYw4^w&NMZOINS%IQZr^
z8=WUEWtPVz;^}jwW~}4r;3x}YrUuQWV#_yNC$ja!1n5SEbF|Wc3Uy_M$kokk0NwV*
zFgu2K<%H(Rx+A}#%kdiy+{n&L%xJ!=uT`=apuI(eJO`*4^0^&B0Qo!h1q}=g%zMgK
zv5vqQekSgAPfF_H7Rk+|=*0muk+uEg8FUB}I|`{TUME4zBK(6fCityP079@}<%niD
z(Y$c;mLVWC>nRfx6uj*G9{Bve*|0gG{D=j6q9aMI#uJIp0|WPSf|t3RZO@3h1p$Pc
zEa#fSTMdFk1DG?cnkBTkWTba-*N=r4r!zG6B`%B`rC#eu>%&DliRn5sAjfFj5PjbU
z>7@sVEd-|6z?)*%ftT%LN(73`ydCq4U7UX>iW$B(7t#M48%l<(3%px4c6W4$e7cB;
zONywN(?*3^dVmw;Bl(_Zc#~T}2VNCCA<dntwEEb&sF*-(-Cv>=iCP3fsNOPs5g?J6
zsHojkcSNbFlZ1r37Z4vA;=TnhgF^>%S6F#?u1SsGKR>eJk(bg~y)2$P+(wad4RL>b
zXTkY&r9@K{tZ6p!^Y*@@)B$(Z3!bq$`2tTX|Iy2ynaZ1eu#XHL^G4v#fTmc278e(R
z_WeawOI0dhxZIquGNSgAFTJ>Vcob2jfRMMu-4P~{bZT-OOYf|)Gv=b`HJN(2tYAvn
zgXIMpTooxgwllw=*hqS?E>eK56f?*UnmKTZX;*tcJ>DHlzDm`u@}3eNuuI1lK0;0(
zNWW93e*wxVhz7a%`7fz5+YaC|F98#HY4V3bZkdpn=vxKUgOQy*-IaI_$*8*@SL{2M
z{K<wk{^j{)4(R3xb_ZwxN>-eqBC52)Ya<G}98`{ceZBlIP+1|XUWo})3Q{i<*@^Da
zHx=org|Hf(|7Z)7!$G9Ntr|m70mroOT|a<yCi$yj!-C{D6@1jA0Yd6A(TbPdhC;l~
zh>DRh_G(%z(GnkmVXF*!kCD&Kim}rEiE6Eh=Nx{%@)Q*l%MM^}!1JpnEukkHYR!kI
z$Qc2=7ihyPK%mLHHlUE@sr&Z+%nX1sD|CO$8GT8$1jgQae0JXvLgGe};CSPK`k@>^
zgQIL?CtMD$KUIPi%i@-u0!j4nR+^~bN!r}Oj404b&zof=H;#c)AJZa$j)7g%^YeDP
z$1sF9cvi0T`#{Z0@7Qo>)>RqrU%hb@p^|l7j$_1hg<cRVf>m`g=RlHHEudX7%_vQ{
zQwrbBL=-;5CZoqamfczBUCH?YDhVtC_uU|1Ihihtc#HQq=H-+Wv7nIuLcewE7aLI|
zuA1Z;?wg*O0R(At&K$}~1S7r-sJ$<3bERXcXzs*WTU+vEe$G6(XK~hX3Dx#f8%LxD
z{6!`g!Jt63cRN!PEJJ;S2qX>dgnYgVCL7@)z*IH<eW27{44R4bL=gl^^q0mz9bX_=
zMaCceb>#GXhm3>BIm7zL&d$!s#LV0TV}(o}qUH(QSHK{U<3#b@+)Ubko!oMR1o?>c
ze+6absix#ctY9qCpKa=Z6nx7zQbw!{NX>*`N;?G2t6S{uF_2A~Ficv2w>M)D2=%d<
zuICQmSA8x*K1wK~s-lkHD;J&wzo56!F1bE@XpkX678P=HbFUf0VPXkFlxs{}*rl_{
z?(d}1vs$vlP)!de9!$E|3ed|X<&GlTR4>RW2+H4GTQxo2K|q^4Cv?f<#hnieKVZBg
zoPeink>s2<#C_lZ5&sY=6a}3i>ke_=+(N5FA3$EPd5v=ZTF=76^r-?&irN*?VQn^x
zfPrvmHS&#z(Gf(;DS}RBZ_1m+`+HT^bO;Vmczzw2d<rw9^CfzEbpxdFD4tA|bpj@d
zaIhHp5Xh;uhY*=XS`GqSWW=3L-$GBnPf*`$$en^z^%>q99LNin7g<k!>C^a3o{j~P
zEeKEutsml4K&M#x&{LghNS#|HASWoDOs^yVhwU-d!n6L<AbXeR^IMZfy5TI7e7v`$
z(_>9VD|Vq!1>i(bnJ@c&V$?G9R$qCtx5F?2VS$~^$=F1au=^8tVrbvhB=g}DlanE~
zpTs-K)9uck^PBW+ha>D|WMzFp`^$qK1IhuNqnrrVF!W5>`C9;q$QdZnyFm2X!{aiC
zDDCRiu(!Q`9)Qm-N}h0O?)|;I**m5{1WcX#&4DcKaX^k{$WTI)Y@wz7HZ~^{t11PB
zB0y3o^K^4*NhD4Xdi3kQ-<r;o3JlOS3j~q~d<hvJru+8^JKN4at4l9Vr-3E~0bxh5
z!d_m+{Tbw;%i;%j9jhQJ;C#NB>2(nb-*=Uj+2vq?QWK^TS~L=x{*Moqzi112d-2ke
zNA3rVTyd*jcP(8&6klOWvHX?a{6IULiXa>W5IUD3=)VaX)vSbaVWEP^^X<p~RUUXG
zV7~nSP9Dh8hLD4m%oD5eYq5SHrv?&mCESP}6M$z<qsfVS4t*Y2JcwzaatG;1r`XO0
zIxfY0pOBBQPG~q*xlen|Ld0$A2H6VIN@M9BsA6O^Mx*|1L|o3x38<P_chxq^I~Bq6
z-{w8d8IvciF7>6@E-B+|i7WqueO9~tCbz(Mt=N`h@e1$=IYVzHaj*D29Vupc9!%`4
z4zX~7%h?MhykTV<$dmB!ex#?MIi&``X?%9mLOkv@kX;MUKlD!h^*>h*TFj+fU_;9)
z2;wvM%aTY0V`Svu1gxWnT7PLnuQ4Yfsp9;G>TE>fQ?*!sWW-WQiq|f>N%gEOdH(<U
z+8U?@k4KMuE=ozgN3jIx^^?al&a(Nl@Lk=yTM&Wi31mVb!;g5&5a*(}>?LIt!4tOO
z0%4b6<Vo>x+2N2lgCYQQ3neG)kRHTi0418jiQ?Z(Oh_t>2F`JaP{<8VO=Sq*AW0M&
zW+&Tw7qfW&aoat@!-P7(5{Hn4wKE8qKat}N*vI5T@%Dr1*9!`uNr;Q%W;90K1prIV
zzkj}RJwlQYPHYts6`OU`op?0*YyQwPD?E%meOi9d6(p<7VfLc(+6pm@P<8{6DD+f-
zs81p_1i%9bvfoiuB<6>~L4C)bvH0>t3DaG_UhF2}?SXJiMZVy`eU0YiE5mSiZ2KrE
zNGv{mw_`pjr-lr4bUtrt|6|eT{148XE)YaW9DO_F@z&X*UC=lDu=4Wat)8PQ)Nf8U
zGH=e4!t5mR?#1d0y{=@x6Z8yC4ybi0{3Q05#TMGS0{g(p@k6tJt}`n<w|QQ9qspwg
zu01*R=bVKPYME9<;EQ0;FZMh6BbgOmL11u@h2G9(`##-_{JcVu*jpN3$IR6qWKS-7
zF8!9PH~oy*VMOE5sRF0Js}pFCYN0&T7wzP#e^;Dl^2X-h${TN5u~{ahaH$`DTvW7G
zjwX8W0^@cEP->{8f>gOPCr>uz;+3mQU1+B!>CiFJzYptBii2fAk^;sSkC>P%@Iw?1
zfurmStA`(y?mO;5P2yFy-qD@9+30t1@-L(Xx{Hp`o?0f#S|FJZ7$;GVBRYCNr~y>H
z12vL~<(1$vkqiLjS>@7wN>cS%9<Hui6Nl@6W;Bm?y%N&D`!jg{7jG8wgWdmC@#$2r
zGt5RsoqxdM0(F8D8Dax!X9)UYNIiif+))j*u$%rNZ9msNJ45_10fE#{Z5|EVRecuB
z5?&J{u}~y1>+pZc1u<iaM8b&Q_lt>%p{mjRm6}r!Rt4m`?d+atHr4#|uw%o68MpBN
zB_gao^-DS?vJyybhuLZSPnQoYEiIwM1tSpvVd0G71wU?UzuzEODRB!mHAB#sFPDTO
zD-qe-AKA9QIOgClB^1L2god}k(}-CQc-fR74!|qmMSy;^7Cdwgst7mc{<SBl!c>+r
zs^7=nv-`%x_YY1Q&R@Ik?c7CEV)tKv%-BRD>OQU$SYfmP@wfnTx^P)4diJ|?L>;2E
z&qTQa57W?HOpAhk8}+!D93~DsP)5o2l}<@GhyMGQ^CD%&UYZga?qBV=*Lh0%{UST@
zfPn7RPb|SQ3R6J_j?5Y@wi*_`|8@;C%8dA5lT!cZmy6$>%H%6RhO-Ui(V*lA77XxS
z2H{~rpztAwjTEhRoK5a`<XZBb;V2?p)#EZBxYA;`cS~p~X3PF!&Tgu*{Qxnu5uE|D
zwE$0IS6sTdgOU`&-nwN<WKAt<M;dv2b?QV_M4h#jukC;`W8g~%g#Lga_U+z1G6N?l
z-m~u_CT$Pjq0gg@@ARgL5$RIEe_tFqX=wPE*jR{=#!GH?*4G~e;wLDW1quW{Te{X3
zl1jUZ0|+Bt3uv^V!j&ceeERS~c$Ub3^7QniM<o#SGSXLU0%3lSX!MCSeG_{5j@aVB
zl;t>)Je%}9-?FQySr`cLv2`UG1SyTTPZse?k|ai?`v8DMzh@S>Pe}N2YDZqJFw(xx
zz*tsQRl)7twe`by7J6uHcKLHVtmOC**OvIqt5c?%p(i%n4^9fDy3XLSVd(*>Hs7Zp
z$seUfc?1$qhSw$=Wc4B(MO1@=SPRf&@h-9oAxSfbg@P0>#qP0iuWz12KKnze7mW*K
zyjWsPFHn#|teGMDBXuTO(YS!h_c=rd(=1CoZAB^U5m99D=??%vT3GNL#XK-GNkXy4
zC^IbX!e1HuUs?e2HRNfCkfocq1pOLv7_uIB(Oq;09bT{M$;qikT_#%CN(84=D6XHP
zM4VDS$Y0W$WIQ6`PJUry2{`)<uZaFu86}vB^;0Qus>&;z%DT0MZvAE$+N7j{^*?hE
ze`ZuA_n{47`{;L{Q-85r&?o)qPQn8RiWBEbte>k(oU6KvZr_Y%@p>cgCJ3FueZ;%D
zTf0NBQ(egBVXr;c@cZ6|L*$ncCca~_etl))`YwWO=LV)54-f8H{}v(hsb(d&?;sOm
ztq?x)Lss+Q#=HmDZ){0ignbuNzgSCp0{I2~L+@`q2m1=~I^a~K)*?<f#VW3^AE%db
zwCOxi5ECTIZ>(-mz-S%+Ktr;46{r1W%&qR4gJs8g<`1P91*D2xd1p#DTm+F63P?yv
zAuh`S$UzJSc%;DhkOI0%T<OxKO8`(aJ!nXiL9%_T-S_R>)}L6#Sh<&gV2LV3aE?)f
z<>rkWy>Rr1`dV<cVRFSFH`^pEiF%~Si@+Gu?tDclAs0h_jjti28~>szAspa&U@}PT
z)3OAB{|96fQ3i?#6aWuH5w=t!cKS3{fm|TApR1~PlLINDxydsx$z9&o^apIA??v$V
zSyN&3bae1dy({~?e0^)yCZPL3<(`~2vw0+L^2$;0?hLu>`>9<WI=$4kVyvV;JGmxa
zeX1gca&_rt-TSB#!c3$b4DPZ7?F9y;Nn`dnyaCXO$pM6X3K75$Y9^BWrMh>{e$Pi%
z-^mf%iyZd498rN>5G^Un4I6A8r*eU@x1O+umxbA-J2?&H;d!lLYr49`=E^zo{ZBr}
z^()m$-#oU9SsH+j6&8U|n*~}xTPWwUYwI56=wOSlt0gkX#T={mlW`tqpPyfNTD?5a
zIV}FbsWftp+E+%*6FbCM<QI#C34UIh4Y$$MKO5m6^+9@Q<^1c$ku#nod0Wb;=g$vw
z78choFD`mr)Y5ux8ZH6=9ceo@0SiCBUeNvOx459imA^(4X6qAUMVEVw*hw(C$})ed
z=Sl8wR@oW#x#V;^-O2{fkct!mpLDYZ2fDt&!NEy}YSFq^GAnFfFJ`pb&{3cx&z(4N
zMGDvgUS9mXu_9IUN_L#VYQeWW+u6B9nb=oG;;!NBInObdi9Vy88XJQz$O0JimQGsf
z&e3zfHW5`%!C?%sL!vsW=N@b%`0TMvtd~B13=9f+9%a?(3+~>!Iee-$=?vHxX6ZX;
z*j`q0VgX0;UY1sxhifcT`So&>XPpH<x>QemId@EuZ@<dnZ&4mL!vG@<8S6Nt-rqR(
zwARq~!@K#Z7&Lst*c=(^oU^>+<Wy9=;m{MP5==D!Hu?=Wpz?*T!iK7JwS&JP=_iU~
zBXz<5j_v!No}Szqc#pOa3>1EX*#7CV!o%QKygn!H4-2)%s4`1uv=)ykzhV3J`?p#~
zSwEKh#6%QcGJK^J?G#{tA#%YX<Axgpc%Ic}aXvm@^d|Pmd{#U3`}bYcW0};f=0=9Z
zfnRvJ6MOu8Vt#29+{PxA1mtQb6nTNkAv%P&nZ|{vIx&FEMB$G|ub1(2s;4wK?3NJL
z7cHn~n|*MzP|q;GIq6xqWptN#MX4$=-6!O2Z_R=P^sd6x3cclmEKp+K)xARRH<CWc
zWX3tZemt+aJI!#ab!z%=Z>&DCMg(5gyH8XygI*YXBI-Bxl>})c_XVp3Y79YU-^t2O
zDqtYn_kj#Gjb2PcA~h3Ej~jsOQFSAL8Q2vELb`F-lzD|aV02dz$D;yv!BI^TFDK8D
z=35`^+i0g`$VR0q4HuCf@Ytz`-7w#QQXsaK#Y68s=knbom-sR&9@m^R2hn~tmFNMm
zi}o5xB$bV1%c&v}FLs}vhUOV438H{2z+V6vz$M^d^xdaoS1G<jDbW@m7yta$y<;Dz
zG@{OKl`-|KYt<%-y^{(<<`L)vyMm(P$=XXuo>va?>l8HDGE%r-UBAAj!T$TG-tdrE
zbx1ysV87+*$ZyZQo3P&-$uH61h0lE=vh2ebMhD5?i-!9}V&>vGa%$`oOOcpJbK?5t
z^a<mxL*NFtu8Q#Qm`hHRsr~It)y{W?Sd=7GAwhK?8!Y&t`#9n8`~unLx$nStdT*Na
z=@zOwv^FP)x4FG(uB;*VarA30gc{+Wvztqld;UJ#y*oH;sQWvfRHX%rqPyt;Z%K!w
zY5c;Ld=0uA9Uw+XI>Zu`^Mjq}Y+hoyjQ)IN3-OD)0=^-!zBo$Afz)%4i6*;KOfp#Z
zJdkVe^nT^W`bWXUO6Va$tp21)NgQtq&i-X_T3f^}H_t?Y{4vi1+{famQD3Q@m_pQF
zx^nqRKVS<a7FsmSz(M;yu~>zSFK#(;|LY$4Qhlpy9mTW1PNz7p^&6e!x|U_2ZTIqf
z+rr+gGg}@8KH{M<Z<D-W*_XhRqrzLtt$5*R=VscRf8_2f+;$Ip6C&A{wL<-n=N|=4
z$<zCnDQVi|*gCw;KNs`V-oEE{iSuE}p**u%&8gPPO0K<ras7S!mt&`1+TfXPuww&=
z+az+gJ|#A-i;>>hy*!^0Awzx@8thi*K6UXFHc<-a>lr^iNp?}(9kFd&FZXg}#>;)_
z47sO1>XL`TF&30CTkp3`IDuP){OOyGl%nnfWao~w13R4s#*x?h3N36b5xhR0G#M$_
zC9J&<m!!J>Y1Ne?-+6}wPUVjqJv{8m7g9<dK{w>^taA6s!|^pXvx|NZOKw5``#9!W
zKc++3&*)nz8V>e<4l0v<B+?l6!u#n?zG-wncJTPAvVI&(e12yAcmsA2{7FsRjjATp
z)%kHT`GyVCcn!@+*LsaZ`vk&GO#S}zF$cDf(A-6CGj4jkbE=O($-=~SAwTk?o#n?z
zW8{0gb7P~&8^^aXcONZMQ?9>gmO+fRJT&T!-P?9{%MKmP1GG(&SjLVcZ#I*g>!LMY
z(va+3k{ED!VEynIp5!Sr`Ih^cDA?V1H}jRyCl&u>TK^KU)5wq{x#4az;LqOVZXoTs
zfIHEW-^D!v56#U^zC!dhobJu6>tKI&Rh&GmNhh(t+x~yXoy+8xZWOAm-}5#*x`py$
zvwGb^kLtYJt<HF&xk)FnGfZ?TF-*24y|7CL$opLnHWyFoPStRa+@7Q(c*O>#_2V?`
zKi)9(pJU>jy!GQd#OGJzQsmL*WFL}5{^Y>=U#GF-AA`~5!uq+l)=Lfd`UTeCT>zuz
zhiog#2gbSlc{Q-Winhhg-5t=eUi8VG!+(1BE;Ug6*Xd<AU8|P2#BJt2Gxe_joLhj-
zzRf~zIC9J1PDayj7)#LoPXlg2>l``1-k3%>PE3CLZdp3~r^WhLSTSzS{O3|q#BH~&
zKiVtY>v^Jc&E@OPLF9oDD1k34ezTpJwH=)iLUF;{>ut@AZF?kQ%~%FUS&xn_mmQxR
zRSI9D+DV>TBqr=Q8O@PsSTfePduo-cS+V%(`o(y0v9I`2-9mRT{XS90p3uW<zrIyd
zJlNijB2=XPZRlUAr2M9mllld&AJsX<E`)>~$TKB|S1?}q%%PND7GE5c#oc`Lvqy>|
zzn2kxhH-cMc7VMh;lh&+V6+75KJc&)BYYRq7woRWeVqKx?zRlQ_qo^S@S7c!qJb{!
zowNhPJJ>fZrgLdhdnTXU6C__WiCiU`eC0kf*}{=D=_%MR(th=@*GO?24S4XtlZQO&
zNjEVV&K5a&EcEvN^EX+%MK{@dVkd6w-Uu^sR#3aJeyO{oho$>U<I0tL>sH|nO1Czg
zCr25ivv`mOic=~MQGfb-qhEPXun^yE!7J_EvOdfoa8?M_qcUGnWMF=nL0iEqUip|m
zH*B-`6@vBiJGOTP&F^0-nIE+LP8Y>r;(OBmSBydqN2T`cS|PXGM?Y?1Pjjg-1Em`^
zyw#tlQJepa=aE}0hgF$BpaN*o(J<z}y9U_uo^s25o-IBz`Hm~lhFXG}gK^7?_%Wx&
z*zLy->9&?>W|Yl7qF(gd+~M`L`gfGP@#On?Y%WKVC}HS*fxU&d&$S4I*Nohy<-lG5
z|GvEuQ}mCqKsX-=|KVAlA};fbGmclfT~B2FVt;CvxhFAx2xZN=<%-hn$8EkRps@ai
z`x55n<}o5S<v<Qk*lpn$9UUbe^JOKF9u?k@pJkfC9TTUkBVShTuqo~e=2pLY^*|9<
zn0|H}zU{$~l5*e|o1uXL&uAKj)DNaVLmrZ^q+Dwm|6br=rZs*Z7I;oGVmd0LGP3l5
zVacZvS-R`@zx`AHPjRnX-;OQy+@n=%T=~cQ*#xPz`B#KU!*4!3wyA0}efEob;1+{B
zdZNXEZ$Y@nS>f+|;GOVzV%sfuf9%kssZ!~&;A<Psm$?d7i_~C8{mcEsy2%ROPQm_$
zZhfXPVcfQ?FSUeOtoXe?>TuovSVebO{--SD&L%P98oS%7UF+<7f&<tD*H<bMR*_4O
z$P>u@5lraz<m(O@FaF#LC4WSs6uq(I;FhT=g}2dP*Y6|2jnVC}K9UJT$GvWINtqbc
zG#AJ=L1LWxmjvhf(^@^pv2%9l^2@2RL;Hly{$nEd@7U-Ny?1?>HeAMW#7Q-Wa?QOb
zxid?N7rw!Q&n8Z(k)NecUeT-^ACfewDcEmopN?u>>egO=f*hFo+b7nK-(wct-u|EC
zofPa|LgYpwwUcms@J0QY%Yx$4&;0+a#iSn|B>ML>>rG?sjpOro`^0K>h2}nuO00kJ
z2j+Lq`T|UvfYr%CL{B`%1o1ibe~!@vFHisHM*Aq(&+L`yTa8`dS#LjgI~eJk>t;cM
zlCk&xuJu+4yp7k}efs*d*{=NcF1v^|v7z7*?bP4oonrzTH<uxyBkiZbR~``)*e>(z
z4Xc~Z_%K*DIiYmq$l4zV<J=w9xi$UFOEGuz`Q^MPce-P9r44iBJ9y8lC0m}Ko^B6h
zZ#%R9LuQ<8E2Dr>jMJw5nbf~37QV|Xoxz|J;6JuL3MQt4w#P5*2ye3N8K%D^eplbI
zy!*lxwuv4-5)EvS5{(Dj^>q8!t7_lWV#)5sZ8s{)`&9nnr{{0S?<k6ozAas@H_^sJ
zI)EkXA=!vu>gTa^RB2_scWI|2)Ag37d%DB>MlTun+@rWCN+}xL{rTPU!Qh15<Yhu)
zIf>SzBxPs|{uynS)Ki=_e$`sd$LjhndDHmm<`%R9`~=y{WDn*?x>Boa%1~p@j7n9a
z@wvGHznkiOv?&KW)N5J4QNEHnJW)HHqUEx@wStNI*G00YF7ZW8cOl|+py|0I1I<lW
z>@VOBDbJ6_Ej0xmjdN5_^LVv#_wMC^Jh$zngbVJnf@DZy>p4YSRyE!_Dk0SL&u7+G
zfy&MDClyW2^qhR-E(XhniTS)V&UVyq<EnX<rQfNkt)3l|l>(u5sQhP&i!vJ>xOWr7
z;_h8P-M4z;Vf!?i7q8ZgeVP$14mRyn*AH;XDt*G)!IPZ1bLmdea%{?I(Z}+mPsbsh
zIBhL)nX`h+`@6iqNRO+6z*$4`ta0}^$y(#9p39r$bm|C0>*c~TEH$&^^64>BVb!rW
z9gU>JSOW$#$NJv%H0IXq^JfuSyzS2^-gG2?Zi=$P2W>$Hy^h6izUAnUnP2dFp(Xl2
z`i%ge!9qJ8x>qYFYpdtu+RXBIvL9_M%h&8GP-~6tn*DQTe4?9+lpx_Q`zdkf&V;fR
zOX96^URd!Tz1loe<gVQ4lae}i{jFY#v5B)Hf578G-3j>wvw7*>^xyg`XDV0ksn1#H
zsu{l$y~b`7UT-)m64*)o>t}-`M@PGIv+}tR^=9&_FHy5-+r`@Per9{h9-lF7A;ucN
zns&c=vuk$W+sie(A6+)j+a393eB6vf>ew#+qeqqR29UUSQLyg~JKnELbeKf{&FfQX
zi%t1`(@eW-W|?_f1%w<nIs0?+O+VG7BkhG1*sJ>BM0cX@YZ1bm3YOfSS#l<cC3*C-
z_n`Tk?ia_6MYGf*Up#*v&wJ>;g!b!&%_KPwQh@&flVry}B@IIz^|p9ZNfGd}Y*HU)
zene}<rOYV|8a&@|>HYhek1ft9XBEipT~kgGJ%<;R*mJm~rqS)NW0uslYd4EGlO%49
z-W6l}@S@<~QmMwAdA~tNpWsy9HlOimH9FD|+L!8U1^d)a-I)1(A=L)RhP2zc{C=iC
zWlxX7ch<*)GRy(|q4p(n-K71y8G8S5{eqY<$5->FV`j>ZP3mqX2g5vX6*7KcE&ewG
z=S1J)s;F$=!&gG~pSKwEhwV(&QdLO{_Fp<#%c`c(d}DveQ}01tW{;b~Bne#aAHr_>
z`Qkq6C2#q@E1XhPY;dx4wEMet1L-CEJ)4S7NsiA`JME-3#)E2c?tbB=lWDw3r$sVX
z9;Q6{@uy7bMXguQwvWW{PUU64W>_jjl6b*cQ5iQ7+jF)|^WeU7yvgS|q<D*;oORK^
z@#R_(KcZT9fA0CYZXfkiWtMx@kJM-L70TaDE=l~<c;ra4fE>2K5SdG*BDp`@Us3t^
za!ZCCERd8=<83}ZTbKDD*X5x=F6ncsW+z7{e-|0=B%<W)CTH=cCP68G)M`_W@1NG3
zI-<j5`Ga=8&WHb#%8`0hZDMYwAe!5Agts_Pq{Zc1?}f{S%I6#vX#IChM2V^^%NdtP
zQb>GD-B;ts%dUR-Xk*Tee-DHWx3hdZO^QZHRH=WrYpG4PmhFdpTIm37n)2HG2qWpP
zgA6~t2GRUP@7e_ry=$a#Iq5!hneLu#-)XnRRr@3a&<xrhEZrc{_%42CRN@$eTEHI^
zORKDT^U1X||Km-^$e4llP^)B7?XbP$ew#@ZBY1Hgv$8D1fS~FZ;+sil<94+=^U<Z0
zIULR3@xKppH4wJ9SvNX7y~O*-_rxvX&7{|{-kN%BO+ISPyrIq)wDx!JbN?7OZFn%u
zUX|GgWCu5PfIq;wlkLv<_iqhcMGQTwgGojPx!U0*Zk331H!W4iq>{^x()QQ(9IR>c
zdHEANW+mZ*YQAc@e_6)OOS`r6y1IFqM=m#SmGki#B*qGWCHckLZQC1c2P^bGJvud$
z8SBS3U0M{-9^l&V?myMjJ>%@F6e6QTH5n3jeOP~0Ih;*aGNU2Aan6cas2KxpeNM31
z-h+f%MrO0~(+#8s3WnaGi<gf@yW}=b4?4ZbKiQeY=M$IWC&m*pLgT-yOQ3W^;4!RT
zqSi4dQ!<rW_lc?IW#>(P=ob4?eANE&_8Qj|J010YXBy{K%}65+>(HNk-$f1yY6rom
zrxaznkdmBd+4l4MjE9;wtHo9W{mrs|+x)APm&`(4JBGX}mGvYu5(`Z;3gk6*<x9xU
zJ1Uy5{$`)Qqh1?)d8$^pt+q@lJF`Sn-S^)V<;ImeM?!tCWQx|O4;+@?zP(0-FY6LN
zf`#K;)UQfEAN23YcX>75DPY_a+Lqgyet%_p^vwy6+5-;}D0;UPN4Zw+S#k(skffGV
zmQ(eN9Emz}O<7Sv<r78DlY{cl+l+tg<LN#$nNy{y_u52;nLfO_zpPuEUnDHMn-w)x
zuWWS4*;m->)jykCeZQ1bXFQlYx2HnluS1%0c@natJU3C%KQS>i3uP&~N$;5T*o;29
zU)8oN(WB(~P$QPz)q<3Sk^X?x(KuJnA2Tf$FODb#i5*-!VsywljNe(mBiW7bpe0|{
zlgkjgUBv$O?U4zGu20={Dby1XX18cGxW+2j>DE&CN=aBzRNX$_B)utbiqm~@%fq|Z
zFA@tbGwoVOYI5Iey#!-D^*Im8P%Yjv%Za6X)B#>U^Y{k`CI>&pl3NQfh%9UH%Dv^f
zH2<Tm3ZX20GF_p2*p8n)HpDMh8&Oe4aeH@py7Ts&ZtXZ4<Kzuj*b3!0XuKh}!p|h?
zU$?jCSx32c99V8q3@#Vw+%=<5HT!9ktp7hOXDiH1B>VWjEB`INq|P<9C*C?ErTK$x
zsgS3lpny)m>%ek#J#~hyr>9@5y_je|CsvRE9E+RPn)$~i0fD<_e{`0d5~*G76<@cM
z8d-h0_V)gTr=RF@5_g6FvtXhX?P||&=~SA3|IndXTA$|*t2HM&<KQv|KBWHTwOv<(
zZrs-FujIMAy~Q$emzTHwj&ij~(hhh(xAo`k=U#&!&Zk{e-(I%WIkHr}T<34&lj)Rv
zmkWyK4+89R`SOQtEo#RwU^SU3*wc87pRdf-g&b{Zmsfb_S5(O;dQ0ubc$-tg!TYWo
z_~YJc>bcqH2r6#LbIui2R-GA_mv?3<HIQ?9WxvqFe&kPd*5T1tuj)dKoa|~0_5iZD
zD8I7pq0B?c>N3la3(Ijn)*O6V!vdp$eXm{$1bG78nm@w2SGz3Gq?|8g#CWoz>iFB^
z{kDmzT_W3_Dn+F4o2BwG9rH0U@f&Mq)tO3DYiyb)3<8ga>SJOBe`;MbbLrZ>O#6lU
z6qD}slx|ivb=|3nH`jaR3;DTp`+h)?7x)_8=AQQY{ZfnRjQ5nC>y`Nlxwbb4&R9QQ
z``7pGwY;V3@fT%pEPMEqO%`{hDOlY2(l**~;vxURm{y79<0*q5Qm`N^@>lA`7;AOF
z&657C9zs{Rb(!0K)T1pD#fLbg5DD7a+QtxDH3weY=pgmvaP^hz_w-Wy%dFLZ7P*ee
zeAYh9Tx@+KgZdMb+&8C%Z36)s%iEaua-P^P{Gm*wokrbnDt%CunB6A}M6~=luefYR
z;{jWeQdqicuwi7CdGOcUsmZFWv-$;YE5fSbEgOcujth?W!EW9WmebiqeD$|gp&<9?
zBa++F7h0|b8DIfMC36Y)PYV}qbLUHoUx-~&zO&ggfckM%QetBMg$#Gz%=gx}-W7_c
zYrNUQA)58lR71(sK4*M<yr`%M`y(vQ<9$e<ACUs4G9!3*Ja74Zyr}V-YvYvdhwgK_
zk~3bm_Brp(wIWS5l0q?n_N!9=T4et`W%+mi5h*9x@$Y*f*dTdrp%g9SN=|Wk9o<&>
zyQ#_U=hjOChr@PKHFK&fj!)ixcKSu;`RB4>*(`163412-Nz&buK@!;4eA<f`yStW;
zTe9s%?$Xexy4jy}GssnLEIFT6kZ_={IQE%cBUHkq=)DZRPa<@5et!R5nB}aIeWgb^
zIIj2S9rMmZ@1Mbo{P%EiEP=R$$+#<DeQwy3|EX-mrm(ll-Y+<AZjx2hCdakXXvVvL
z+qt7wb@VwU786Z!Ck5uazWS${xU~L5|H(R5virM}<+wlkg!FzV#krA-xu@nQwAr+U
z<XHw`K_*o^BN(g1`9d16QNNlW3C|i8TW%GResA!tsAe$g?s=<$ryK8{HGXPkpEJ38
zcN<-?Y*>6sN>bNXT#uA<kkW-IGc>e4-v#Bbq%S|{`>2v|@Qn2(E2*+cYY$e;hRGj%
zl-ui{&`;!thL2^vw*RcYJl{IuZ#QX^oUnIGih$f`{~PUv!($0>z~2=L+soF})gke#
zC_P+eE6g_n%K%c;#=E-MJvnn}%V0%!l~2+CG&2utTY6_DBVRWpWll8L+_ntS5ICn=
zls!3+XJ_%k{L#VwJOSYd7GYIgAEy4v+ueaXsCo8>+2`^-%eq##j{9&`*y~^3AMH}w
zn0e<x+H3C4q?)YL`VKcbH>TNGP%gJ@)b-M)_4zw-B*5%7tw$u^Y+6#1PJ4c9L&)h8
z2KV(=c)FdU+sSP3oWZhTd~Xu<qO{MIiLWOdhHi^-%{qTtESaBpG0|~a&epb1M(orP
zhoK}aK3x&){mv|#nz`Oa*BFjXSrxRKj?O-b?@s#7ZYSAn$bPQKPNJuA;!*_fH}ELy
zdjy;bW8GfGsy&tFY9~8eWli`+EOxROy(9*aEgKgx@QLC&ka?hh8tUupJ7v;@RPRHd
z?98o9BTZ#qsZ@uL?KKtm?wnElJ0Fep@b7Hm?&9Lln`FmdoXel;IwILKL7gyYx6Pj|
zHuAV+_^s8Wf9<53*VvD?*^biMd2+AvR=PS{442V4uiT5=ywP^GZ+pBVn-w>M%U6wi
zu6eTW^F*`?cbsM4dO+VkN6E)5QHwx+_bYv2?^TeO-@&hGg{)8xDMP~{LTC=U8tG}`
zf2RXRMruQXv}bi}Xcj|rquT8)vO`an<i3bAt_z*2{@`K#^|<!Hyk&6QJnOEieHn8j
z_F9$>&VADd6-s-?{!K5xrdODf^~a;T>tI5B<llOR)cTZ^nkGZLxbos(@BfY2e(IG&
zf^{*sqrP?+KdbEwzx4cfvpzSY&pxtw1~pl-dB-!^@R$;kPsJ*ZYUTVyMhTJ^%Ll_i
z<5^=z6gw#KK=;q0dd@Qudy5XSR_^N4u6J&?`bWyr3E$Ktb(Y;SUCfL8v6XHg<1QhN
z(Qkh=O9oQC<fr~of7OuN!`^W$$uhB#Ci8iExTmC^`mD+GJC#PSC+$YlHLld;2Hp`0
zo9kY2y7=BMqWD5^7EAn-%E<Jw#BT-AW+i479nn^O>@{hj&m)e9Bpsu<LbZs!4Im$>
zoUkLd1xR{kj~#p=z+b7gyhpAn2+0q$wC2+~C(gY2$9(?doS3PpQhxZW*Gg}{-g>|_
zYa7qgS~Mmp>QEBry}{*qUR<2nJ&R#YJATVp@4;XOU1pw-U0=Ulz7nBi1RWw9!Q){X
z);CeygIrV63ly}B-gM-P7b2hxk*sL({Y@Xzkj;IE4|Jy6_ps*}Y1kcW_{mtru&p-j
zD%IAMh`%SkesgT)&~>?{ZO`3sDR=EF(^NlC_u{0cz{9FQbM_bGZ{(I&{MLFUsW}o7
zim#7%iIq(eLx*(0Xq1FnQujEdkZPjleDB_0q)7Y8cKj&x&NW)lh`g}u_gjyL&qDE3
zyQz5=-(3@H2QHe@x@TJArLB*CU)h<Os;78D;Oc_$bLA7XLmU|ij_S8c^9>I~vEJ-C
zSNqQsLGzv$n)gQ2kBHWbb_ys2GjPWFe(f0--=X%xp-th$k=%!b^(DQ3ki3DYc>9n6
zwXjtuWE7yeU=<U}4B2K%rLMjw{{o}2&J#1qms`Jca4e1c3wm~$+B;163-Ww0$`2R5
zT=uzc`zC4hym3$Mi2&zhonR^Y;;vgdui1~Mc5>Co+*-6~A5rGdm-|uISywKs_5C|V
zWyB#3re5Oa-22euqoXDqQmZWe*fEA7CL!TKS-lt}^5C||@=-qI#HrkEt-1kf8X5<e
zP0bU-ZZ~?UCG+OfvybG8RFBg|b*Y*tPq*zo=iEQr!P<7d)YwjLhSxeTroil5%V>K7
z{ke__y;Ei?Te^8<TI1qbXpJ#W<}{w|`(Z7>DtK)6ZeA8j%ml~f<m8;0KbCwku}u00
z8N#+Jj@bTPu+L9i*z*Xq77rN&!W9A*77FD!MXB-8ff~VbUqHORnqdZY93;HPjbpqI
ztqOI}f-mQnSEpSq?%bKSsTb2IUwv+xp1gL}d;qvW?nLpXZu-Q8`HXaztXB81QmrC>
zqmEEB>Cb|H(|FtLobzJzJN!E2!>nh%{`OKIoE5U@x~$$+Q4{gpVXd*V&fHaUlD)xR
zYxpIVKb!8zk&Kp;9Pb;{T2DFZ4&9v!JDjgyxM;V0^t-BD@Oxo1)K9Xo{NXJ0UIMa&
z<8xE|Pi)Rccu)R2c!y5gyN4?cm^Y*%msUTiKmbucoSNn*K+QTgo3LPQS!=WXqTLzE
zyeWYfYy7D-UbIms;+iAm=jK(aVmD_!xX<4oC2>q)&S=xWhl=JGx;NC1wVB!lq+?FU
z<lH`0Otrl5EYDnHgY#<zN$TSbgMSqBOTt#n5=UNs`EBs*f-OhM`DUID9kW?t>E2)X
z>}-XRk<qeC88H7OyI3ARKEASgFj_=)oUwe<=U1;?bJP;E%v3AxdgXCaF4$I=I#aCc
zilIIK+RT;wJ)J?5h4<cHa~)0D*HL)KdStrF$hAS(a{gDQ$Dwpn(cMNDC2yxRG_@xP
zsNAl0kk;!g{#3%XX$}dmQNc4>YuzTcwjvR991j`nIMg__R{ZHaEBmaP_R)9+bxeeD
z%eC6p{o`BMH7T*5f%;^2Ddip{#Hrq)-AJi;-NEDIuHiRDMQ52!dW)}}tQoZ4t>I=x
zr~PHnS}wTh_kiq4nPBc)o0AfZX+%8}((X}w{@6s(-9JsuF*?;dG3VsLfD|pCDH}2N
zdI+CJ^TWf!@`kVbsYevMnAu-Cbcm^_AnRw9+v8`8A3xTAQ=fZWJ@dhp%2Ihj=j7j&
zx6Ld@&I%$_D<Awr=B`C7#20!n^k3G!#U`&)^e6jwPSC28_|;L}YpUug#*$tfbs@W-
zjg(J@gdVtSy1MO1O;)=0>-vP?urR|fkg4QpLL!u4{IVkKt8EuG<R?2$)cv!cq2XqS
z$wpP|>hYKrA1TPm1#x??d1!yM8|8ctHl@PmBIty7P2oHXqs<D`nb1uq&12=G;lN(#
z9WwQbuPufZx5-_;-^jy#JU8x*y<+XP?MMH(={hF||6XaCYdc@ga8p_9qm9^~b4`|2
z<<7rPeREEaS&5Zx-7&e=9f*hiR5U!nezv#cp9i#_^nc7Gq6#BKa&M?3g21|_^Ym;0
zDkSPrXQI|gRR$~H;lr>UeO6}<C`Wb@w@<%zEpj%&<>iwF$Eyc9-ftHvUzwi1l`Rw5
zWjo5!>MvSsZ?0IqkkOM(b4SkbPxaqY=cu~&@l2bxxUkxIg7YK24@y4zO1%H^32gCL
z0$m%*jMl#pQVugH?<u$Hdw1j58H4JOD!-FqN3XNfdZ6<9*&7w^jjri5+a?-?G8Dp<
z%f9~#$|!kKZda0#<6A!2FVIz@J9Ad}is4zd*q4V61*dVw8GN2ugoP_kUK1h65R#U%
z;F5m=w&Ei8M@0Ej6la7s4moWM)EiL|Vjg?f$S8d%M3I~g_bZ6XAzz-}DXKTWYHVyA
zQ+F9z!kJULNbcvmk$VZhV4L3ZJ(M>?A|lMQ+wSTf#g>+LufIMlB2t{5?%(`E>tml4
zWLDH1CZ?vTsjTgNOtr`BUbT)H{I#pW=V#dHI5vXE!p7s=a)${VR+b;TTp1V{ZQ%ev
zM_K<5Hk*p+-Xp<%<+|k2huCv|tmp4M6vYXJI_ao9;3ygiQ6Qld%;f&nB1+JiB+`}4
ziuhLSu8Q(YHjST3uvO%ddjjOY3>nSS64DenexP@WAAV3;#Q*Mugn~GD0}zM!pF}7k
zi2o}nK{OLTn!tR?2V2015eGLY{_7%O^NA1se|?c??xGDlZe+JfzEY9wZl7A9dO!fE
z%)#7;I9z{{?&=PY=W0qzdvMFWtZZq34SY*uwp(2tTNP{q=}7nB&8d{sR8v#aO%L9%
z?7E-!YV7R32c&o0{ah?Vzfp*DHEL}ok>Lk)PzAmJZei-$%N>hO3JncKVP+S|wY(>{
z1b;<23?vI4d{{8_oY=_X$<N9CyP>eOoC8hc1x-~c1kK))biqp3uC*Zt=ph4S?L-oM
zO!0>OZwns&e{cPY)7A-gP;$L{C&}?1dw^zTXKyDS4L40qN#T`;Zxplqul~RsQGcB8
zVO34dEvlp3?=f)tXO|u_AXUzaSkv%C$97>iOkW00Xk(Gifsfobf-PMi1v--due~24
zWl!3cnwsj|=dlyX=_;zKU*VlWm#B;4rCIz3i>@EbMT#nFYnq`t3AS2Bg)~&_d5+5V
zO+L9q#VTYmeD7^sn+|ZhJJl`bBJ}b6uAa&p#s~K4ZhhxJfs|Iq|54q!2SS<leOx_j
zk!;feBb6pvdNK}=S{)o}oU)P>i=2{8Du<nt^HOVUj#)~iuo3A*bQ-7R&`=1ANC;EU
zI)_Y-GxL70df(@5pZBll&$s`YRQFu>b^lKH_5FN}jEt6x>EKZy`lW1zoVyvsWHeGa
z&9=m!39&feq<%iK`tDb%CIs2Gyi3dwNcEZKbipLxC^j}W*R!&S*YIbM3$!c2(1K+I
ziJd2KOh?EhAauig;L4R%FcSx)gFu|ycv=lQTEXp~kAk=NY;-vBY(0+UaYm?OF2PH}
zS}z6oXo_zy!M*KzKc&rl0`cqcxIe2<?C+|BP&%uss+LUtVb<m4*J9u%sB*Pk_=BNn
zcMNonJ}_#6=<&^oTKF(*CD#O9GgKTi@NOf{xlN^Ea*0Vv`V%#ZDcR6n48U&@4e-lN
zgdY+Dizebrto34xD^Mn!cq|Pgf>5_%xp1T<+EDQD9FxMC(y&dvWFi}Ow=xI>H+c^s
zz)i%}4bh%A&|K}V`C&)+J0MC+Aht!<1t%}9&^}Go1se~2Q!)U9n;={kTxW4!5SPe^
z0OzC&XO#<7Lm?mpw0f71n2RwE_9D?yp+ka91lup1QlN@c{(Q)H+fMOQ{kkO-j?Ta9
zfX?!WZi13rb=4MeZV+{wNb@<huU-C;qd$0M&vn&W!mW8dBFGcY=8eT?ug<x95mK1$
zz@a<1HQ<%nxmG97ERS=Mz&Nkl2T(I*78Tk%9Ia++YHRyp!xrVjwUx$n3=2TvV<Ap_
zrH$(CxBq$NWJTN}+QKtCU}jX!Wo}C~FF6nsrt$b}`Cy$+5tU^+1YRDI3B%jxIQO!-
z%jlU!$G4(=K~0amjj&ze%s15hpuDDX*<F=|Se7?s&f@Xu=?bl0_o;CUEv>BUpFJA@
z&j!_%mn_3Q#jTUBruOnwY598Jbx%Ki+1oGbGNX4UU-?EsZytE`cYi`fBADql)>>j^
zTMGt+uwp~J+?Iq(`s37#r;Rklx_!`Fe#_FxJskDpeHbY#{cfRAN=%H)KZ4-nAdV+1
zl$*=!&8g<v!LbBf)qPMFoIl@;W2QJTIyuy)K7aem{dv9_CNOe~!a+2Ve&znpGg0TN
z>(c!fn?1=8AIskF=tyVf2(5GTbB~&vn=hRG>Ual4<me#6DK-ww@TNoa?~2ImK>j<G
zsD$`<CRjd@g+o6%ngr|T!%2>At7N!qW8S;m-KG_Xo4Tw6NtS#A^OP1`QH?LYr;pC(
z-7jNwT08Y$wLa($jEL;1ngxqP%7hq51i)*=;A$iExs|VL<U$LJx<<hWjDbPB<pmr>
zflhA*IFotAf~_9y8m{Nx{^d&&6^!@2SGkX82Xp?<|B3U#=?yma_VTS>!O|V=N*mXD
zV9Ef7A-r58n+HnZQ><3hpD|_tqXn$tZWIg^Np)FSStKwx<@q6k=iwg)7{RFSi|zBw
z4GRNs*JV&w`H4I0boMrl|Ibd1ZohOcf1j7NwY61o6J%qFr}fI$!B|58GcUqCq;J@;
z6(n$lpyB+QNi|kKpzjFRd<k!k3xL-eqns}{j|ceRG!`3vzC-#yet8FTo?Jdx&0Wd>
zEX(LYP^lBm5a;KYH7eVd1N`b;Jn|423Q$zLf<M#)lN!k{#4kofLBT&9+t!;#eF_pj
zR&G=T*=r}jDnj)M+(WCgQ?Y~_HK*Ud?sG(qT2wMH`ENC9h75PXV%RZbfozT*9Y6(+
zu-0-gFz?2V0K`w?95oc!VgP^+BLY_-xBb=UZ=_oPmOKttkP2iC%|a~e`LJLA#BRMW
zmJSa&Cr&|~)y`(RzMY_9Y6XlHtY4sy-Q&=9&<WBwRTUM4fMHM-VleB0Hz~P25FNqI
zXBDpzXO~r3R2~1j%)jK3YXxrgD`{1BM^deYwDbSAElDs%P{z$3?|Yh(EkT2oFFGg?
zK$e7zg`veY^rN^<0>Ot-ghl$AV=M|I;Tp1*t@wD2VjT5N-lx-2P5>1UYb+?;N*qpp
z4mE%5)<dXymU;jFBxsZcs_<#OE6Ds-;eXfn`<tQCa?5h>-W>zyn2s{h+<XKM>|n4V
zbU|daCU|V8jHuDiDcSYOnw7f-qy|wa;@ie3l28KgBLnexxLBMhgeSq(A;iU5L}Q%9
zULNxOmzQAL2VD!{$O210*=Vq2v;XO5BvXCPu1^sqE+r+&@_zdqilm<r!FR^oX?HNG
zp`Ug#H?Ie`W|5>M;PEim4y>g=Vt0)i`boL@VC$gTz7_d50G!lIC->1<^=|TrtEfr-
zzHwvi^XDj!=XgWffW^M{DyAQMX#bb%Is~;fm~kmE${|!TohNVqCN0AQ2z8DxfMoW*
zAaw74r=|=Xk^WSPYKZ;k<<fsx@s8kXB@&cO6DASqvIZ1G$D^HhCE!VG)mweM0%n|E
zlXlJb4X3gmkYQl`=C(=jkm+>K@?S*F6^ORqEO!5jELcY%s1$-rgJtNrF2vR2bd<=*
zFfuW~9^M&2@eaRWlZ8%K5`qDAH+OE_IC^*xsUGkVnD7=sqX!7D5KzymTj}`g=KGgH
zIkQU*Q4@aE2f$a$*q&gB1OBCOf}m9)etH)TWCL4JJzYhD89?hd^9u?lu=${dq^qcm
z%>)B5*af+uYp6NOdKr{@JA&+?t|wYj_}>}&DcQj}qk*08-jqSugMun}5_AGvDy4TH
zecQ%db{~`odAYd+3P}(hTYR9W{oRz5iX+!Ug@R8N15`{vZX5b3+vVIm*16@*WcQ||
zS{^E#T?tC8_JYG^B~DPU|1N<x!(AT}MPkfE=>zat_(5OT?DgIrpV`>7d3m?3Ov#3t
zTc@cs!W5Nk5N~iji$EUUytnYd8^aydICgc~7s7?xA6{&?5FZDP4CGU_i`U`%AwQI^
z?pSnRkO}Mg?D2jSU6sWCCqaqs{5TjD4)Iy5tMe&zz=sgj&mlS=UbWxSa=cn!z(hbm
zmcmS$sIhEr+cp_0Kp{==BT2Q1=g$MdGT(zkOkTr1NKjy>{Dd8Yp~UwV7C<AuHP`L}
zK{0LCq+`U_G=UhQ(yEYHO=4HR^V#+Wt7*HZ7ReMpqU+PY_tEQA04`8_poSxmtoZ6q
zAa4?sNbKe~at--{sj|3tJlEcF|Cpr2!-b+<gaN2zioW&Cj<p5o7Dsxe-=m3Nr?3r*
z$MHrf?M{jp4*xg<gNwqg4QU%;3=2T~GLZhpXElMWx9i6daz9j*k~HGXFf@x|cI2lm
zED9{;;ByhPZ35w;kW#@NXW0V`9b+*JnEf5WMs@p#b^epsqP@nCUej4@)UUDfmj7%>
z55=NE>f93A4|_4GvKrSoT{l=(vh%U|cuO9`R6JPE*(wy&-al|fvtnimJ}_%HT76q3
ziqZ#6hh$k2r4MDj`^R84`-iINv<1bm4})={QgA!rNeh#SZ5zK2<M%Y?TZ1LKEmba8
z_ttDF#k&n}yD0(INqz)-Au@Ot59yka=ipb!&wQ#6?qe&oxTEJccfI`0i_6-11L@ee
zwvy7=f!;`vg(DANvnWU`w!p`}E^R^qYiDhJ3b2hb>?}}G=g?2!ux~g84g<ois4r1`
z4grmXrvcnp9?3@ooj?TV-MkqUn2O$A0V4U^_g)l%G?|82je-U=JPBGi?mi;hrimKt
zKZl>nMe*EAGv<en=-WeIOdzIjP|pKIRg{ZaK9L&o<_cl)`-VSC7##RUb55{st5qw$
zJ&+D&OLq!PrcmTo_Og3_l}`gLkU+MzJAVqR>=pky3}b((&>KwT<)Ni-45?)^eWzX!
zCMDN_Kao`CVK;^9XC0S=yt!&-5tdHk4N~^0w+8r`8x)4+1xfXiz2cVedp@5&Up2(b
z&dRF4r1dB4wgRzHVN@W1s!5Wz1}Z<l?f_*qeO0`W2saD~5)m;~ELfN~6zJ(Wf?U9&
z;Ax6aJgJ*2?N6)1y<8qJ2Cun_bRk6I=YsnmmGSR13D87Q>G9U%m%WNity{;+u{5i-
z++xNXA*HGu0i;NVp%O)m^=f4G;dq7DGT|67Z!AbR7VVD6<VDX&?Xx}wz8oXs%9Se^
zsu}6sN8ZmI)de-JrrTI_+HJGrnY^5wc5EX8v`?yC&r05ObRS-LM;9Mi9ELBeF;N2r
z<02Y}21+c#)PHLLw7{nB7WN#%7uAsuh_~XB*WkP{6RXujXKJaT$1pGnvFcf{b^%%%
z=8O-DY5!k+JHbm<Id-Fjz-HJ)4w*AmI(YL|j6BjEHV8OiI<x(m@y#W8b0fgZ5;;0a
zy3>(x)LhD#AN4CMMsud<d=)(vr@`Ldeh#6{s5C@n#Vj1VFJ8JNSG$A<4{mg8_To$g
zb@OBm+Qe0)>24547!jl@!d;4b*P8P2tp5DYo63rcn8v+e!?A@QA_OWJL9N$kn|)`g
zf?(_{BLd46e!U!qRMW^7j0mjIBHCkILPy5|?$wHl3MgC`(LyFO73AnpMN2a<;L2R9
z03m!Cz^Od6HzV+#8*0GhKtlq9CM6DxB5)DMXjPD`e?Ufm^wb_sYg>;oH*=rkQ>fM4
z9kdi=(%6{)WjGxH4Us2%R36al@l#KLs)+fy1EsIawY`b~#)X)?#0)a|Pg}o5odf$H
zYN!f#U`^bg=#6g)`w{LAI{e)DjTt-Z+BF{-a3?1=_KnKfG_}2a83N3%aWs=Z57BJs
z$9w4#q$ixdI`(eTcvzM}Jt>FZ@;8gZ-((EroNi|I>N|p+BsgMfw$=sv+9h*X2Ov-A
zCD~;mUS3{S(e=I)t^{lrt84X!&E&51&_^_2;RyUn$sYNqy?7wB*%7IfLx3qTBJeiM
z_(A#)iNr$=(`18RsWx{_n8jV3wyz+lkK=Ea*Mila@}OOb|2V{cz<Ar)+XoS7jrnQ)
zuF+IP4$gy7Fy|wi4t+IENZY6(LF4LcDVZ9Jg+V+<Epl{y{(uGWnJfyP`$}BXD55DC
zt_9a|DrdDcrhCi2RabwH!bFyVzH)xRBJ^0pBd>M2ZEZb)ymjmR<(Vxtx$t}7sc+XV
z0S}ur3I#cBn%TkCHU3GQq7@-3Y70V;;<R6C;S#|+L>8%?cH)vPHvfH}PccUD-vJ1^
zH2+-c(jI`M9cZk4?m_k#!he58bcKz@;#FIILa~QPEz9tLX#=28ThIjof`YzfwraVB
zq32GFC5awOB+>_%^t4ki?HCcnuMnK6$k9EO8EOxgNIL6!tvC}M9qk(MV6V!L4kr|3
zww^6XbX>-uN^)CUTT@4mr(K+!oP@1jEt~kO=9pOA3gYm_;EXsm$48A1z!oC{>nK6P
zu-ijt#p18=A{r0N#-*AIx-~GAM$V@@77;(Q&AB=FJLow|2C^?>M7{FCV_nQqaiU8^
z30B02*O98VDcN0!=)6DEwQq&&+&(&}e*B2W&3cA7w_ps|*hMrTIU9!;-ZYNzlyiOc
zkyE6^+5#t4wwdQK^wXT!8gI1&jk>NYgB=~a&}zTTb-Xnc=;f6=zRmb_-km#LxS7Pc
zd#I@1?J`_8T8?L5yEgw;SwDc}ZAtH4g*4E-s!R=cT1l4S#u1FY_za4hxj5l;99qEX
z`6qZ))PzazZwH|{pYM7fqD<N=E)rySReVuyn|CK>H1@pV3n1@rZue5ny_lBPon<DA
z8)(^yMr7sibqb?-lX>%jvGqnuim!+<sMO1LOib|~%G0WT-x+q`gId10x@OOZnFGzG
z@X}qG_yN;RC7lJ`{_Y6JoZ{O~r5(v4`e5<925qfGA+kh4Z`I8D7kDTMr4~RO5)2s;
zFonuVeT&EM6hugC#|3El<v%=m4?Ax#hdJMDhi|%-*;<M}((>zvCmCcW$lE`Kf;YCb
zu|m|v^~iKzUmxj51ko2RT=?<u;SEtI@GL#d*5Y(~=%uTdhz4H6*yNAoY0ure-S_vt
z-+)(V2VRw8q8k|@;=TK{Hq~ev{8;&|IwgV<7s&Zx9|2wL5B#YV_~kMcCy3zS;8E1A
zyL2AC_G~dj)ir2w0EYqY<xjOEe^9mtIkfRGw5en(6`K<yRLqcn{_$6J3<PHfZ&_PE
zc;$ht;T*$32CuVi8?h;ywf67zF<48PZp{m^wzJz{y04{R<ob|o*>t$&pb$j-dE`Yy
zz5AHOh)T+c)6m1U9L)Aqj6eSL!R?^lqhp_J7o`1{F71Kk=eL9AryCj?WEr<XxfgT@
z@GM?oauM4AL&x(~m+|f%Wq}CO2fPDk-&cpFUgpBu1kE_w_QzNTNi{a<NDP29z>OI~
z@j_w=fSGqnRCHyKtP!U_VD|G8*%;760g4r87niG)LqM_j(mU6ilJtZ~P=!f@YtkH=
z2lPfK@vkATy($fQL^*`1A}3*ho~`m6O?Do;FA8-K62lBqgmzRnDlIazj9uNb#=qyx
z%iZ_lN|iBp#UvnpWA~%mK^Pxn1v!}*Ph$Mtdp0J#KXzTHudi>*#SjwwK)i9ZvC&2y
z=c-(Y21a(p3ROJKFp1vm6o9?x>5EeYJLZo0-k^!>7?QIl^S<wxb=Z=z?>fUNYvPnA
zt`ajsQa6Rf=)-oK6N~wEc%1g%Kl*htO<L?(>K{Xf^Q$w~EXphXh-s%hJx=#yGUX81
zha`iC1x(|^yj!=T15+`}1PHQJQ2yOzPY_Bh#pKlqWPg*NAOns)Ipu129aIKCbo_z-
zuHmu?V0(4B(`VNN<_svB#87Fh-;|Cb_|Zqxh#N9T6ClRuK$ZD)!OdH_H-%MbsPE((
zy>HN)^GI)y<yzo?T3wu3|AHr~z~S>q@m_v5@Z)eH*cA3&g@?&{{g!^|(hRUv{6U+t
zW6e%5zCffr4(7dZCS8d3!{TgZQ<hnh0>9rRuk<4<1#2-3%{W(v`IFypI}`OL!A#(W
zpl8B63^W8Icw-4>!>T-|dgMCj<IDIX-eOWKFI*(QzC)RI+52$-S`@I%;@XNg)g}w`
zhGwm899y4{hATfZ*@I`!rYSmbd5;vZXq|XFf;JX~UutUVh;K158$r)`2&`vB@P@iN
zk9Q5?_?iZP$lGnCiX1$EpQO|2tHJU1SN0Wwj>P|8udc8X_~8G?d#@)#F{<U)2W-zp
R<dbVQHQHmC%Q*PUe*snNFwp=2

diff --git a/Perf-measure.md b/Perf-measure.md
index 6afd518b..dc2143c8 100644
--- a/Perf-measure.md
+++ b/Perf-measure.md
@@ -68,7 +68,7 @@ rm
 #>     }
 #>     .Internal(remove(list, envir, inherits))
 #> }
-#> <bytecode: 0x56294c35adf0>
+#> <bytecode: 0x56191ab34df0>
 #> <environment: namespace:base>
 ```
 
@@ -144,8 +144,8 @@ full_join(t_bench_df, t_systime_df, by = "expression")
 #> # A tibble: 2 × 4
 #>   expression bench_mean systime_with_gc systime_with_nogc
 #>   <bch:expr>   <bch:tm>           <dbl>             <dbl>
-#> 1 sqrt(x)      670.07ns           0.629             0.443
-#> 2 x^0.5          2.19µs           2.01              2.02
+#> 1 sqrt(x)       677.2ns           0.629             0.428
+#> 2 x^0.5          2.23µs           2.02              2.00
 ```
 
 The comparison reveals that these two approaches yield quite similar results. Slight differences in exact values is possibly due to differences in the precision of timers used internally by these functions.
@@ -180,10 +180,10 @@ bench::mark(
 #> # A tibble: 4 × 2
 #>   expression      median
 #>   <bch:expr>    <bch:tm>
-#> 1 sqrt(x)         3.03µs
-#> 2 exp(log(x)/2)  12.62µs
-#> 3 x^0.5          18.86µs
-#> 4 x^(1/2)           19µs
+#> 1 sqrt(x)         3.02µs
+#> 2 exp(log(x)/2)  12.66µs
+#> 3 x^0.5          18.96µs
+#> 4 x^(1/2)        19.01µs
 ```
 
 The specialized primitive function `sqrt()` (written in `C`) is the fastest way to compute square root.
@@ -205,7 +205,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Quotation.md b/Quotation.md
index 4f9eec32..f71682ea 100644
--- a/Quotation.md
+++ b/Quotation.md
@@ -171,7 +171,7 @@ rlang::expr
 #> {
 #>     enexpr(expr)
 #> }
-#> <bytecode: 0x55b385e4f448>
+#> <bytecode: 0x5584f9fd1728>
 #> <environment: namespace:rlang>
 ```
 
@@ -193,7 +193,7 @@ rlang::enexpr
 #> {
 #>     .Call(ffi_enexpr, substitute(arg), parent.frame())
 #> }
-#> <bytecode: 0x55b3829aaae0>
+#> <bytecode: 0x5584f6b2cae0>
 #> <environment: namespace:rlang>
 ```
 
@@ -824,7 +824,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/R6.md b/R6.md
index 988cee34..4c63dc0e 100644
--- a/R6.md
+++ b/R6.md
@@ -288,7 +288,7 @@ typeof(myExample)
 #> [1] "environment"
 
 rlang::env_print(myExample)
-#> <environment: 0x55ffdaaac290> [L]
+#> <environment: 0x5565e0c0a290> [L]
 #> Parent: <environment: empty>
 #> Class: Example, R6
 #> Bindings:
@@ -556,7 +556,7 @@ gc()
 #> [1] "Closing the file connection!"
 #>           used (Mb) gc trigger (Mb) max used (Mb)
 #> Ncells  746818 39.9    1365334   73  1365334 73.0
-#> Vcells 1372826 10.5    8388608   64  2594866 19.8
+#> Vcells 1372826 10.5    8388608   64  2594870 19.8
 ```
 
 ## Session information
@@ -574,7 +574,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       Europe/Berlin
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Rcpp.md b/Rcpp.md
index 73dfc463..7c75ddcc 100644
--- a/Rcpp.md
+++ b/Rcpp.md
@@ -216,12 +216,12 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression                                      min
 #>   <bch:expr>                                 <bch:tm>
-#> 1 all(c(rep(TRUE, 1000), rep(FALSE, 1000)))    6.16µs
-#> 2 allC(c(rep(TRUE, 1000), rep(FALSE, 1000)))   8.59µs
+#> 1 all(c(rep(TRUE, 1000), rep(FALSE, 1000)))    6.13µs
+#> 2 allC(c(rep(TRUE, 1000), rep(FALSE, 1000)))   8.73µs
 #>     median `itr/sec` mem_alloc `gc/sec`
 #>   <bch:tm>     <dbl> <bch:byt>    <dbl>
-#> 1   6.46µs   152676.    15.8KB        0
-#> 2   8.95µs   109377.    18.3KB        0
+#> 1   6.49µs   152211.    15.8KB        0
+#> 2   9.15µs   106881.    18.3KB        0
 ```
 
 - `cumprod()`
@@ -263,8 +263,8 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression        min   median `itr/sec` mem_alloc
 #>   <bch:expr>   <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 cumprod(v1)  110.01ns 130.04ns  6991133.        0B
-#> 2 cumprodC(v1)   1.44µs   1.53µs   619348.    6.62KB
+#> 1 cumprod(v1)   89.99ns 101.05ns  8156808.        0B
+#> 2 cumprodC(v1)   1.45µs   1.55µs   612672.    6.62KB
 #>   `gc/sec`
 #>      <dbl>
 #> 1        0
@@ -310,8 +310,8 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression       min   median `itr/sec` mem_alloc `gc/sec`
 #>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
-#> 1 cummin(v1)  130.97ns 151.57ns  5389797.        0B        0
-#> 2 cumminC(v1)   1.52µs   1.83µs   514890.    6.62KB        0
+#> 1 cummin(v1)  120.02ns 131.08ns  5672695.        0B        0
+#> 2 cumminC(v1)   1.56µs   1.87µs   507394.    6.62KB        0
 ```
 
 - `cummaxC()`
@@ -353,8 +353,8 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression       min   median `itr/sec` mem_alloc `gc/sec`
 #>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
-#> 1 cummax(v1)  110.01ns 120.14ns  6760800.        0B        0
-#> 2 cummaxC(v1)   1.55µs   1.72µs   549982.    6.62KB        0
+#> 1 cummax(v1)  110.01ns 121.07ns  6496158.        0B        0
+#> 2 cummaxC(v1)   1.58µs   1.76µs   545922.    6.62KB        0
 ```
 
 - `diff()`
@@ -417,8 +417,8 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression        min   median `itr/sec` mem_alloc
 #>   <bch:expr>   <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 diff(v1, 2)    3.78µs   4.44µs   218299.        0B
-#> 2 diffC(v1, 2)   1.91µs   2.12µs   450551.    2.49KB
+#> 1 diff(v1, 2)    3.88µs   4.46µs   217285.        0B
+#> 2 diffC(v1, 2)   1.89µs   2.13µs   452394.    2.49KB
 #>   `gc/sec`
 #>      <dbl>
 #> 1        0
@@ -464,8 +464,8 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression      min   median `itr/sec` mem_alloc `gc/sec`
 #>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
-#> 1 range(v1)    1.64µs   1.81µs   526003.        0B        0
-#> 2 rangeC(v1)   1.51µs   1.66µs   482154.    6.62KB        0
+#> 1 range(v1)    1.68µs   1.94µs   481017.        0B        0
+#> 2 rangeC(v1)   1.47µs   1.68µs   569929.    6.62KB        0
 ```
 
 - `var()`
@@ -513,8 +513,8 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression        min   median `itr/sec` mem_alloc
 #>   <bch:expr>   <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 var(v1)         7.4µs   8.31µs   117887.        0B
-#> 2 variance(v1)   1.46µs   1.61µs   587188.    6.62KB
+#> 1 var(v1)        7.54µs   8.51µs   111092.        0B
+#> 2 variance(v1)   1.49µs   1.66µs   567349.    6.62KB
 #>   `gc/sec`
 #>      <dbl>
 #> 1        0
@@ -695,8 +695,8 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression              min   median `itr/sec` mem_alloc
 #>   <bch:expr>         <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 median.default(v2)  21.91µs  23.52µs    40942.        0B
-#> 2 medianC(v2)          1.45µs   1.67µs   565586.    2.49KB
+#> 1 median.default(v2)  22.17µs  24.29µs    40480.        0B
+#> 2 medianC(v2)          1.49µs   1.73µs   546971.    2.49KB
 #>   `gc/sec`
 #>      <dbl>
 #> 1        0
@@ -746,8 +746,8 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression          min   median `itr/sec` mem_alloc
 #>   <bch:expr>     <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 x1 %in% x2     881.03ns   1.11µs   703925.        0B
-#> 2 matchC(x1, x2)   2.07µs   2.35µs   409169.    6.62KB
+#> 1 x1 %in% x2     881.03ns   1.11µs   870819.        0B
+#> 2 matchC(x1, x2)   2.15µs   2.41µs   399261.    6.62KB
 #>   `gc/sec`
 #>      <dbl>
 #> 1        0
@@ -822,8 +822,8 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression        min   median `itr/sec` mem_alloc
 #>   <bch:expr>   <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 unique(v1)     2.13µs   2.43µs   343230.        0B
-#> 2 uniqueC2(v1)   1.69µs   1.85µs   512921.    6.62KB
+#> 1 unique(v1)      2.1µs   2.41µs   355209.        0B
+#> 2 uniqueC2(v1)   1.63µs   1.85µs   502571.    6.62KB
 #>   `gc/sec`
 #>      <dbl>
 #> 1        0
@@ -871,8 +871,8 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression      min   median `itr/sec` mem_alloc `gc/sec`
 #>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
-#> 1 min(v1)    240.05ns 251.11ns  3549516.        0B        0
-#> 2 minC(v1)     1.39µs   1.51µs   612452.    6.62KB        0
+#> 1 min(v1)    230.04ns 250.93ns  3402682.        0B        0
+#> 2 minC(v1)     1.42µs   1.56µs   606873.    6.62KB        0
 
 max(v1)
 #> [1] 9
@@ -888,8 +888,8 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression      min   median `itr/sec` mem_alloc `gc/sec`
 #>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
-#> 1 max(v1)    220.03ns  250.1ns  3345068.        0B        0
-#> 2 maxC(v1)     1.39µs    1.6µs   595400.    6.62KB        0
+#> 1 max(v1)    211.06ns  241.1ns  3455199.        0B        0
+#> 2 maxC(v1)     1.41µs   1.66µs   571011.    6.62KB        0
 ```
 
 1. `which.min()` using `min_element`, or `which.max()` using `max_element`.
@@ -939,8 +939,8 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression          min   median `itr/sec` mem_alloc
 #>   <bch:expr>     <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 which.min(v1)   401.1ns 431.09ns  2009835.        0B
-#> 2 which_minC(v1)    1.4µs   1.62µs   589790.    6.62KB
+#> 1 which.min(v1)  401.05ns  421.1ns  2181122.        0B
+#> 2 which_minC(v1)   1.44µs    1.6µs   591193.    6.62KB
 #>   `gc/sec`
 #>      <dbl>
 #> 1        0
@@ -960,8 +960,8 @@ bench::mark(
 #> # A tibble: 2 × 6
 #>   expression          min   median `itr/sec` mem_alloc
 #>   <bch:expr>     <bch:tm> <bch:tm>     <dbl> <bch:byt>
-#> 1 which.max(v1)  390.11ns 461.12ns  1940961.        0B
-#> 2 which_maxC(v1)   1.39µs   1.55µs   608448.    6.62KB
+#> 1 which.max(v1)  410.01ns 490.92ns  1870056.        0B
+#> 2 which_maxC(v1)   1.41µs   1.57µs   585441.    6.62KB
 #>   `gc/sec`
 #>      <dbl>
 #> 1        0
@@ -1101,7 +1101,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/S3.md b/S3.md
index 50741d15..80ccb997 100644
--- a/S3.md
+++ b/S3.md
@@ -579,7 +579,7 @@ t(x)
 t
 #> function (x) 
 #> UseMethod("t")
-#> <bytecode: 0x56246846b798>
+#> <bytecode: 0x56119bea5208>
 #> <environment: namespace:base>
 sloop::is_s3_generic("t")
 #> [1] TRUE
@@ -587,7 +587,7 @@ sloop::is_s3_generic("t")
 t.test
 #> function (x, ...) 
 #> UseMethod("t.test")
-#> <bytecode: 0x562463507140>
+#> <bytecode: 0x561196f41140>
 #> <environment: namespace:stats>
 sloop::is_s3_generic("t.test")
 #> [1] TRUE
@@ -1006,7 +1006,7 @@ sloop::s3_get_method("[.Date")
 #> {
 #>     .Date(NextMethod("["), oldClass(x))
 #> }
-#> <bytecode: 0x5624687fc3c0>
+#> <bytecode: 0x56119c235db0>
 #> <environment: namespace:base>
 ```
 
@@ -1017,7 +1017,7 @@ The `.Date` function looks like this:
 .Date
 #> function (xx, cl = "Date") 
 #> `class<-`(xx, cl)
-#> <bytecode: 0x562466349c10>
+#> <bytecode: 0x561199d84c50>
 #> <environment: namespace:base>
 ```
 
@@ -1040,7 +1040,7 @@ attributes(x[1])
 #> [1] "subDate" "Date"
 
 x[1]
-#> [1] "2023-12-31"
+#> [1] "2024-01-07"
 ```
 
 
@@ -1409,7 +1409,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/S4.md b/S4.md
index baaa03f1..2320d4c3 100644
--- a/S4.md
+++ b/S4.md
@@ -625,7 +625,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Subsetting.md b/Subsetting.md
index c630c71f..541d4411 100644
--- a/Subsetting.md
+++ b/Subsetting.md
@@ -418,7 +418,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Translation.md b/Translation.md
index 0adf8abb..cb728a3f 100644
--- a/Translation.md
+++ b/Translation.md
@@ -322,7 +322,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/Vectors.md b/Vectors.md
index d4bb2003..490f65a3 100644
--- a/Vectors.md
+++ b/Vectors.md
@@ -210,7 +210,7 @@ setNames
 #>     names(object) <- nm
 #>     object
 #> }
-#> <bytecode: 0x5635a7f8a880>
+#> <bytecode: 0x55c0caac1880>
 #> <environment: namespace:stats>
 ```
 
@@ -240,7 +240,7 @@ unname
 #>         dimnames(obj) <- NULL
 #>     obj
 #> }
-#> <bytecode: 0x5635a7395578>
+#> <bytecode: 0x55c0c9ecc578>
 #> <environment: namespace:base>
 ```
 
@@ -841,7 +841,7 @@ sessioninfo::session_info(include_base = TRUE)
 #>  collate  C.UTF-8
 #>  ctype    C.UTF-8
 #>  tz       UTC
-#>  date     2023-12-31
+#>  date     2024-01-07
 #>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)
 #> 
 #> ─ Packages ───────────────────────────────────────────────
diff --git a/base-types.html b/base-types.html
index adcd3178..3c1ab03b 100644
--- a/base-types.html
+++ b/base-types.html
@@ -111,7 +111,7 @@ <h1>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/big-picture.html b/big-picture.html
index 125a2583..c4a9fe95 100644
--- a/big-picture.html
+++ b/big-picture.html
@@ -111,7 +111,7 @@ <h1>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/conditions.html b/conditions.html
index 5a8b04bc..be8b39b6 100644
--- a/conditions.html
+++ b/conditions.html
@@ -314,7 +314,7 @@ <h2>
 <span><span class="co">#&gt;         return(NULL)</span></span>
 <span><span class="co">#&gt;     })))</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x564fa21357a8&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x55f2c19fa068&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:rlang&gt;</span></span></code></pre></div>
 <p>As mentioned in the function docs:</p>
 <blockquote>
@@ -633,7 +633,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -726,7 +726,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/control-flow.html b/control-flow.html
index 933ab094..43eea36b 100644
--- a/control-flow.html
+++ b/control-flow.html
@@ -277,7 +277,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -352,7 +352,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/debugging.html b/debugging.html
index 874babb8..b1a1d483 100644
--- a/debugging.html
+++ b/debugging.html
@@ -111,7 +111,7 @@ <h1>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/environments.html b/environments.html
index a7ea6801..ad81f7df 100644
--- a/environments.html
+++ b/environments.html
@@ -143,15 +143,15 @@ <h2>
 <span><span class="va">e</span> <span class="op">&lt;-</span> <span class="fu"><a href="https://rlang.r-lib.org/reference/env.html">env</a></span><span class="op">(</span><span class="op">)</span></span>
 <span><span class="va">e</span><span class="op">$</span><span class="va">loop</span> <span class="op">&lt;-</span> <span class="va">e</span></span>
 <span><span class="fu"><a href="https://rlang.r-lib.org/reference/env_print.html">env_print</a></span><span class="op">(</span><span class="va">e</span><span class="op">)</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x558a72b692e8&gt;</span></span>
+<span><span class="co">#&gt; &lt;environment: 0x5592b255c2e8&gt;</span></span>
 <span><span class="co">#&gt; Parent: &lt;environment: global&gt;</span></span>
 <span><span class="co">#&gt; Bindings:</span></span>
 <span><span class="co">#&gt; • loop: &lt;env&gt;</span></span></code></pre></div>
 <p>The binding <code>loop</code> should have the same memory address as the environment <code>e</code>:</p>
 <div class="sourceCode" id="cb193"><pre class="downlit sourceCode r">
 <code class="sourceCode R"><span><span class="fu">lobstr</span><span class="fu">::</span><span class="fu"><a href="https://lobstr.r-lib.org/reference/ref.html">ref</a></span><span class="op">(</span><span class="va">e</span><span class="op">$</span><span class="va">loop</span><span class="op">)</span></span>
-<span><span class="co">#&gt; █ [1:0x558a72b692e8] &lt;env&gt; </span></span>
-<span><span class="co">#&gt; └─loop = [1:0x558a72b692e8]</span></span></code></pre></div>
+<span><span class="co">#&gt; █ [1:0x5592b255c2e8] &lt;env&gt; </span></span>
+<span><span class="co">#&gt; └─loop = [1:0x5592b255c2e8]</span></span></code></pre></div>
 <p><strong>Q3.</strong> Create a pair of environments as illustrated by this picture.</p>
 <div class="inline-figure"><img src="diagrams/environments/recursive-2.png" width="307"></div>
 <p><strong>A3.</strong> Creating the specified environment:</p>
@@ -164,9 +164,9 @@ <h2>
 <span></span>
 <span><span class="co"># following should be the same</span></span>
 <span><span class="fu">lobstr</span><span class="fu">::</span><span class="fu"><a href="https://lobstr.r-lib.org/reference/obj_addr.html">obj_addrs</a></span><span class="op">(</span><span class="fu"><a href="https://rdrr.io/r/base/list.html">list</a></span><span class="op">(</span><span class="va">e1</span>, <span class="va">e2</span><span class="op">$</span><span class="va">deloop</span><span class="op">)</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "0x558a751c8be8" "0x558a751c8be8"</span></span>
+<span><span class="co">#&gt; [1] "0x5592b4bbb1e8" "0x5592b4bbb1e8"</span></span>
 <span><span class="fu">lobstr</span><span class="fu">::</span><span class="fu"><a href="https://lobstr.r-lib.org/reference/obj_addr.html">obj_addrs</a></span><span class="op">(</span><span class="fu"><a href="https://rdrr.io/r/base/list.html">list</a></span><span class="op">(</span><span class="va">e2</span>, <span class="va">e1</span><span class="op">$</span><span class="va">loop</span><span class="op">)</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "0x558a75216348" "0x558a75216348"</span></span></code></pre></div>
+<span><span class="co">#&gt; [1] "0x5592b4c08948" "0x5592b4c08948"</span></span></code></pre></div>
 <p><strong>Q4.</strong> Explain why <code>e[[1]]</code> and <code>e[c("a", "b")]</code> don’t make sense when <code>e</code> is an environment.</p>
 <p><strong>A4.</strong> An environment is a non-linear data structure, and has no concept of ordered elements. Therefore, indexing it (e.g. <code>e[[1]]</code>) doesn’t make sense.</p>
 <p>Subsetting a list or a vector returns a subset of the underlying data structure. For example, subsetting a vector returns another vector. But it’s unclear what subsetting an environment (e.g. <code>e[c("a", "b")]</code>) should return because there is no data structure to contain its returns. It can’t be another environment since environments have reference semantics.</p>
@@ -349,7 +349,7 @@ <h2>
 <span><span class="fu">fget</span><span class="op">(</span><span class="st">"mean"</span>, inherits <span class="op">=</span> <span class="cn">TRUE</span><span class="op">)</span></span>
 <span><span class="co">#&gt; function (x, ...) </span></span>
 <span><span class="co">#&gt; UseMethod("mean")</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x558a7059e5a8&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x5592aff915a8&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:base&gt;</span></span>
 <span></span>
 <span><span class="va">mean</span> <span class="op">&lt;-</span> <span class="fl">5</span></span>
@@ -516,7 +516,7 @@ <h2>
 <span><span class="co">#&gt; {</span></span>
 <span><span class="co">#&gt;     parent.frame(n + 1)</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x558a710bd8d0&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x5592b0ab08d0&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:rlang&gt;</span></span></code></pre></div>
 <p>Let’s try it out:</p>
 <ul>
@@ -564,7 +564,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -655,7 +655,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/evaluation.html b/evaluation.html
index d8f93574..4f2c3fcd 100644
--- a/evaluation.html
+++ b/evaluation.html
@@ -122,7 +122,7 @@ <h2>
 <span>    <span class="fu">foo</span><span class="op">(</span><span class="op">)</span></span>
 <span>  <span class="op">}</span></span>
 <span><span class="op">)</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x56269fbe7e18&gt;</span></span>
+<span><span class="co">#&gt; &lt;environment: 0x564fa1bed6a8&gt;</span></span>
 <span><span class="co">#&gt; Parent: &lt;environment: global&gt;</span></span></code></pre></div>
 <p>To specify a custom environment, the <code><a href="https://rdrr.io/r/base/sys.source.html">sys.source()</a></code> function can be used, which provides an <code>envir</code> parameter.</p>
 <hr>
@@ -275,17 +275,17 @@ <h2>
 <span><span class="va">q1</span></span>
 <span><span class="co">#&gt; &lt;quosure&gt;</span></span>
 <span><span class="co">#&gt; expr: ^x</span></span>
-<span><span class="co">#&gt; env:  0x56269fa39110</span></span>
+<span><span class="co">#&gt; env:  0x564fa1a3e9a0</span></span>
 <span><span class="va">q2</span> <span class="op">&lt;-</span> <span class="fu"><a href="https://rlang.r-lib.org/reference/new_quosure.html">new_quosure</a></span><span class="op">(</span><span class="fu"><a href="https://rlang.r-lib.org/reference/expr.html">expr</a></span><span class="op">(</span><span class="va">x</span> <span class="op">+</span> <span class="op">!</span><span class="op">!</span><span class="va">q1</span><span class="op">)</span>, <span class="fu"><a href="https://rlang.r-lib.org/reference/env.html">env</a></span><span class="op">(</span>x <span class="op">=</span> <span class="fl">10</span><span class="op">)</span><span class="op">)</span></span>
 <span><span class="va">q2</span></span>
 <span><span class="co">#&gt; &lt;quosure&gt;</span></span>
 <span><span class="co">#&gt; expr: ^x + (^x)</span></span>
-<span><span class="co">#&gt; env:  0x5626a008bd48</span></span>
+<span><span class="co">#&gt; env:  0x564fa20915d8</span></span>
 <span><span class="va">q3</span> <span class="op">&lt;-</span> <span class="fu"><a href="https://rlang.r-lib.org/reference/new_quosure.html">new_quosure</a></span><span class="op">(</span><span class="fu"><a href="https://rlang.r-lib.org/reference/expr.html">expr</a></span><span class="op">(</span><span class="va">x</span> <span class="op">+</span> <span class="op">!</span><span class="op">!</span><span class="va">q2</span><span class="op">)</span>, <span class="fu"><a href="https://rlang.r-lib.org/reference/env.html">env</a></span><span class="op">(</span>x <span class="op">=</span> <span class="fl">100</span><span class="op">)</span><span class="op">)</span></span>
 <span><span class="va">q3</span></span>
 <span><span class="co">#&gt; &lt;quosure&gt;</span></span>
 <span><span class="co">#&gt; expr: ^x + (^x + (^x))</span></span>
-<span><span class="co">#&gt; env:  0x5626a034a820</span></span></code></pre></div>
+<span><span class="co">#&gt; env:  0x564fa23500b0</span></span></code></pre></div>
 <p><strong>A1.</strong> Correctly predicted 😉</p>
 <div class="sourceCode" id="cb598"><pre class="downlit sourceCode r">
 <code class="sourceCode R"><span><span class="va">q1</span> <span class="op">&lt;-</span> <span class="fu"><a href="https://rlang.r-lib.org/reference/new_quosure.html">new_quosure</a></span><span class="op">(</span><span class="fu"><a href="https://rlang.r-lib.org/reference/expr.html">expr</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span>, <span class="fu"><a href="https://rlang.r-lib.org/reference/env.html">env</a></span><span class="op">(</span>x <span class="op">=</span> <span class="fl">1</span><span class="op">)</span><span class="op">)</span></span>
@@ -313,7 +313,7 @@ <h2>
 <span></span>
 <span><span class="va">foo</span> <span class="op">&lt;-</span> <span class="kw">function</span><span class="op">(</span><span class="va">x</span><span class="op">)</span> <span class="fu">enenv</span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
 <span><span class="fu">foo</span><span class="op">(</span><span class="op">)</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x5626a12b70c8&gt;</span></span></code></pre></div>
+<span><span class="co">#&gt; &lt;environment: 0x564fa32bc708&gt;</span></span></code></pre></div>
 <hr>
 </div>
 <div id="data-masks-exercises-20.4.6" class="section level2" number="20.3">
@@ -608,7 +608,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -692,7 +692,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/expressions.html b/expressions.html
index 79414d46..6c1437c7 100644
--- a/expressions.html
+++ b/expressions.html
@@ -359,7 +359,7 @@ <h2>
 <code class="sourceCode R"><span><span class="va">mean</span></span>
 <span><span class="co">#&gt; function (x, ...) </span></span>
 <span><span class="co">#&gt; UseMethod("mean")</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x55d34a3275a8&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x562fc13895a8&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:base&gt;</span></span></code></pre></div>
 <p>As mentioned in the respective <a href="If%20the%20function%20uses%20...%20it%E2%80%99s%20not%20possible%20to%20standardise%20all%20arguments.">section</a>:</p>
 <blockquote>
@@ -833,7 +833,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -918,7 +918,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/function-factories.html b/function-factories.html
index 39e45d78..b397185b 100644
--- a/function-factories.html
+++ b/function-factories.html
@@ -100,7 +100,7 @@ <h2>
 <code class="sourceCode R"><span><span class="va">force</span></span>
 <span><span class="co">#&gt; function (x) </span></span>
 <span><span class="co">#&gt; x</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x556cfd712ab0&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x56479c5e3ab0&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:base&gt;</span></span></code></pre></div>
 <p>Why is it better to <code>force(x)</code> instead of just <code>x</code>?</p>
 <p><strong>A1.</strong> Due to lazy evaluation, argument to a function won’t be evaluated until its value is needed. But sometimes we may want to have eager evaluation, and using <code><a href="https://rdrr.io/r/base/force.html">force()</a></code> makes this intent clearer.</p>
@@ -118,8 +118,8 @@ <h2>
 <span><span class="va">f</span></span>
 <span><span class="co">#&gt; function (v) </span></span>
 <span><span class="co">#&gt; .approxfun(x, y, v, method, yleft, yright, f, na.rm)</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x556d01983cb0&gt;</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x556d01982ba0&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x5647a0854cb0&gt;</span></span>
+<span><span class="co">#&gt; &lt;environment: 0x5647a0853ba0&gt;</span></span>
 <span><span class="fu">f</span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
 <span><span class="co">#&gt;  [1] -0.7786629 -0.3894764 -2.0337983 -0.9823731  0.2478901</span></span>
 <span><span class="co">#&gt;  [6] -2.1038646 -0.3814180  2.0749198  1.0271384  0.4730142</span></span>
@@ -235,15 +235,15 @@ <h2>
 <span><span class="co">#&gt;     i &lt;- i + 1</span></span>
 <span><span class="co">#&gt;     i</span></span>
 <span><span class="co">#&gt;   }</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x556d01b28e68&gt;</span></span>
+<span><span class="co">#&gt; &lt;environment: 0x5647a09f9e68&gt;</span></span>
 <span></span>
 <span><span class="fu">new_counter3</span><span class="op">(</span><span class="op">)</span></span>
 <span><span class="co">#&gt; function() {</span></span>
 <span><span class="co">#&gt;     i &lt;- i + 1</span></span>
 <span><span class="co">#&gt;     i</span></span>
 <span><span class="co">#&gt;   }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x556d01ced600&gt;</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x556d01ba0c80&gt;</span></span></code></pre></div>
+<span><span class="co">#&gt; &lt;bytecode: 0x5647a0bbe600&gt;</span></span>
+<span><span class="co">#&gt; &lt;environment: 0x5647a0a71c80&gt;</span></span></code></pre></div>
 <hr>
 </div>
 <div id="graphical-factories-exercises-10.3.4" class="section level2" number="10.2">
@@ -278,7 +278,7 @@ <h2>
 <span><span class="co">#&gt;     }</span></span>
 <span><span class="co">#&gt;     structure(fun, class = "labeller")</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x556d01e7f0b0&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x5647a0d500b0&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:ggplot2&gt;</span></span></code></pre></div>
 <ul>
 <li><code><a href="https://scales.r-lib.org/reference/comma.html">scales::number_format()</a></code></li>
@@ -299,7 +299,7 @@ <h2>
 <span><span class="co">#&gt;             scale_cut = scale_cut, trim = trim, ...)</span></span>
 <span><span class="co">#&gt;     }</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x556d020cfbc8&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x5647a0fa0bc8&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:scales&gt;</span></span></code></pre></div>
 <p>Both of these functions return formatting functions used to style the facets labels and other labels to have the desired format in <a href="https://ggplot2.tidyverse.org">ggplot2</a> plots.</p>
 <p>For example, using plotmath expression in the facet label:</p>
@@ -365,7 +365,7 @@ <h2>
 <div class="sourceCode" id="cb322"><pre class="downlit sourceCode r">
 <code class="sourceCode R"><span><span class="fu">boxcox2</span><span class="op">(</span><span class="fl">1</span><span class="op">)</span></span>
 <span><span class="co">#&gt; function(x) (x^lambda - 1) / lambda</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x556cfdac76a0&gt;</span></span>
+<span><span class="co">#&gt; &lt;environment: 0x56479c998358&gt;</span></span>
 <span></span>
 <span><span class="fu">boxcox3</span><span class="op">(</span><span class="va">mtcars</span><span class="op">$</span><span class="va">wt</span><span class="op">)</span></span>
 <span><span class="co">#&gt; function(lambda) {</span></span>
@@ -375,7 +375,7 @@ <h2>
 <span><span class="co">#&gt;       (x^lambda - 1) / lambda</span></span>
 <span><span class="co">#&gt;     }</span></span>
 <span><span class="co">#&gt;   }</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x556cfefae4d0&gt;</span></span></code></pre></div>
+<span><span class="co">#&gt; &lt;environment: 0x56479de7f188&gt;</span></span></code></pre></div>
 <p>As can be seen:</p>
 <ul>
 <li>in <code>boxcox2()</code>, we can vary <code>x</code> for the same value of <code>lambda</code>, while</li>
@@ -401,12 +401,12 @@ <h2>
 <span><span class="co">#&gt;     col &lt;- df[[var]]</span></span>
 <span><span class="co">#&gt;     col[sample(n, replace = TRUE)]</span></span>
 <span><span class="co">#&gt;   }</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x556d00bef850&gt;</span></span></code></pre></div>
+<span><span class="co">#&gt; &lt;environment: 0x56479fac0b28&gt;</span></span></code></pre></div>
 <p>This is why we don’t need to worry about a copy being made because the <code>df</code> in the function environment points to the memory address of the data frame. We can confirm this by comparing their memory addresses:</p>
 <div class="sourceCode" id="cb324"><pre class="downlit sourceCode r">
 <code class="sourceCode R"><span><span class="va">boot_permute_env</span> <span class="op">&lt;-</span> <span class="fu">rlang</span><span class="fu">::</span><span class="fu"><a href="https://rlang.r-lib.org/reference/fn_env.html">fn_env</a></span><span class="op">(</span><span class="fu">boot_permute</span><span class="op">(</span><span class="va">mtcars</span>, <span class="st">"mpg"</span><span class="op">)</span><span class="op">)</span></span>
 <span><span class="fu">rlang</span><span class="fu">::</span><span class="fu"><a href="https://rlang.r-lib.org/reference/env_print.html">env_print</a></span><span class="op">(</span><span class="va">boot_permute_env</span><span class="op">)</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x556d0181e450&gt;</span></span>
+<span><span class="co">#&gt; &lt;environment: 0x5647a06ef728&gt;</span></span>
 <span><span class="co">#&gt; Parent: &lt;environment: global&gt;</span></span>
 <span><span class="co">#&gt; Bindings:</span></span>
 <span><span class="co">#&gt; • n: &lt;int&gt;</span></span>
@@ -455,8 +455,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression      min   median `itr/sec` mem_alloc `gc/sec`</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt; &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;    &lt;dbl&gt;</span></span>
-<span><span class="co">#&gt; 1 LL1          27.5µs   29.3µs    33174.    12.9KB     43.2</span></span>
-<span><span class="co">#&gt; 2 LL2          14.6µs   15.3µs    62143.        0B     49.8</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 LL1          28.3µs   30.2µs    31979.    12.9KB     41.6</span></span>
+<span><span class="co">#&gt; 2 LL2          14.6µs   15.7µs    60147.        0B     48.2</span></span></code></pre></div>
 <p>As can be seen, the second version is much faster than the first version.</p>
 <p>We can also vary the length of the vector and confirm that across a wide range of vector lengths, this performance advantage is observed.</p>
 <div class="sourceCode" id="cb327"><pre class="downlit sourceCode r">
@@ -477,16 +477,16 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 10 × 5</span></span>
 <span><span class="co">#&gt;    length expression      min   median `itr/sec`</span></span>
 <span><span class="co">#&gt;     &lt;dbl&gt; &lt;bch:expr&gt; &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt;</span></span>
-<span><span class="co">#&gt;  1     10 LL1          38.9µs   40.9µs    23209.</span></span>
-<span><span class="co">#&gt;  2     10 LL2          17.2µs     18µs    54710.</span></span>
-<span><span class="co">#&gt;  3     20 LL1          42.1µs     44µs    22427.</span></span>
-<span><span class="co">#&gt;  4     20 LL2          16.9µs   17.7µs    55356.</span></span>
-<span><span class="co">#&gt;  5     50 LL1          46.2µs   47.9µs    20558.</span></span>
-<span><span class="co">#&gt;  6     50 LL2          16.2µs   17.2µs    57361.</span></span>
-<span><span class="co">#&gt;  7    100 LL1          61.9µs   63.6µs    15410.</span></span>
-<span><span class="co">#&gt;  8    100 LL2          17.6µs   18.4µs    53603.</span></span>
-<span><span class="co">#&gt;  9   1000 LL1         854.3µs  872.3µs     1092.</span></span>
-<span><span class="co">#&gt; 10   1000 LL2          54.6µs   55.9µs    17653.</span></span>
+<span><span class="co">#&gt;  1     10 LL1          39.9µs   42.1µs    22543.</span></span>
+<span><span class="co">#&gt;  2     10 LL2          17.3µs   18.3µs    53968.</span></span>
+<span><span class="co">#&gt;  3     20 LL1            43µs   44.6µs    22159.</span></span>
+<span><span class="co">#&gt;  4     20 LL2          17.1µs     18µs    54815.</span></span>
+<span><span class="co">#&gt;  5     50 LL1          46.8µs   48.3µs    20374.</span></span>
+<span><span class="co">#&gt;  6     50 LL2          16.3µs   17.4µs    56753.</span></span>
+<span><span class="co">#&gt;  7    100 LL1          62.4µs   64.1µs    15369.</span></span>
+<span><span class="co">#&gt;  8    100 LL2          17.6µs   18.4µs    53652.</span></span>
+<span><span class="co">#&gt;  9   1000 LL1         857.4µs    965µs     1081.</span></span>
+<span><span class="co">#&gt; 10   1000 LL2          54.6µs   56.2µs    17530.</span></span>
 <span></span>
 <span><span class="fu"><a href="https://ggplot2.tidyverse.org/reference/ggplot.html">ggplot</a></span><span class="op">(</span></span>
 <span>  <span class="va">df_bench</span>,</span>
@@ -598,7 +598,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -698,7 +698,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/function-operators.html b/function-operators.html
index b8a77da7..de65ce11 100644
--- a/function-operators.html
+++ b/function-operators.html
@@ -159,7 +159,7 @@ <h2>
 <span><span class="co">#&gt;         })</span></span>
 <span><span class="co">#&gt;     }</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x55d90290c0b8&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x55d5181c6328&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:purrr&gt;</span></span></code></pre></div>
 <p>Looking at this code, we can see that <code><a href="https://purrr.tidyverse.org/reference/possibly.html">possibly()</a></code>:</p>
 <ul>
@@ -179,7 +179,7 @@ <h2>
 <span><span class="co">#&gt;     check_bool(quiet)</span></span>
 <span><span class="co">#&gt;     function(...) capture_error(.f(...), otherwise, quiet)</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x55d902aa9948&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x55d518363bb8&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:purrr&gt;</span></span>
 <span></span>
 <span><span class="fu">purrr</span><span class="fu">:::</span><span class="va">capture_error</span></span>
@@ -191,7 +191,7 @@ <h2>
 <span><span class="co">#&gt;         list(result = otherwise, error = e)</span></span>
 <span><span class="co">#&gt;     })</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x55d902b04180&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x55d5183be3f0&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:purrr&gt;</span></span></code></pre></div>
 <p>Looking at this code, we can see that <code><a href="https://purrr.tidyverse.org/reference/safely.html">safely()</a></code>:</p>
 <ul>
@@ -316,10 +316,10 @@ <h2>
 <span></span>
 <span>  <span class="fu"><a href="https://rdrr.io/r/base/cat.html">cat</a></span><span class="op">(</span><span class="fu"><a href="https://rdrr.io/r/base/readLines.html">readLines</a></span><span class="op">(</span><span class="va">logfile</span><span class="op">)</span>, sep <span class="op">=</span> <span class="st">"\n"</span><span class="op">)</span></span>
 <span><span class="op">}</span><span class="op">)</span></span>
-<span><span class="co">#&gt; Function created at: 2023-12-31 00:43:08.117083</span></span>
-<span><span class="co">#&gt; Function called at:  2023-12-31 00:43:13.123988</span></span>
-<span><span class="co">#&gt; Function called at:  2023-12-31 00:43:18.129344</span></span>
-<span><span class="co">#&gt; Function called at:  2023-12-31 00:43:26.137767</span></span></code></pre></div>
+<span><span class="co">#&gt; Function created at: 2024-01-07 00:44:00.426587</span></span>
+<span><span class="co">#&gt; Function called at:  2024-01-07 00:44:05.433612</span></span>
+<span><span class="co">#&gt; Function called at:  2024-01-07 00:44:10.439013</span></span>
+<span><span class="co">#&gt; Function called at:  2024-01-07 00:44:18.447574</span></span></code></pre></div>
 <hr>
 <p><strong>Q5.</strong> Modify <code>delay_by()</code> so that instead of delaying by a fixed amount of time, it ensures that a certain amount of time has elapsed since the function was last called. That is, if you called <code>g &lt;- delay_by(1, f); g(); Sys.sleep(2); g()</code> there shouldn’t be an extra delay.</p>
 <p><strong>A5.</strong> Modified version of the function meeting the specified requirements:</p>
@@ -361,7 +361,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -447,7 +447,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/functionals.html b/functionals.html
index bb424d76..8854748f 100644
--- a/functionals.html
+++ b/functionals.html
@@ -115,7 +115,7 @@ <h2>
 <span><span class="fu"><a href="https://purrr.tidyverse.org/reference/as_mapper.html">as_mapper</a></span><span class="op">(</span><span class="fl">1</span><span class="op">)</span></span>
 <span><span class="co">#&gt; function (x, ...) </span></span>
 <span><span class="co">#&gt; pluck_raw(x, list(1), .default = NULL)</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x555ecae48430&gt;</span></span>
+<span><span class="co">#&gt; &lt;environment: 0x55d958cd1430&gt;</span></span>
 <span></span>
 <span><span class="fu"><a href="https://purrr.tidyverse.org/reference/map.html">map</a></span><span class="op">(</span><span class="va">x</span>, <span class="fu"><a href="https://rdrr.io/r/base/list.html">list</a></span><span class="op">(</span><span class="fl">2</span>, <span class="fl">1</span><span class="op">)</span><span class="op">)</span></span>
 <span><span class="co">#&gt; [[1]]</span></span>
@@ -126,7 +126,7 @@ <h2>
 <span><span class="fu"><a href="https://purrr.tidyverse.org/reference/as_mapper.html">as_mapper</a></span><span class="op">(</span><span class="fu"><a href="https://rdrr.io/r/base/list.html">list</a></span><span class="op">(</span><span class="fl">2</span>, <span class="fl">1</span><span class="op">)</span><span class="op">)</span></span>
 <span><span class="co">#&gt; function (x, ...) </span></span>
 <span><span class="co">#&gt; pluck_raw(x, list(2, 1), .default = NULL)</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x555ecaf2f158&gt;</span></span>
+<span><span class="co">#&gt; &lt;environment: 0x55d958db8158&gt;</span></span>
 <span></span>
 <span><span class="co"># mapping by name -----------------------</span></span>
 <span></span>
@@ -144,7 +144,7 @@ <h2>
 <span><span class="fu"><a href="https://purrr.tidyverse.org/reference/as_mapper.html">as_mapper</a></span><span class="op">(</span><span class="st">"m"</span><span class="op">)</span></span>
 <span><span class="co">#&gt; function (x, ...) </span></span>
 <span><span class="co">#&gt; pluck_raw(x, list("m"), .default = NULL)</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x555ecb07d330&gt;</span></span>
+<span><span class="co">#&gt; &lt;environment: 0x55d958f06330&gt;</span></span>
 <span></span>
 <span><span class="co"># mixing position and name</span></span>
 <span><span class="fu"><a href="https://purrr.tidyverse.org/reference/map.html">map</a></span><span class="op">(</span><span class="va">y</span>, <span class="fu"><a href="https://rdrr.io/r/base/list.html">list</a></span><span class="op">(</span><span class="fl">2</span>, <span class="st">"m"</span><span class="op">)</span><span class="op">)</span></span>
@@ -156,7 +156,7 @@ <h2>
 <span><span class="fu"><a href="https://purrr.tidyverse.org/reference/as_mapper.html">as_mapper</a></span><span class="op">(</span><span class="fu"><a href="https://rdrr.io/r/base/list.html">list</a></span><span class="op">(</span><span class="fl">2</span>, <span class="st">"m"</span><span class="op">)</span><span class="op">)</span></span>
 <span><span class="co">#&gt; function (x, ...) </span></span>
 <span><span class="co">#&gt; pluck_raw(x, list(2, "m"), .default = NULL)</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x555ecb163178&gt;</span></span>
+<span><span class="co">#&gt; &lt;environment: 0x55d958fec178&gt;</span></span>
 <span></span>
 <span><span class="co"># compact functions ----------------------------</span></span>
 <span></span>
@@ -210,7 +210,7 @@ <h2>
 <span><span class="fu"><a href="https://purrr.tidyverse.org/reference/as_mapper.html">as_mapper</a></span><span class="op">(</span><span class="fu"><a href="https://rdrr.io/r/stats/Uniform.html">runif</a></span><span class="op">(</span><span class="fl">2</span><span class="op">)</span><span class="op">)</span></span>
 <span><span class="co">#&gt; function (x, ...) </span></span>
 <span><span class="co">#&gt; pluck_raw(x, list(0.597890264587477, 0.587997315218672), .default = NULL)</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x555ecc0f46f8&gt;</span></span></code></pre></div>
+<span><span class="co">#&gt; &lt;environment: 0x55d959f7d6f8&gt;</span></span></code></pre></div>
 <hr>
 <p><strong>Q3.</strong> Use the appropriate <code><a href="https://purrr.tidyverse.org/reference/map.html">map()</a></code> function to:</p>
 <pre><code>a) Compute the standard deviation of every column in a numeric data frame.
@@ -787,7 +787,7 @@ <h2>
 <span></span>
 <span><span class="va">e</span> <span class="op">&lt;-</span> <span class="fu"><a href="https://rlang.r-lib.org/reference/env.html">env</a></span><span class="op">(</span><span class="st">"x"</span> <span class="op">=</span> <span class="fl">1</span>, <span class="st">"y"</span> <span class="op">=</span> <span class="fl">2</span><span class="op">)</span></span>
 <span><span class="fu">rlang</span><span class="fu">::</span><span class="fu"><a href="https://rlang.r-lib.org/reference/env_print.html">env_print</a></span><span class="op">(</span><span class="va">e</span><span class="op">)</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x555ec937cf20&gt;</span></span>
+<span><span class="co">#&gt; &lt;environment: 0x55d9572060a8&gt;</span></span>
 <span><span class="co">#&gt; Parent: &lt;environment: global&gt;</span></span>
 <span><span class="co">#&gt; Bindings:</span></span>
 <span><span class="co">#&gt; • x: &lt;dbl&gt;</span></span>
@@ -901,7 +901,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -990,7 +990,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/functions.html b/functions.html
index d8ef65d2..9c3fb11d 100644
--- a/functions.html
+++ b/functions.html
@@ -99,7 +99,7 @@ <h2>
 <code class="sourceCode R"><span><span class="fu"><a href="https://rdrr.io/r/base/match.fun.html">match.fun</a></span><span class="op">(</span><span class="st">"mean"</span><span class="op">)</span></span>
 <span><span class="co">#&gt; function (x, ...) </span></span>
 <span><span class="co">#&gt; UseMethod("mean")</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x5610be5e25a8&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x55734d1505a8&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:base&gt;</span></span></code></pre></div>
 <p>But, given a function, it doesn’t make sense to find its name because there can be multiple names bound to the same function.</p>
 <div class="sourceCode" id="cb119"><pre class="downlit sourceCode r">
@@ -365,7 +365,7 @@ <h2>
 <span><span class="va">mean</span></span>
 <span><span class="co">#&gt; function (x, ...) </span></span>
 <span><span class="co">#&gt; UseMethod("mean")</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x5610be5e25a8&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x55734d1505a8&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:base&gt;</span></span>
 <span></span>
 <span><span class="co"># other package function</span></span>
@@ -374,7 +374,7 @@ <h2>
 <span><span class="co">#&gt; {</span></span>
 <span><span class="co">#&gt;     map_("list", .x, .f, ..., .progress = .progress)</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x5610c05c5a28&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x55734f135980&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:purrr&gt;</span></span></code></pre></div>
 <p>There are two exceptions where the enclosing environment won’t be printed:</p>
 <ul>
@@ -530,7 +530,7 @@ <h2>
 <span><span class="op">}</span></span>
 <span></span>
 <span><span class="fu">f2</span><span class="op">(</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "0x5610bed32bc0" "0x5610bed32bc0"</span></span>
+<span><span class="co">#&gt; [1] "0x55734d8a0d10" "0x55734d8a0d10"</span></span>
 <span><span class="co">#&gt; [1] 100</span></span></code></pre></div>
 <p><strong>Q3.</strong> What does this function return? Why? Which principle does it illustrate?</p>
 <div class="sourceCode" id="cb146"><pre class="downlit sourceCode r">
@@ -591,7 +591,7 @@ <h2>
 <span><span class="op">}</span></span>
 <span></span>
 <span><span class="fu">show_time</span><span class="op">(</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "2023-12-31 00:42:55 UTC"</span></span></code></pre></div>
+<span><span class="co">#&gt; [1] "2024-01-07 00:43:46 UTC"</span></span></code></pre></div>
 <p><strong>A5.</strong> Let’s take this step-by-step.</p>
 <p>The function argument <code>x</code> is missing in the function call. This means that <code>stop("Error!")</code> is evaluated in the function environment, and not global environment.</p>
 <p>But, due to lazy evaluation, the promise <code>stop("Error!")</code> is evaluated only when <code>x</code> is accessed. This happens only when <code>print(x)</code> is called.</p>
@@ -742,7 +742,7 @@ <h2>
 <span><span class="co">#&gt;     on.exit(setwd(old))</span></span>
 <span><span class="co">#&gt;     force(code)</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x5610c7d2faa0&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x55735689cc20&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:withr&gt;</span></span></code></pre></div>
 <p>More importantly, its parameters <code>dir</code> allows temporarily changing working directory to <em>any</em> directory.</p>
 <p><strong>Q4.</strong> Write a function that opens a graphics device, runs the supplied code, and closes the graphics device (always, regardless of whether or not the plotting code works).</p>
@@ -811,7 +811,7 @@ <h2>
 <span><span class="co">#&gt;         invisible(NULL)</span></span>
 <span><span class="co">#&gt;     else rval</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x5610c81bfaa0&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x557356d2cc20&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:utils&gt;</span></span></code></pre></div>
 <p>Here are few key differences:</p>
 <ul>
@@ -1070,7 +1070,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -1178,7 +1178,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/improving-performance.html b/improving-performance.html
index 52c59687..18ccc08c 100644
--- a/improving-performance.html
+++ b/improving-performance.html
@@ -136,10 +136,10 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 4 × 5</span></span>
 <span><span class="co">#&gt;   expression      min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt; &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 lm         860.49µs 897.15µs     1094.    1.27MB</span></span>
-<span><span class="co">#&gt; 2 speedglm     1.56ms   1.63ms      611.   75.01MB</span></span>
-<span><span class="co">#&gt; 3 biglm      767.92µs 797.64µs     1189.   589.1KB</span></span>
-<span><span class="co">#&gt; 4 fastLm     987.36µs   1.01ms      982.    4.48MB</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 lm          881.5µs  916.5µs     1045.    1.27MB</span></span>
+<span><span class="co">#&gt; 2 speedglm      1.6ms   1.66ms      601.   75.01MB</span></span>
+<span><span class="co">#&gt; 3 biglm       772.1µs 809.23µs     1174.   589.1KB</span></span>
+<span><span class="co">#&gt; 4 fastLm        993µs   1.02ms      978.    4.48MB</span></span></code></pre></div>
 <p>The results might change depending on the size of the dataset, with the performance benefits accruing bigger the dataset.</p>
 <p>You will have to experiment with different algorithms and find the one that fits the needs of your dataset the best.</p>
 <p><strong>Q2.</strong> What package implements a version of <code><a href="https://rdrr.io/r/base/match.html">match()</a></code> that’s faster for repeated look ups? How much faster is it?</p>
@@ -164,8 +164,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 5</span></span>
 <span><span class="co">#&gt;   expression      min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt; &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 base         1.14µs   1.18µs   796519.     2.8KB</span></span>
-<span><span class="co">#&gt; 2 fastmatch  990.93ns   1.04µs   929883.    2.66KB</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 base         1.18µs   1.24µs   756897.     2.8KB</span></span>
+<span><span class="co">#&gt; 2 fastmatch    1.02µs   1.06µs   904225.    2.66KB</span></span></code></pre></div>
 <p>But, with a larger vector, <code><a href="https://rdrr.io/pkg/fastmatch/man/fmatch.html">fmatch()</a></code> is orders of magnitude faster! ⚡</p>
 <div class="sourceCode" id="cb651"><pre class="downlit sourceCode r">
 <code class="sourceCode R"><span><span class="va">large_vec</span> <span class="op">&lt;-</span> <span class="fu"><a href="https://rdrr.io/r/base/c.html">c</a></span><span class="op">(</span><span class="fu"><a href="https://rdrr.io/r/base/rep.html">rep</a></span><span class="op">(</span><span class="fu"><a href="https://rdrr.io/r/base/c.html">c</a></span><span class="op">(</span><span class="st">"a"</span>, <span class="st">"b"</span><span class="op">)</span>, <span class="fl">1e4</span><span class="op">)</span>, <span class="st">"x"</span>, <span class="fu"><a href="https://rdrr.io/r/base/rep.html">rep</a></span><span class="op">(</span><span class="fu"><a href="https://rdrr.io/r/base/c.html">c</a></span><span class="op">(</span><span class="st">"m"</span>, <span class="st">"n"</span><span class="op">)</span>, <span class="fl">1e6</span><span class="op">)</span>, <span class="st">"y"</span><span class="op">)</span></span>
@@ -180,8 +180,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 5</span></span>
 <span><span class="co">#&gt;   expression      min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt; &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 base         23.3ms   23.4ms      42.7    31.4MB</span></span>
-<span><span class="co">#&gt; 2 fastmatch   990.9ns   1.04µs  917170.         0B</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 base         20.6ms  21.95ms      43.0    31.4MB</span></span>
+<span><span class="co">#&gt; 2 fastmatch   992.1ns   1.04µs  897837.         0B</span></span></code></pre></div>
 <p>We can also look at the hash table:</p>
 <div class="sourceCode" id="cb652"><pre class="downlit sourceCode r">
 <code class="sourceCode R"><span><span class="fu"><a href="https://rdrr.io/pkg/fastmatch/man/fmatch.html">fmatch.hash</a></span><span class="op">(</span><span class="fu"><a href="https://rdrr.io/r/base/c.html">c</a></span><span class="op">(</span><span class="st">"x"</span>, <span class="st">"y"</span><span class="op">)</span>, <span class="va">small_vec</span><span class="op">)</span></span>
@@ -238,10 +238,10 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 4 × 6</span></span>
 <span><span class="co">#&gt;   expression       min   median `itr/sec` mem_alloc `gc/sec`</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;  &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;    &lt;dbl&gt;</span></span>
-<span><span class="co">#&gt; 1 as.POSIXct   29.97µs   31.6µs    31256.        0B     0   </span></span>
-<span><span class="co">#&gt; 2 as.POSIXlt   20.65µs  21.83µs    45281.        0B     0   </span></span>
-<span><span class="co">#&gt; 3 ymd_hms       2.22ms   2.29ms      432.    21.5KB     3.48</span></span>
-<span><span class="co">#&gt; 4 fastPOSIXct   1.17µs   1.22µs   779859.        0B     0</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 as.POSIXct   29.74µs   31.4µs    31531.        0B     0   </span></span>
+<span><span class="co">#&gt; 2 as.POSIXlt   20.59µs  21.79µs    45434.        0B     0   </span></span>
+<span><span class="co">#&gt; 3 ymd_hms       2.25ms   2.33ms      425.    21.5KB     3.43</span></span>
+<span><span class="co">#&gt; 4 fastPOSIXct   1.16µs   1.22µs   766780.        0B     0</span></span></code></pre></div>
 <p>There are many more packages that implement a way to convert from string to a date time object. For more, see <a href="https://cran.r-project.org/web/views/TimeSeries.html">CRAN Task View: Time Series Analysis</a></p>
 <p><strong>Q4.</strong> Which packages provide the ability to compute a rolling mean?</p>
 <p><strong>A4.</strong> Here are a few packages and respective functions that provide a way to compute a rolling mean:</p>
@@ -297,7 +297,7 @@ <h2>
 <span><span class="co">#&gt;     else names(z) &lt;- dimnames(x)[[1L]]</span></span>
 <span><span class="co">#&gt;     z</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x55a2a63ce4b0&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x563353356400&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:base&gt;</span></span></code></pre></div>
 <ul>
 <li>
@@ -307,7 +307,7 @@ <h2>
 <code class="sourceCode R"><span><span class="va">.rowSums</span></span>
 <span><span class="co">#&gt; function (x, m, n, na.rm = FALSE) </span></span>
 <span><span class="co">#&gt; .Internal(rowSums(x, m, n, na.rm))</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x55a2ac309f60&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x56335cd3f7a8&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:base&gt;</span></span></code></pre></div>
 <p>But they have comparable performance:</p>
 <div class="sourceCode" id="cb661"><pre class="downlit sourceCode r">
@@ -320,8 +320,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 5</span></span>
 <span><span class="co">#&gt;   expression      min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt; &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 rowSums       821µs   1.29ms      761.     859KB</span></span>
-<span><span class="co">#&gt; 2 .rowSums      818µs   1.28ms      837.     859KB</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 rowSums       821µs   1.29ms      859.     859KB</span></span>
+<span><span class="co">#&gt; 2 .rowSums      817µs   1.28ms      833.     859KB</span></span></code></pre></div>
 <p><strong>Q2.</strong> Make a faster version of <code><a href="https://rdrr.io/r/stats/chisq.test.html">chisq.test()</a></code> that only computes the chi-square test statistic when the input is two numeric vectors with no missing values. You can try simplifying <code><a href="https://rdrr.io/r/stats/chisq.test.html">chisq.test()</a></code> or by coding from the <a href="http://en.wikipedia.org/wiki/Pearson%27s_chi-squared_test">mathematical definition</a>.</p>
 <p><strong>A2.</strong> If the function is supposed to accept only two numeric vectors without missing values, then we can make <code><a href="https://rdrr.io/r/stats/chisq.test.html">chisq.test()</a></code> do less work by removing code corresponding to the following :</p>
 <ul>
@@ -379,8 +379,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 5</span></span>
 <span><span class="co">#&gt;   expression      min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt; &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 base          904µs    927µs     1008.    1.57MB</span></span>
-<span><span class="co">#&gt; 2 custom        708µs    721µs     1376.    1.13MB</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 base          906µs    926µs     1071.    1.57MB</span></span>
+<span><span class="co">#&gt; 2 custom        706µs    722µs     1371.    1.13MB</span></span></code></pre></div>
 <p><strong>Q3.</strong> Can you make a faster version of <code><a href="https://rdrr.io/r/base/table.html">table()</a></code> for the case of an input of two integer vectors with no missing values? Can you use it to speed up your chi-square test?</p>
 <p><strong>A3.</strong> In order to make a leaner version of <code><a href="https://rdrr.io/r/base/table.html">table()</a></code>, we can take a similar approach and trim the unnecessary input checks in light of our new API of accepting just two vectors without missing values. We can remove the following components from the code:</p>
 <ul>
@@ -428,8 +428,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 5</span></span>
 <span><span class="co">#&gt;   expression      min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt; &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 base          638µs    652µs     1519.     960KB</span></span>
-<span><span class="co">#&gt; 2 custom        355µs    362µs     2736.     488KB</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 base          638µs    653µs     1515.     960KB</span></span>
+<span><span class="co">#&gt; 2 custom        354µs    359µs     2730.     488KB</span></span></code></pre></div>
 <p>We can also use this function in our custom chi-squared test function and see if the performance improves any further:</p>
 <div class="sourceCode" id="cb666"><pre class="downlit sourceCode r">
 <code class="sourceCode R"><span><span class="va">my_chisq_test2</span> <span class="op">&lt;-</span> <span class="kw">function</span><span class="op">(</span><span class="va">x</span>, <span class="va">y</span><span class="op">)</span> <span class="op">{</span></span>
@@ -479,8 +479,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 5</span></span>
 <span><span class="co">#&gt;   expression      min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt; &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 base          909µs    940µs     1059.    1.28MB</span></span>
-<span><span class="co">#&gt; 2 custom        418µs    427µs     2314.  594.45KB</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 base          912µs    940µs     1057.    1.28MB</span></span>
+<span><span class="co">#&gt; 2 custom        420µs    429µs     2296.  594.45KB</span></span></code></pre></div>
 </div>
 <div id="exercises-24.5.1" class="section level2" number="24.3">
 <h2>
@@ -569,8 +569,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 5</span></span>
 <span><span class="co">#&gt;   expression                min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;           &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 crossprod(x, w)[[1]]    871ns    922ns  1001700.        0B</span></span>
-<span><span class="co">#&gt; 2 sum(x * w)[[1]]         460ns    511ns  1796479.        0B</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 crossprod(x, w)[[1]]    901ns    961ns   966191.        0B</span></span>
+<span><span class="co">#&gt; 2 sum(x * w)[[1]]         461ns    511ns  1759060.        0B</span></span></code></pre></div>
 
 </div>
 </div>
@@ -602,7 +602,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/index.html b/index.html
index 98d3d87f..7aa4e766 100644
--- a/index.html
+++ b/index.html
@@ -113,7 +113,7 @@ <h1>About<a class="anchor" aria-label="anchor" href="#about"><i class="fas fa-li
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/index.md b/index.md
index 0abfc2d2..931bccde 100644
--- a/index.md
+++ b/index.md
@@ -1,7 +1,7 @@
 --- 
 title: "Advanced R Exercises"
 author: '<a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>'
-date: "2023-12-31"
+date: "2024-01-07"
 site: bookdown::bookdown_site
 documentclass: book
 bibliography: [book.bib, packages.bib]
diff --git a/introduction.html b/introduction.html
index 99150122..17cf53a4 100644
--- a/introduction.html
+++ b/introduction.html
@@ -111,7 +111,7 @@ <h1>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/measuring-performance.html b/measuring-performance.html
index b7bc82b3..40df4ca7 100644
--- a/measuring-performance.html
+++ b/measuring-performance.html
@@ -131,7 +131,7 @@ <h2>
 <span><span class="co">#&gt;     }</span></span>
 <span><span class="co">#&gt;     .Internal(remove(list, envir, inherits))</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x56294c35adf0&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x56191ab34df0&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:base&gt;</span></span></code></pre></div>
 <p>I still couldn’t figure out why. I would recommend checking out the <a href="https://advanced-r-solutions.rbind.io/measuring-performance.html#profiling">official answer</a>.</p>
 <hr>
@@ -194,8 +194,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 4</span></span>
 <span><span class="co">#&gt;   expression bench_mean systime_with_gc systime_with_nogc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;   &lt;bch:tm&gt;           &lt;dbl&gt;             &lt;dbl&gt;</span></span>
-<span><span class="co">#&gt; 1 sqrt(x)      670.07ns           0.629             0.443</span></span>
-<span><span class="co">#&gt; 2 x^0.5          2.19µs           2.01              2.02</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 sqrt(x)       677.2ns           0.629             0.428</span></span>
+<span><span class="co">#&gt; 2 x^0.5          2.23µs           2.02              2.00</span></span></code></pre></div>
 <p>The comparison reveals that these two approaches yield quite similar results. Slight differences in exact values is possibly due to differences in the precision of timers used internally by these functions.</p>
 <hr>
 <p><strong>Q2.</strong> Here are two other ways to compute the square root of a vector. Which do you think will be fastest? Which will be slowest? Use microbenchmarking to test your answers.</p>
@@ -219,10 +219,10 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 4 × 2</span></span>
 <span><span class="co">#&gt;   expression      median</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;    &lt;bch:tm&gt;</span></span>
-<span><span class="co">#&gt; 1 sqrt(x)         3.03µs</span></span>
-<span><span class="co">#&gt; 2 exp(log(x)/2)  12.62µs</span></span>
-<span><span class="co">#&gt; 3 x^0.5          18.86µs</span></span>
-<span><span class="co">#&gt; 4 x^(1/2)           19µs</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 sqrt(x)         3.02µs</span></span>
+<span><span class="co">#&gt; 2 exp(log(x)/2)  12.66µs</span></span>
+<span><span class="co">#&gt; 3 x^0.5          18.96µs</span></span>
+<span><span class="co">#&gt; 4 x^(1/2)        19.01µs</span></span></code></pre></div>
 <p>The specialized primitive function <code><a href="https://rdrr.io/r/base/MathFun.html">sqrt()</a></code> (written in <code>C</code>) is the fastest way to compute square root.</p>
 <hr>
 </div>
@@ -242,7 +242,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -334,7 +334,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/names-and-values.html b/names-and-values.html
index f5e7e151..e8a70a9a 100644
--- a/names-and-values.html
+++ b/names-and-values.html
@@ -104,11 +104,11 @@ <h2>
 <div class="sourceCode" id="cb3"><pre class="downlit sourceCode r">
 <code class="sourceCode R"><span><span class="va">obj_addrs</span> <span class="op">&lt;-</span> <span class="fu"><a href="https://lobstr.r-lib.org/reference/obj_addr.html">obj_addrs</a></span><span class="op">(</span><span class="fu"><a href="https://rdrr.io/r/base/list.html">list</a></span><span class="op">(</span><span class="va">a</span>, <span class="va">b</span>, <span class="va">c</span><span class="op">)</span><span class="op">)</span></span>
 <span><span class="fu"><a href="https://rdrr.io/r/base/unique.html">unique</a></span><span class="op">(</span><span class="va">obj_addrs</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "0x556dced402c8"</span></span></code></pre></div>
+<span><span class="co">#&gt; [1] "0x56426e6d12c8"</span></span></code></pre></div>
 <p>Except <code>d</code>, which is a different object, even if it has the same value as <code>a</code>, <code>b</code>, and <code>c</code>:</p>
 <div class="sourceCode" id="cb4"><pre class="downlit sourceCode r">
 <code class="sourceCode R"><span><span class="fu"><a href="https://lobstr.r-lib.org/reference/obj_addr.html">obj_addr</a></span><span class="op">(</span><span class="va">d</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "0x556dce9f9fc8"</span></span></code></pre></div>
+<span><span class="co">#&gt; [1] "0x56426e38afc8"</span></span></code></pre></div>
 <hr>
 <p><strong>Q2.</strong> The following code accesses the mean function in multiple ways. Do they all point to the same underlying function object? Verify this with <code><a href="https://lobstr.r-lib.org/reference/obj_addr.html">lobstr::obj_addr()</a></code>.</p>
 <div class="sourceCode" id="cb5"><pre class="downlit sourceCode r">
@@ -128,7 +128,7 @@ <h2>
 <span><span class="op">)</span><span class="op">)</span></span>
 <span></span>
 <span><span class="fu"><a href="https://rdrr.io/r/base/unique.html">unique</a></span><span class="op">(</span><span class="va">obj_addrs</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "0x556dcc722500"</span></span></code></pre></div>
+<span><span class="co">#&gt; [1] "0x56426c0b3500"</span></span></code></pre></div>
 <hr>
 <p><strong>Q3.</strong> By default, base R data import functions, like <code><a href="https://rdrr.io/r/utils/read.table.html">read.csv()</a></code>, will automatically convert non-syntactic names to syntactic ones. Why might this be problematic? What option allows you to suppress this behaviour?</p>
 <p><strong>A3.</strong> The conversion of non-syntactic names to syntactic ones can sometimes corrupt the data. Some datasets may require non-syntactic names.</p>
@@ -170,7 +170,7 @@ <h2>
 <code class="sourceCode R"><span><span class="va">x</span> <span class="op">&lt;-</span> <span class="fl">1</span><span class="op">:</span><span class="fl">10</span></span>
 <span></span>
 <span><span class="fu"><a href="https://rdrr.io/r/base/tracemem.html">tracemem</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "&lt;0x556dd0825628&gt;"</span></span>
+<span><span class="co">#&gt; [1] "&lt;0x5642701b6628&gt;"</span></span>
 <span></span>
 <span><span class="va">x</span> <span class="op">&lt;-</span> <span class="va">x</span> <span class="op">+</span> <span class="fl">1</span></span>
 <span></span>
@@ -178,10 +178,10 @@ <h2>
 <p>But since the object created in memory by <code>1:10</code> is not assigned a name, it can’t be addressed or modified from R, and so there is nothing to trace.</p>
 <div class="sourceCode" id="cb10"><pre class="downlit sourceCode r">
 <code class="sourceCode R"><span><span class="fu"><a href="https://lobstr.r-lib.org/reference/obj_addr.html">obj_addr</a></span><span class="op">(</span><span class="fl">1</span><span class="op">:</span><span class="fl">10</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "0x556dd02fa0f8"</span></span>
+<span><span class="co">#&gt; [1] "0x56426fc8b0f8"</span></span>
 <span></span>
 <span><span class="fu"><a href="https://rdrr.io/r/base/tracemem.html">tracemem</a></span><span class="op">(</span><span class="fl">1</span><span class="op">:</span><span class="fl">10</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "&lt;0x556dd03667a0&gt;"</span></span></code></pre></div>
+<span><span class="co">#&gt; [1] "&lt;0x56426fcf77a0&gt;"</span></span></code></pre></div>
 <hr>
 <p><strong>Q2.</strong> Explain why <code><a href="https://rdrr.io/r/base/tracemem.html">tracemem()</a></code> shows two copies when you run this code. Hint: carefully look at the difference between this code and the code shown earlier in the section.</p>
 <div class="sourceCode" id="cb11"><pre class="downlit sourceCode r">
@@ -196,11 +196,11 @@ <h2>
 <span><span class="fu"><a href="https://rdrr.io/r/base/typeof.html">typeof</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
 <span><span class="co">#&gt; [1] "integer"</span></span>
 <span><span class="fu"><a href="https://rdrr.io/r/base/tracemem.html">tracemem</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "&lt;0x556dd0baaad8&gt;"</span></span>
+<span><span class="co">#&gt; [1] "&lt;0x56427053bad8&gt;"</span></span>
 <span></span>
 <span><span class="va">x</span><span class="op">[[</span><span class="fl">3</span><span class="op">]</span><span class="op">]</span> <span class="op">&lt;-</span> <span class="fl">4</span></span>
-<span><span class="co">#&gt; tracemem[0x556dd0baaad8 -&gt; 0x556dd0cd6558]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call &lt;Anonymous&gt; evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file &lt;Anonymous&gt; &lt;Anonymous&gt; do.call eval eval eval eval eval.parent local </span></span>
-<span><span class="co">#&gt; tracemem[0x556dd0cd6558 -&gt; 0x556dd0ce41d8]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call &lt;Anonymous&gt; evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file &lt;Anonymous&gt; &lt;Anonymous&gt; do.call eval eval eval eval eval.parent local</span></span>
+<span><span class="co">#&gt; tracemem[0x56427053bad8 -&gt; 0x564270667558]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call &lt;Anonymous&gt; evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file &lt;Anonymous&gt; &lt;Anonymous&gt; do.call eval eval eval eval eval.parent local </span></span>
+<span><span class="co">#&gt; tracemem[0x564270667558 -&gt; 0x564270675188]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call &lt;Anonymous&gt; evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file &lt;Anonymous&gt; &lt;Anonymous&gt; do.call eval eval eval eval eval.parent local</span></span>
 <span><span class="fu"><a href="https://rdrr.io/r/base/tracemem.html">untracemem</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
 <span></span>
 <span><span class="fu"><a href="https://rdrr.io/r/base/typeof.html">typeof</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
@@ -211,10 +211,10 @@ <h2>
 <span><span class="fu"><a href="https://rdrr.io/r/base/typeof.html">typeof</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
 <span><span class="co">#&gt; [1] "integer"</span></span>
 <span><span class="fu"><a href="https://rdrr.io/r/base/tracemem.html">tracemem</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "&lt;0x556dd11a3198&gt;"</span></span>
+<span><span class="co">#&gt; [1] "&lt;0x564270b34198&gt;"</span></span>
 <span></span>
 <span><span class="va">x</span><span class="op">[[</span><span class="fl">3</span><span class="op">]</span><span class="op">]</span> <span class="op">&lt;-</span> <span class="fl">4L</span></span>
-<span><span class="co">#&gt; tracemem[0x556dd11a3198 -&gt; 0x556dd12cee18]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call &lt;Anonymous&gt; evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file &lt;Anonymous&gt; &lt;Anonymous&gt; do.call eval eval eval eval eval.parent local</span></span>
+<span><span class="co">#&gt; tracemem[0x564270b34198 -&gt; 0x564270c5fe18]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call &lt;Anonymous&gt; evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file &lt;Anonymous&gt; &lt;Anonymous&gt; do.call eval eval eval eval eval.parent local</span></span>
 <span><span class="fu"><a href="https://rdrr.io/r/base/tracemem.html">untracemem</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
 <span></span>
 <span><span class="fu"><a href="https://rdrr.io/r/base/typeof.html">typeof</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
@@ -236,20 +236,20 @@ <h2>
 <span><span class="va">c</span> <span class="op">&lt;-</span> <span class="fu"><a href="https://rdrr.io/r/base/list.html">list</a></span><span class="op">(</span><span class="va">b</span>, <span class="va">a</span>, <span class="fl">1</span><span class="op">:</span><span class="fl">10</span><span class="op">)</span></span>
 <span></span>
 <span><span class="fu"><a href="https://lobstr.r-lib.org/reference/ref.html">ref</a></span><span class="op">(</span><span class="va">a</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1:0x556dd198acc8] &lt;int&gt;</span></span>
+<span><span class="co">#&gt; [1:0x56427131bcc8] &lt;int&gt;</span></span>
 <span></span>
 <span><span class="fu"><a href="https://lobstr.r-lib.org/reference/ref.html">ref</a></span><span class="op">(</span><span class="va">b</span><span class="op">)</span></span>
-<span><span class="co">#&gt; █ [1:0x556dd1eb64c8] &lt;list&gt; </span></span>
-<span><span class="co">#&gt; ├─[2:0x556dd198acc8] &lt;int&gt; </span></span>
-<span><span class="co">#&gt; └─[2:0x556dd198acc8]</span></span>
+<span><span class="co">#&gt; █ [1:0x5642718493e8] &lt;list&gt; </span></span>
+<span><span class="co">#&gt; ├─[2:0x56427131bcc8] &lt;int&gt; </span></span>
+<span><span class="co">#&gt; └─[2:0x56427131bcc8]</span></span>
 <span></span>
 <span><span class="fu"><a href="https://lobstr.r-lib.org/reference/ref.html">ref</a></span><span class="op">(</span><span class="va">c</span><span class="op">)</span></span>
-<span><span class="co">#&gt; █ [1:0x556dd1f3ee48] &lt;list&gt; </span></span>
-<span><span class="co">#&gt; ├─█ [2:0x556dd1eb64c8] &lt;list&gt; </span></span>
-<span><span class="co">#&gt; │ ├─[3:0x556dd198acc8] &lt;int&gt; </span></span>
-<span><span class="co">#&gt; │ └─[3:0x556dd198acc8] </span></span>
-<span><span class="co">#&gt; ├─[3:0x556dd198acc8] </span></span>
-<span><span class="co">#&gt; └─[4:0x556dd1f41178] &lt;int&gt;</span></span></code></pre></div>
+<span><span class="co">#&gt; █ [1:0x5642718cfda8] &lt;list&gt; </span></span>
+<span><span class="co">#&gt; ├─█ [2:0x5642718493e8] &lt;list&gt; </span></span>
+<span><span class="co">#&gt; │ ├─[3:0x56427131bcc8] &lt;int&gt; </span></span>
+<span><span class="co">#&gt; │ └─[3:0x56427131bcc8] </span></span>
+<span><span class="co">#&gt; ├─[3:0x56427131bcc8] </span></span>
+<span><span class="co">#&gt; └─[4:0x5642718d2178] &lt;int&gt;</span></span></code></pre></div>
 <p>Here is what we learn:</p>
 <ul>
 <li>The name <code>a</code> references object <code>1:10</code> in the memory.</li>
@@ -269,7 +269,7 @@ <h2>
 <span><span class="co">#&gt; [[1]]</span></span>
 <span><span class="co">#&gt;  [1]  1  2  3  4  5  6  7  8  9 10</span></span>
 <span><span class="fu"><a href="https://lobstr.r-lib.org/reference/obj_addr.html">obj_addr</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "0x556dcfcf8e40"</span></span>
+<span><span class="co">#&gt; [1] "0x56426f689e78"</span></span>
 <span></span>
 <span><span class="va">x</span><span class="op">[[</span><span class="fl">2</span><span class="op">]</span><span class="op">]</span> <span class="op">&lt;-</span> <span class="va">x</span></span>
 <span><span class="va">x</span></span>
@@ -280,13 +280,13 @@ <h2>
 <span><span class="co">#&gt; [[2]][[1]]</span></span>
 <span><span class="co">#&gt;  [1]  1  2  3  4  5  6  7  8  9 10</span></span>
 <span><span class="fu"><a href="https://lobstr.r-lib.org/reference/obj_addr.html">obj_addr</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "0x556dd04f3148"</span></span>
+<span><span class="co">#&gt; [1] "0x56426fe84148"</span></span>
 <span></span>
 <span><span class="fu"><a href="https://lobstr.r-lib.org/reference/ref.html">ref</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
-<span><span class="co">#&gt; █ [1:0x556dd04f3148] &lt;list&gt; </span></span>
-<span><span class="co">#&gt; ├─[2:0x556dcf554b38] &lt;int&gt; </span></span>
-<span><span class="co">#&gt; └─█ [3:0x556dcfcf8e40] &lt;list&gt; </span></span>
-<span><span class="co">#&gt;   └─[2:0x556dcf554b38]</span></span></code></pre></div>
+<span><span class="co">#&gt; █ [1:0x56426fe84148] &lt;list&gt; </span></span>
+<span><span class="co">#&gt; ├─[2:0x56426eee5b38] &lt;int&gt; </span></span>
+<span><span class="co">#&gt; └─█ [3:0x56426f689e78] &lt;list&gt; </span></span>
+<span><span class="co">#&gt;   └─[2:0x56426eee5b38]</span></span></code></pre></div>
 <p>I don’t have access to OmniGraffle software, so I am including here the figure from the <a href="https://advanced-r-solutions.rbind.io/names-and-values.html#copy-on-modify">official solution manual</a>:</p>
 <div class="inline-figure"><img src="https://raw.githubusercontent.com/Tazinho/Advanced-R-Solutions/main/images/names_values/copy_on_modify_fig2.png" width="180pt"></div>
 <hr>
@@ -397,16 +397,16 @@ <h2>
 <code class="sourceCode R"><span><span class="va">x</span> <span class="op">&lt;-</span> <span class="fu"><a href="https://rdrr.io/r/base/list.html">list</a></span><span class="op">(</span><span class="op">)</span></span>
 <span></span>
 <span><span class="fu"><a href="https://lobstr.r-lib.org/reference/obj_addr.html">obj_addr</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "0x556dd095be30"</span></span>
+<span><span class="co">#&gt; [1] "0x5642702ece68"</span></span>
 <span></span>
 <span><span class="fu"><a href="https://rdrr.io/r/base/tracemem.html">tracemem</a></span><span class="op">(</span><span class="va">x</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "&lt;0x556dd095be30&gt;"</span></span>
+<span><span class="co">#&gt; [1] "&lt;0x5642702ece68&gt;"</span></span>
 <span></span>
 <span><span class="va">x</span><span class="op">[[</span><span class="fl">1</span><span class="op">]</span><span class="op">]</span> <span class="op">&lt;-</span> <span class="va">x</span></span>
-<span><span class="co">#&gt; tracemem[0x556dd095be30 -&gt; 0x556dd02523e8]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call &lt;Anonymous&gt; evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file &lt;Anonymous&gt; &lt;Anonymous&gt; do.call eval eval eval eval eval.parent local</span></span>
+<span><span class="co">#&gt; tracemem[0x5642702ece68 -&gt; 0x56426fbe3420]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call &lt;Anonymous&gt; evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file &lt;Anonymous&gt; &lt;Anonymous&gt; do.call eval eval eval eval eval.parent local</span></span>
 <span></span>
 <span><span class="fu"><a href="https://lobstr.r-lib.org/reference/obj_addr.html">obj_addr</a></span><span class="op">(</span><span class="va">x</span><span class="op">[[</span><span class="fl">1</span><span class="op">]</span><span class="op">]</span><span class="op">)</span></span>
-<span><span class="co">#&gt; [1] "0x556dd095be30"</span></span></code></pre></div>
+<span><span class="co">#&gt; [1] "0x5642702ece68"</span></span></code></pre></div>
 <hr>
 <p><strong>Q2.</strong> Wrap the two methods for subtracting medians into two functions, then use the ‘bench’ package to carefully compare their speeds. How does performance change as the number of columns increase?</p>
 <p><strong>A2.</strong> Let’s first microbenchmark functions that do and do not create copies for varying lengths of number of columns.</p>
@@ -507,7 +507,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -619,7 +619,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/quasiquotation.html b/quasiquotation.html
index 1d79a3ee..99562fd6 100644
--- a/quasiquotation.html
+++ b/quasiquotation.html
@@ -218,7 +218,7 @@ <h2>
 <span><span class="co">#&gt; {</span></span>
 <span><span class="co">#&gt;     enexpr(expr)</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x55b385e4f448&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x5584f9fd1728&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:rlang&gt;</span></span></code></pre></div>
 <p>For example:</p>
 <div class="sourceCode" id="cb548"><pre class="downlit sourceCode r">
@@ -232,7 +232,7 @@ <h2>
 <span><span class="co">#&gt; {</span></span>
 <span><span class="co">#&gt;     .Call(ffi_enexpr, substitute(arg), parent.frame())</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x55b3829aaae0&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x5584f6b2cae0&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:rlang&gt;</span></span></code></pre></div>
 <hr>
 <p><strong>Q2.</strong> Compare and contrast the following two functions. Can you predict the output before running them?</p>
@@ -742,7 +742,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -846,7 +846,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/r6.html b/r6.html
index 172c3a58..f53c8c2b 100644
--- a/r6.html
+++ b/r6.html
@@ -302,7 +302,7 @@ <h2>
 <span><span class="co">#&gt; [1] "environment"</span></span>
 <span></span>
 <span><span class="fu">rlang</span><span class="fu">::</span><span class="fu"><a href="https://rlang.r-lib.org/reference/env_print.html">env_print</a></span><span class="op">(</span><span class="va">myExample</span><span class="op">)</span></span>
-<span><span class="co">#&gt; &lt;environment: 0x55ffdaaac290&gt; [L]</span></span>
+<span><span class="co">#&gt; &lt;environment: 0x5565e0c0a290&gt; [L]</span></span>
 <span><span class="co">#&gt; Parent: &lt;environment: empty&gt;</span></span>
 <span><span class="co">#&gt; Class: Example, R6</span></span>
 <span><span class="co">#&gt; Bindings:</span></span>
@@ -530,7 +530,7 @@ <h2>
 <span><span class="co">#&gt; [1] "Closing the file connection!"</span></span>
 <span><span class="co">#&gt;           used (Mb) gc trigger (Mb) max used (Mb)</span></span>
 <span><span class="co">#&gt; Ncells  746818 39.9    1365334   73  1365334 73.0</span></span>
-<span><span class="co">#&gt; Vcells 1372826 10.5    8388608   64  2594866 19.8</span></span></code></pre></div>
+<span><span class="co">#&gt; Vcells 1372826 10.5    8388608   64  2594870 19.8</span></span></code></pre></div>
 </div>
 <div id="session-information-11" class="section level2" number="14.4">
 <h2>
@@ -548,7 +548,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       Europe/Berlin</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -629,7 +629,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/rewriting-r-code-in-c.html b/rewriting-r-code-in-c.html
index 48175f44..c86e7a48 100644
--- a/rewriting-r-code-in-c.html
+++ b/rewriting-r-code-in-c.html
@@ -261,12 +261,12 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression                                      min</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;                                 &lt;bch:tm&gt;</span></span>
-<span><span class="co">#&gt; 1 all(c(rep(TRUE, 1000), rep(FALSE, 1000)))    6.16µs</span></span>
-<span><span class="co">#&gt; 2 allC(c(rep(TRUE, 1000), rep(FALSE, 1000)))   8.59µs</span></span>
+<span><span class="co">#&gt; 1 all(c(rep(TRUE, 1000), rep(FALSE, 1000)))    6.13µs</span></span>
+<span><span class="co">#&gt; 2 allC(c(rep(TRUE, 1000), rep(FALSE, 1000)))   8.73µs</span></span>
 <span><span class="co">#&gt;     median `itr/sec` mem_alloc `gc/sec`</span></span>
 <span><span class="co">#&gt;   &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;    &lt;dbl&gt;</span></span>
-<span><span class="co">#&gt; 1   6.46µs   152676.    15.8KB        0</span></span>
-<span><span class="co">#&gt; 2   8.95µs   109377.    18.3KB        0</span></span></code></pre></div>
+<span><span class="co">#&gt; 1   6.49µs   152211.    15.8KB        0</span></span>
+<span><span class="co">#&gt; 2   9.15µs   106881.    18.3KB        0</span></span></code></pre></div>
 <ul>
 <li><code><a href="https://rdrr.io/r/base/cumsum.html">cumprod()</a></code></li>
 </ul>
@@ -301,8 +301,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression        min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;   &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 cumprod(v1)  110.01ns 130.04ns  6991133.        0B</span></span>
-<span><span class="co">#&gt; 2 cumprodC(v1)   1.44µs   1.53µs   619348.    6.62KB</span></span>
+<span><span class="co">#&gt; 1 cumprod(v1)   89.99ns 101.05ns  8156808.        0B</span></span>
+<span><span class="co">#&gt; 2 cumprodC(v1)   1.45µs   1.55µs   612672.    6.62KB</span></span>
 <span><span class="co">#&gt;   `gc/sec`</span></span>
 <span><span class="co">#&gt;      &lt;dbl&gt;</span></span>
 <span><span class="co">#&gt; 1        0</span></span>
@@ -342,8 +342,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression       min   median `itr/sec` mem_alloc `gc/sec`</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;  &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;    &lt;dbl&gt;</span></span>
-<span><span class="co">#&gt; 1 cummin(v1)  130.97ns 151.57ns  5389797.        0B        0</span></span>
-<span><span class="co">#&gt; 2 cumminC(v1)   1.52µs   1.83µs   514890.    6.62KB        0</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 cummin(v1)  120.02ns 131.08ns  5672695.        0B        0</span></span>
+<span><span class="co">#&gt; 2 cumminC(v1)   1.56µs   1.87µs   507394.    6.62KB        0</span></span></code></pre></div>
 <ul>
 <li><code>cummaxC()</code></li>
 </ul>
@@ -379,8 +379,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression       min   median `itr/sec` mem_alloc `gc/sec`</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;  &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;    &lt;dbl&gt;</span></span>
-<span><span class="co">#&gt; 1 cummax(v1)  110.01ns 120.14ns  6760800.        0B        0</span></span>
-<span><span class="co">#&gt; 2 cummaxC(v1)   1.55µs   1.72µs   549982.    6.62KB        0</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 cummax(v1)  110.01ns 121.07ns  6496158.        0B        0</span></span>
+<span><span class="co">#&gt; 2 cummaxC(v1)   1.58µs   1.76µs   545922.    6.62KB        0</span></span></code></pre></div>
 <ul>
 <li><code><a href="https://rdrr.io/r/base/diff.html">diff()</a></code></li>
 </ul>
@@ -437,8 +437,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression        min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;   &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 diff(v1, 2)    3.78µs   4.44µs   218299.        0B</span></span>
-<span><span class="co">#&gt; 2 diffC(v1, 2)   1.91µs   2.12µs   450551.    2.49KB</span></span>
+<span><span class="co">#&gt; 1 diff(v1, 2)    3.88µs   4.46µs   217285.        0B</span></span>
+<span><span class="co">#&gt; 2 diffC(v1, 2)   1.89µs   2.13µs   452394.    2.49KB</span></span>
 <span><span class="co">#&gt;   `gc/sec`</span></span>
 <span><span class="co">#&gt;      &lt;dbl&gt;</span></span>
 <span><span class="co">#&gt; 1        0</span></span>
@@ -478,8 +478,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression      min   median `itr/sec` mem_alloc `gc/sec`</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt; &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;    &lt;dbl&gt;</span></span>
-<span><span class="co">#&gt; 1 range(v1)    1.64µs   1.81µs   526003.        0B        0</span></span>
-<span><span class="co">#&gt; 2 rangeC(v1)   1.51µs   1.66µs   482154.    6.62KB        0</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 range(v1)    1.68µs   1.94µs   481017.        0B        0</span></span>
+<span><span class="co">#&gt; 2 rangeC(v1)   1.47µs   1.68µs   569929.    6.62KB        0</span></span></code></pre></div>
 <ul>
 <li><code><a href="https://rdrr.io/r/stats/cor.html">var()</a></code></li>
 </ul>
@@ -520,8 +520,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression        min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;   &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 var(v1)         7.4µs   8.31µs   117887.        0B</span></span>
-<span><span class="co">#&gt; 2 variance(v1)   1.46µs   1.61µs   587188.    6.62KB</span></span>
+<span><span class="co">#&gt; 1 var(v1)        7.54µs   8.51µs   111092.        0B</span></span>
+<span><span class="co">#&gt; 2 variance(v1)   1.49µs   1.66µs   567349.    6.62KB</span></span>
 <span><span class="co">#&gt;   `gc/sec`</span></span>
 <span><span class="co">#&gt;      &lt;dbl&gt;</span></span>
 <span><span class="co">#&gt; 1        0</span></span>
@@ -680,8 +680,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression              min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;         &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 median.default(v2)  21.91µs  23.52µs    40942.        0B</span></span>
-<span><span class="co">#&gt; 2 medianC(v2)          1.45µs   1.67µs   565586.    2.49KB</span></span>
+<span><span class="co">#&gt; 1 median.default(v2)  22.17µs  24.29µs    40480.        0B</span></span>
+<span><span class="co">#&gt; 2 medianC(v2)          1.49µs   1.73µs   546971.    2.49KB</span></span>
 <span><span class="co">#&gt;   `gc/sec`</span></span>
 <span><span class="co">#&gt;      &lt;dbl&gt;</span></span>
 <span><span class="co">#&gt; 1        0</span></span>
@@ -726,8 +726,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression          min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;     &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 x1 %in% x2     881.03ns   1.11µs   703925.        0B</span></span>
-<span><span class="co">#&gt; 2 matchC(x1, x2)   2.07µs   2.35µs   409169.    6.62KB</span></span>
+<span><span class="co">#&gt; 1 x1 %in% x2     881.03ns   1.11µs   870819.        0B</span></span>
+<span><span class="co">#&gt; 2 matchC(x1, x2)   2.15µs   2.41µs   399261.    6.62KB</span></span>
 <span><span class="co">#&gt;   `gc/sec`</span></span>
 <span><span class="co">#&gt;      &lt;dbl&gt;</span></span>
 <span><span class="co">#&gt; 1        0</span></span>
@@ -788,8 +788,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression        min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;   &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 unique(v1)     2.13µs   2.43µs   343230.        0B</span></span>
-<span><span class="co">#&gt; 2 uniqueC2(v1)   1.69µs   1.85µs   512921.    6.62KB</span></span>
+<span><span class="co">#&gt; 1 unique(v1)      2.1µs   2.41µs   355209.        0B</span></span>
+<span><span class="co">#&gt; 2 uniqueC2(v1)   1.63µs   1.85µs   502571.    6.62KB</span></span>
 <span><span class="co">#&gt;   `gc/sec`</span></span>
 <span><span class="co">#&gt;      &lt;dbl&gt;</span></span>
 <span><span class="co">#&gt; 1        0</span></span>
@@ -832,8 +832,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression      min   median `itr/sec` mem_alloc `gc/sec`</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt; &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;    &lt;dbl&gt;</span></span>
-<span><span class="co">#&gt; 1 min(v1)    240.05ns 251.11ns  3549516.        0B        0</span></span>
-<span><span class="co">#&gt; 2 minC(v1)     1.39µs   1.51µs   612452.    6.62KB        0</span></span>
+<span><span class="co">#&gt; 1 min(v1)    230.04ns 250.93ns  3402682.        0B        0</span></span>
+<span><span class="co">#&gt; 2 minC(v1)     1.42µs   1.56µs   606873.    6.62KB        0</span></span>
 <span></span>
 <span><span class="fu"><a href="https://rdrr.io/r/base/Extremes.html">max</a></span><span class="op">(</span><span class="va">v1</span><span class="op">)</span></span>
 <span><span class="co">#&gt; [1] 9</span></span>
@@ -849,8 +849,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression      min   median `itr/sec` mem_alloc `gc/sec`</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt; &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;    &lt;dbl&gt;</span></span>
-<span><span class="co">#&gt; 1 max(v1)    220.03ns  250.1ns  3345068.        0B        0</span></span>
-<span><span class="co">#&gt; 2 maxC(v1)     1.39µs    1.6µs   595400.    6.62KB        0</span></span></code></pre></div>
+<span><span class="co">#&gt; 1 max(v1)    211.06ns  241.1ns  3455199.        0B        0</span></span>
+<span><span class="co">#&gt; 2 maxC(v1)     1.41µs   1.66µs   571011.    6.62KB        0</span></span></code></pre></div>
 <ol style="list-style-type: decimal">
 <li>
 <code><a href="https://rdrr.io/r/base/which.min.html">which.min()</a></code> using <code>min_element</code>, or <code><a href="https://rdrr.io/r/base/which.min.html">which.max()</a></code> using <code>max_element</code>.</li>
@@ -894,8 +894,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression          min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;     &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 which.min(v1)   401.1ns 431.09ns  2009835.        0B</span></span>
-<span><span class="co">#&gt; 2 which_minC(v1)    1.4µs   1.62µs   589790.    6.62KB</span></span>
+<span><span class="co">#&gt; 1 which.min(v1)  401.05ns  421.1ns  2181122.        0B</span></span>
+<span><span class="co">#&gt; 2 which_minC(v1)   1.44µs    1.6µs   591193.    6.62KB</span></span>
 <span><span class="co">#&gt;   `gc/sec`</span></span>
 <span><span class="co">#&gt;      &lt;dbl&gt;</span></span>
 <span><span class="co">#&gt; 1        0</span></span>
@@ -915,8 +915,8 @@ <h2>
 <span><span class="co">#&gt; # A tibble: 2 × 6</span></span>
 <span><span class="co">#&gt;   expression          min   median `itr/sec` mem_alloc</span></span>
 <span><span class="co">#&gt;   &lt;bch:expr&gt;     &lt;bch:tm&gt; &lt;bch:tm&gt;     &lt;dbl&gt; &lt;bch:byt&gt;</span></span>
-<span><span class="co">#&gt; 1 which.max(v1)  390.11ns 461.12ns  1940961.        0B</span></span>
-<span><span class="co">#&gt; 2 which_maxC(v1)   1.39µs   1.55µs   608448.    6.62KB</span></span>
+<span><span class="co">#&gt; 1 which.max(v1)  410.01ns 490.92ns  1870056.        0B</span></span>
+<span><span class="co">#&gt; 2 which_maxC(v1)   1.41µs   1.57µs   585441.    6.62KB</span></span>
 <span><span class="co">#&gt;   `gc/sec`</span></span>
 <span><span class="co">#&gt;      &lt;dbl&gt;</span></span>
 <span><span class="co">#&gt; 1        0</span></span>
@@ -1039,7 +1039,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -1138,7 +1138,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/s3.html b/s3.html
index 97b04087..24d37b1a 100644
--- a/s3.html
+++ b/s3.html
@@ -512,7 +512,7 @@ <h2>
 <code class="sourceCode R"><span><span class="va">t</span></span>
 <span><span class="co">#&gt; function (x) </span></span>
 <span><span class="co">#&gt; UseMethod("t")</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x56246846b798&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x56119bea5208&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:base&gt;</span></span>
 <span><span class="fu">sloop</span><span class="fu">::</span><span class="fu"><a href="https://sloop.r-lib.org/reference/is_s3_generic.html">is_s3_generic</a></span><span class="op">(</span><span class="st">"t"</span><span class="op">)</span></span>
 <span><span class="co">#&gt; [1] TRUE</span></span>
@@ -520,7 +520,7 @@ <h2>
 <span><span class="va">t.test</span></span>
 <span><span class="co">#&gt; function (x, ...) </span></span>
 <span><span class="co">#&gt; UseMethod("t.test")</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x562463507140&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x561196f41140&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:stats&gt;</span></span>
 <span><span class="fu">sloop</span><span class="fu">::</span><span class="fu"><a href="https://sloop.r-lib.org/reference/is_s3_generic.html">is_s3_generic</a></span><span class="op">(</span><span class="st">"t.test"</span><span class="op">)</span></span>
 <span><span class="co">#&gt; [1] TRUE</span></span></code></pre></div>
@@ -848,14 +848,14 @@ <h2>
 <span><span class="co">#&gt; {</span></span>
 <span><span class="co">#&gt;     .Date(NextMethod("["), oldClass(x))</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x5624687fc3c0&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x56119c235db0&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:base&gt;</span></span></code></pre></div>
 <p>The <code>.Date</code> function looks like this:</p>
 <div class="sourceCode" id="cb402"><pre class="downlit sourceCode r">
 <code class="sourceCode R"><span><span class="va">.Date</span></span>
 <span><span class="co">#&gt; function (xx, cl = "Date") </span></span>
 <span><span class="co">#&gt; `class&lt;-`(xx, cl)</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x562466349c10&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x561199d84c50&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:base&gt;</span></span></code></pre></div>
 <p>Here, <code>oldClass</code> is the same as <code><a href="https://rdrr.io/r/base/class.html">class()</a></code>.</p>
 <p>Therefore, by reading this code, we can surmise that:</p>
@@ -875,7 +875,7 @@ <h2>
 <span><span class="co">#&gt; [1] "subDate" "Date"</span></span>
 <span></span>
 <span><span class="va">x</span><span class="op">[</span><span class="fl">1</span><span class="op">]</span></span>
-<span><span class="co">#&gt; [1] "2023-12-31"</span></span></code></pre></div>
+<span><span class="co">#&gt; [1] "2024-01-07"</span></span></code></pre></div>
 <p><strong>Q2.</strong> R has two classes for representing date time data, <code>POSIXct</code> and <code>POSIXlt</code>, which both inherit from <code>POSIXt</code>. Which generics have different behaviours for the two classes? Which generics share the same behaviour?</p>
 <p><strong>A2.</strong> First, let’s demonstrate that <code>POSIXct</code> and <code>POSIXlt</code> are indeed subclasses and <code>POSIXt</code> is the superclass.</p>
 <div class="sourceCode" id="cb404"><pre class="downlit sourceCode r">
@@ -1166,7 +1166,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -1260,7 +1260,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/s4.html b/s4.html
index f1c451c9..538fa606 100644
--- a/s4.html
+++ b/s4.html
@@ -582,7 +582,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -672,7 +672,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/search.json b/search.json
index fc0160dc..b314223f 100644
--- a/search.json
+++ b/search.json
@@ -1 +1 @@
-[{"path":"index.html","id":"about","chapter":"About","heading":"About","text":"book provides solutions exercises Hadley Wickham’s Advanced R (2nd edition) book.started working book part process learn solving book’s exercises. comparing solutions official solutions manual, realized solutions took different approaches least explained differently. ’m sharing solutions case others might find another perspective explanation official solution manual helpful building understanding.Although tried make sure solutions correct, blame inaccuracies lies solely . ’d much appreciate suggestions corrections.","code":""},{"path":"introduction.html","id":"introduction","chapter":"1 Introduction","heading":"1 Introduction","text":"exercises.","code":""},{"path":"names-and-values.html","id":"names-and-values","chapter":"2 Names and values","heading":"2 Names and values","text":"Loading needed libraries:","code":"\nlibrary(lobstr)"},{"path":"names-and-values.html","id":"binding-basics-exercise-2.2.2","chapter":"2 Names and values","heading":"2.1 Binding basics (Exercise 2.2.2)","text":"Q1. Explain relationship , b, c d following code:A1. names (, b, c) values point object memory, can seen identical memory addresses:Except d, different object, even value , b, c:Q2. following code accesses mean function multiple ways. point underlying function object? Verify lobstr::obj_addr().A2. listed function calls point underlying function object memory, shown object’s memory address:Q3. default, base R data import functions, like read.csv(), automatically convert non-syntactic names syntactic ones. might problematic? option allows suppress behaviour?A3. conversion non-syntactic names syntactic ones can sometimes corrupt data. datasets may require non-syntactic names.suppress behavior, one can set check.names = FALSE.Q4. rules make.names() use convert non-syntactic names syntactic ones?A4. make.names() uses following rules convert non-syntactic names syntactic ones:prepends non-syntactic names Xit converts invalid characters (like @) .adds . suffix name reserved keywordQ5. slightly simplified rules govern syntactic names. .123e1 syntactic name? Read ?make.names full details.A5. .123e1 syntacti name parsed number, string:docs mention (emphasis mine):syntactically valid name consists letters, numbers dot underline characters starts letter dot followed number.","code":"\na <- 1:10\nb <- a\nc <- b\nd <- 1:10\nobj_addrs <- obj_addrs(list(a, b, c))\nunique(obj_addrs)\n#> [1] \"0x556dced402c8\"\nobj_addr(d)\n#> [1] \"0x556dce9f9fc8\"\nmean\nbase::mean\nget(\"mean\")\nevalq(mean)\nmatch.fun(\"mean\")\nobj_addrs <- obj_addrs(list(\n  mean,\n  base::mean,\n  get(\"mean\"),\n  evalq(mean),\n  match.fun(\"mean\")\n))\n\nunique(obj_addrs)\n#> [1] \"0x556dcc722500\"\nmake.names(c(\"123abc\", \"@me\", \"_yu\", \"  gh\", \"else\"))\n#> [1] \"X123abc\" \"X.me\"    \"X_yu\"    \"X..gh\"   \"else.\"\ntypeof(.123e1)\n#> [1] \"double\""},{"path":"names-and-values.html","id":"copy-on-modify-exercise-2.3.6","chapter":"2 Names and values","heading":"2.2 Copy-on-modify (Exercise 2.3.6)","text":"Q1. tracemem(1:10) useful?A1. tracemem() traces copying objects R. example:since object created memory 1:10 assigned name, can’t addressed modified R, nothing trace.Q2. Explain tracemem() shows two copies run code. Hint: carefully look difference code code shown earlier section.A2. initial atomic vector type integer, 4 (4L) type double. new copy created.Trying integer create another copy:understand still produces copy, explanation official solutions manual:Please aware running code RStudio result additional copies reference environment pane.Q3. Sketch relationship following objects:A3. can understand relationship objects looking memory addresses:learn:name references object 1:10 memory.name b bound list two references memory address .name c also bound list references b, 1:10 object (bound name).Q4. happens run code?Draw picture.A4.don’t access OmniGraffle software, including figure official solution manual:","code":"\nx <- 1:10\n\ntracemem(x)\n#> [1] \"<0x556dd0825628>\"\n\nx <- x + 1\n\nuntracemem(x)\nobj_addr(1:10)\n#> [1] \"0x556dd02fa0f8\"\n\ntracemem(1:10)\n#> [1] \"<0x556dd03667a0>\"\nx <- c(1L, 2L, 3L)\ntracemem(x)\n\nx[[3]] <- 4\nuntracemem(x)\nx <- c(1L, 2L, 3L)\ntypeof(x)\n#> [1] \"integer\"\ntracemem(x)\n#> [1] \"<0x556dd0baaad8>\"\n\nx[[3]] <- 4\n#> tracemem[0x556dd0baaad8 -> 0x556dd0cd6558]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local \n#> tracemem[0x556dd0cd6558 -> 0x556dd0ce41d8]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local\nuntracemem(x)\n\ntypeof(x)\n#> [1] \"double\"\nx <- c(1L, 2L, 3L)\ntypeof(x)\n#> [1] \"integer\"\ntracemem(x)\n#> [1] \"<0x556dd11a3198>\"\n\nx[[3]] <- 4L\n#> tracemem[0x556dd11a3198 -> 0x556dd12cee18]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local\nuntracemem(x)\n\ntypeof(x)\n#> [1] \"integer\"\na <- 1:10\nb <- list(a, a)\nc <- list(b, a, 1:10)\na <- 1:10\nb <- list(a, a)\nc <- list(b, a, 1:10)\n\nref(a)\n#> [1:0x556dd198acc8] <int>\n\nref(b)\n#> █ [1:0x556dd1eb64c8] <list> \n#> ├─[2:0x556dd198acc8] <int> \n#> └─[2:0x556dd198acc8]\n\nref(c)\n#> █ [1:0x556dd1f3ee48] <list> \n#> ├─█ [2:0x556dd1eb64c8] <list> \n#> │ ├─[3:0x556dd198acc8] <int> \n#> │ └─[3:0x556dd198acc8] \n#> ├─[3:0x556dd198acc8] \n#> └─[4:0x556dd1f41178] <int>\nx <- list(1:10)\nx[[2]] <- x\nx <- list(1:10)\nx\n#> [[1]]\n#>  [1]  1  2  3  4  5  6  7  8  9 10\nobj_addr(x)\n#> [1] \"0x556dcfcf8e40\"\n\nx[[2]] <- x\nx\n#> [[1]]\n#>  [1]  1  2  3  4  5  6  7  8  9 10\n#> \n#> [[2]]\n#> [[2]][[1]]\n#>  [1]  1  2  3  4  5  6  7  8  9 10\nobj_addr(x)\n#> [1] \"0x556dd04f3148\"\n\nref(x)\n#> █ [1:0x556dd04f3148] <list> \n#> ├─[2:0x556dcf554b38] <int> \n#> └─█ [3:0x556dcfcf8e40] <list> \n#>   └─[2:0x556dcf554b38]"},{"path":"names-and-values.html","id":"object-size-exercise-2.4.1","chapter":"2 Names and values","heading":"2.3 Object size (Exercise 2.4.1)","text":"Q1. following example, object.size(y) obj_size(y) radically different? Consult documentation object.size().A1. mentioned docs object.size():function…detect elements list shared.sizes different:Q2. Take following list. size somewhat misleading?A2. functions externally created objects R, always available part base packages, doesn’t make much sense measure size never going available.Q3. Predict output following code:A3. Correctly predicted 😉Key pieces information keep mind make correct predictions:Size empty vectorSize single double: 8 bytesCopy--modify semantics","code":"\ny <- rep(list(runif(1e4)), 100)\n\nobject.size(y)\nobj_size(y)\ny <- rep(list(runif(1e4)), 100)\n\nobject.size(y)\n#> 8005648 bytes\n\nobj_size(y)\n#> 80.90 kB\nfuns <- list(mean, sd, var)\nobj_size(funs)\nfuns <- list(mean, sd, var)\nobj_size(funs)\n#> 17.55 kB\na <- runif(1e6)\nobj_size(a)\n\nb <- list(a, a)\nobj_size(b)\nobj_size(a, b)\n\nb[[1]][[1]] <- 10\nobj_size(b)\nobj_size(a, b)\n\nb[[2]][[1]] <- 10\nobj_size(b)\nobj_size(a, b)\na <- runif(1e6)\nobj_size(a)\n#> 8.00 MB\n\nb <- list(a, a)\nobj_size(b)\n#> 8.00 MB\nobj_size(a, b)\n#> 8.00 MB\n\nb[[1]][[1]] <- 10\nobj_size(b)\n#> 16.00 MB\nobj_size(a, b)\n#> 16.00 MB\n\nb[[2]][[1]] <- 10\nobj_size(b)\n#> 16.00 MB\nobj_size(a, b)\n#> 24.00 MB\nobj_size(double())\n#> 48 B\nobj_size(double(1))\n#> 56 B"},{"path":"names-and-values.html","id":"modify-in-place-exercise-2.5.3","chapter":"2 Names and values","heading":"2.4 Modify-in-place (Exercise 2.5.3)","text":"Q1. Explain following code doesn’t create circular list.A1. Copy--modify prevents creation circular list.Q2. Wrap two methods subtracting medians two functions, use ‘bench’ package carefully compare speeds. performance change number columns increase?A2. Let’s first microbenchmark functions create copies varying lengths number columns.Plotting benchmarks reveals performance gets increasingly worse number data frames increases:Q3. happens attempt use tracemem() environment?A3. doesn’t work documentation tracemem() makes clear :useful trace NULL, environments, promises, weak references, external pointer objects, duplicated","code":"\nx <- list()\nx[[1]] <- x\nx <- list()\n\nobj_addr(x)\n#> [1] \"0x556dd095be30\"\n\ntracemem(x)\n#> [1] \"<0x556dd095be30>\"\n\nx[[1]] <- x\n#> tracemem[0x556dd095be30 -> 0x556dd02523e8]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local\n\nobj_addr(x[[1]])\n#> [1] \"0x556dd095be30\"\nlibrary(bench)\nlibrary(tidyverse)\n\ngenerateDataFrame <- function(ncol) {\n  as.data.frame(matrix(runif(100 * ncol), nrow = 100))\n}\n\nwithCopy <- function(ncol) {\n  x <- generateDataFrame(ncol)\n  medians <- vapply(x, median, numeric(1))\n\n  for (i in seq_along(medians)) {\n    x[[i]] <- x[[i]] - medians[[i]]\n  }\n\n  return(x)\n}\n\nwithoutCopy <- function(ncol) {\n  x <- generateDataFrame(ncol)\n  medians <- vapply(x, median, numeric(1))\n\n  y <- as.list(x)\n\n  for (i in seq_along(medians)) {\n    y[[i]] <- y[[i]] - medians[[i]]\n  }\n\n  return(y)\n}\n\nbenchComparison <- function(ncol) {\n  bench::mark(\n    withCopy(ncol),\n    withoutCopy(ncol),\n    iterations = 100,\n    check = FALSE\n  ) %>%\n    dplyr::select(expression:total_time)\n}\n\nnColList <- list(1, 10, 50, 100, 250, 500, 1000)\n\nnames(nColList) <- as.character(nColList)\n\nbenchDf <- purrr::map_dfr(\n  .x = nColList,\n  .f = benchComparison,\n  .id = \"nColumns\"\n)\nggplot(\n  benchDf,\n  aes(\n    x = as.numeric(nColumns),\n    y = median,\n    group = as.character(expression),\n    color = as.character(expression)\n  )\n) +\n  geom_line() +\n  labs(\n    x = \"Number of Columns\",\n    y = \"Median Execution Time (ms)\",\n    colour = \"Type of function\"\n  )\ne <- rlang::env(a = 1, b = \"3\")\ntracemem(e)\n#> Error in tracemem(e): 'tracemem' is not useful for promise and environment objects"},{"path":"names-and-values.html","id":"session-information","chapter":"2 Names and values","heading":"2.5 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bench       * 1.1.3   2023-05-04 [1] RSPM\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  colorspace    2.1-0   2023-01-23 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  crayon        1.5.2   2022-09-29 [1] RSPM\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr       * 1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  farver        2.1.1   2022-07-06 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  forcats     * 1.0.0   2023-01-29 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  ggplot2     * 3.4.4   2023-10-12 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  grid          4.3.2   2023-11-01 [3] local\n#>  gtable        0.3.4   2023-08-21 [1] RSPM\n#>  highr         0.10    2022-12-22 [1] RSPM\n#>  hms           1.1.3   2023-03-21 [1] RSPM\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  labeling      0.4.3   2023-08-29 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  lobstr      * 1.1.2   2022-06-22 [1] RSPM\n#>  lubridate   * 1.9.3   2023-09-27 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  munsell       0.5.0   2018-06-12 [1] RSPM\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  prettyunits   1.2.0   2023-09-24 [1] RSPM\n#>  profmem       0.6.0   2020-12-13 [1] RSPM\n#>  purrr       * 1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  readr       * 2.1.4   2023-02-10 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  scales        1.3.0   2023-11-28 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  stringi       1.8.3   2023-12-11 [1] RSPM\n#>  stringr     * 1.5.1   2023-11-14 [1] RSPM\n#>  tibble      * 3.2.1   2023-03-20 [1] RSPM\n#>  tidyr       * 1.3.0   2023-01-24 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tidyverse   * 2.0.0   2023-02-22 [1] RSPM\n#>  timechange    0.2.0   2023-01-11 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  tzdb          0.4.0   2023-05-12 [1] RSPM\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"vectors.html","id":"vectors","chapter":"3 Vectors","heading":"3 Vectors","text":"","code":""},{"path":"vectors.html","id":"atomic-vectors-exercises-3.2.5","chapter":"3 Vectors","heading":"3.1 Atomic vectors (Exercises 3.2.5)","text":"Q1. create raw complex scalars? (See ?raw ?complex.)A1. R, scalars nothing vectors length 1, can created using constructor.Raw vectorsThe raw type holds raw bytes, can created using charToRaw(). example,alternative use .raw():Complex vectorsComplex vectors used represent (surprise!) complex numbers.Example complex scalar:Q2. Test knowledge vector coercion rules predicting output following uses c():A2. vector coercion rules dictate data type smaller size converted data type bigger size.Q3. 1 == \"1\" true? -1 < FALSE true? \"one\" < 2 false?A3. coercion rules vectors reveal comparisons return results .Q4. default missing value, NA, logical vector? ’s special logical vectors? (Hint: think c(FALSE, NA_character_).)A4. \"logical\" type lowest coercion hierarchy.NA defaulting type (e.g. \"numeric\") mean time missing element vector, rest elements converted type higher hierarchy, problematic types lower hierarchy.Q5. Precisely .atomic(), .numeric(), .vector() test ?A5. Let’s discuss one--one..atomic()function checks object vector atomic type (NULL).Quoting docs:.atomic true atomic types (“logical”, “integer”, “numeric”, “complex”, “character” “raw”) NULL..numeric()documentation says:.numeric return true base type class double integer values can reasonably regarded numericTherefore, function checks double integer base types types based top types (factor, Date, POSIXt, difftime)..vector()per documentation:.vector returns TRUE x vector specified mode attributes names. returns FALSE otherwise.Thus, function can incorrectif object attributes names.better way check vector:","code":"\nx <- \"A string\"\n\n(y <- charToRaw(x))\n#> [1] 41 20 73 74 72 69 6e 67\n\ntypeof(y)\n#> [1] \"raw\"\nas.raw(\"–\") # en-dash\n#> Warning: NAs introduced by coercion\n#> Warning: out-of-range values treated as 0 in coercion to\n#> raw\n#> [1] 00\nas.raw(\"—\") # em-dash\n#> Warning: NAs introduced by coercion\n\n#> Warning: out-of-range values treated as 0 in coercion to\n#> raw\n#> [1] 00\n(x <- complex(length.out = 1, real = 1, imaginary = 8))\n#> [1] 1+8i\n\ntypeof(x)\n#> [1] \"complex\"\nc(1, FALSE)\nc(\"a\", 1)\nc(TRUE, 1L)\nc(1, FALSE)\n#> [1] 1 0\n\nc(\"a\", 1)\n#> [1] \"a\" \"1\"\n\nc(TRUE, 1L)\n#> [1] 1 1\n1 == \"1\"\n#> [1] TRUE\n\nc(1, \"1\")\n#> [1] \"1\" \"1\"\n-1 < FALSE\n#> [1] TRUE\n\nc(-1, FALSE)\n#> [1] -1  0\n\"one\" < 2\n#> [1] FALSE\n\nc(\"one\", 2)\n#> [1] \"one\" \"2\"\n\nsort(c(\"one\", 2))\n#> [1] \"2\"   \"one\"\ntypeof(NA)\n#> [1] \"logical\"\n\nc(FALSE, NA_character_)\n#> [1] \"FALSE\" NA\nis.atomic(NULL)\n#> [1] TRUE\n\nis.atomic(list(NULL))\n#> [1] FALSE\nis.numeric(1L)\n#> [1] TRUE\n\nis.numeric(factor(1L))\n#> [1] FALSE\nx <- c(\"x\" = 1, \"y\" = 2)\n\nis.vector(x)\n#> [1] TRUE\n\nattr(x, \"m\") <- \"abcdef\"\n\nis.vector(x)\n#> [1] FALSE\nis.null(dim(x))\n#> [1] TRUE"},{"path":"vectors.html","id":"attributes-exercises-3.3.4","chapter":"3 Vectors","heading":"3.2 Attributes (Exercises 3.3.4)","text":"Q1. setNames() implemented? unname() implemented? Read source code.A1. Let’s look implementations functions.setNames()Given function signature, can see , first argument given, result still named vector.unname()unname() removes existing names (dimnames) setting NULL.Q2. dim() return applied 1-dimensional vector? might use NROW() NCOL()?A2. Dimensions 1-dimensional vector NULL. example,NROW() NCOL() helpful getting dimensions 1D vectors treating matrices dataframes.Q3. describe following three objects? makes different 1:5?A3. x1, x2, x3 one-dimensional arrays, different “orientations”, mentally visualize .x1 5 entries third dimension, x2 second dimension, x1 first dimension.Q4. early draft used code illustrate structure():print object don’t see comment attribute. ? attribute missing, something else special ? (Hint: try using help.)A4. ?attributes (emphasis mine):Note attributes (namely class, comment, dim, dimnames, names, row.names tsp) treated specially restrictions values can set.","code":"\nsetNames\n#> function (object = nm, nm) \n#> {\n#>     names(object) <- nm\n#>     object\n#> }\n#> <bytecode: 0x5635a7f8a880>\n#> <environment: namespace:stats>\nsetNames(, c(\"a\", \"b\"))\n#>   a   b \n#> \"a\" \"b\"\n\nsetNames(c(1, 2), c(\"a\", \"b\"))\n#> a b \n#> 1 2\nunname\n#> function (obj, force = FALSE) \n#> {\n#>     if (!is.null(names(obj))) \n#>         names(obj) <- NULL\n#>     if (!is.null(dimnames(obj)) && (force || !is.data.frame(obj))) \n#>         dimnames(obj) <- NULL\n#>     obj\n#> }\n#> <bytecode: 0x5635a7395578>\n#> <environment: namespace:base>\nunname(setNames(, c(\"a\", \"b\")))\n#> [1] \"a\" \"b\"\ndim(c(1, 2))\n#> NULL\n# example-1\nx <- character(0)\n\ndim(x)\n#> NULL\n\nnrow(x)\n#> NULL\nNROW(x)\n#> [1] 0\n\nncol(x)\n#> NULL\nNCOL(x)\n#> [1] 1\n\n# example-2\ny <- 1:4\n\ndim(y)\n#> NULL\n\nnrow(y)\n#> NULL\nNROW(y)\n#> [1] 4\n\nncol(y)\n#> NULL\nNCOL(y)\n#> [1] 1\nx1 <- array(1:5, c(1, 1, 5))\nx2 <- array(1:5, c(1, 5, 1))\nx3 <- array(1:5, c(5, 1, 1))\nstructure(1:5, comment = \"my attribute\")\n#> [1] 1 2 3 4 5\nstructure(1:5, x = \"my attribute\")\n#> [1] 1 2 3 4 5\n#> attr(,\"x\")\n#> [1] \"my attribute\"\n\nstructure(1:5, comment = \"my attribute\")\n#> [1] 1 2 3 4 5"},{"path":"vectors.html","id":"s3-atomic-vectors-exercises-3.4.5","chapter":"3 Vectors","heading":"3.3 S3 atomic vectors (Exercises 3.4.5)","text":"Q1. sort object table() return? type? attributes ? dimensionality change tabulate variables?A1. table() returns array integer type dimensions scale number variables present.Q2. happens factor modify levels?A2. levels change underlying integer values remain .Q3. code ? f2 f3 differ f1?A3. code:f2: underlying integers reversed, levels remain unchanged.f3: levels underlying integers reversed.","code":"\n(x <- table(mtcars$am))\n#> \n#>  0  1 \n#> 19 13\n(y <- table(mtcars$am, mtcars$cyl))\n#>    \n#>      4  6  8\n#>   0  3  4 12\n#>   1  8  3  2\n(z <- table(mtcars$am, mtcars$cyl, mtcars$vs))\n#> , ,  = 0\n#> \n#>    \n#>      4  6  8\n#>   0  0  0 12\n#>   1  1  3  2\n#> \n#> , ,  = 1\n#> \n#>    \n#>      4  6  8\n#>   0  3  4  0\n#>   1  7  0  0\n\n# type\npurrr::map(list(x, y, z), typeof)\n#> [[1]]\n#> [1] \"integer\"\n#> \n#> [[2]]\n#> [1] \"integer\"\n#> \n#> [[3]]\n#> [1] \"integer\"\n\n# attributes\npurrr::map(list(x, y, z), attributes)\n#> [[1]]\n#> [[1]]$dim\n#> [1] 2\n#> \n#> [[1]]$dimnames\n#> [[1]]$dimnames[[1]]\n#> [1] \"0\" \"1\"\n#> \n#> \n#> [[1]]$class\n#> [1] \"table\"\n#> \n#> \n#> [[2]]\n#> [[2]]$dim\n#> [1] 2 3\n#> \n#> [[2]]$dimnames\n#> [[2]]$dimnames[[1]]\n#> [1] \"0\" \"1\"\n#> \n#> [[2]]$dimnames[[2]]\n#> [1] \"4\" \"6\" \"8\"\n#> \n#> \n#> [[2]]$class\n#> [1] \"table\"\n#> \n#> \n#> [[3]]\n#> [[3]]$dim\n#> [1] 2 3 2\n#> \n#> [[3]]$dimnames\n#> [[3]]$dimnames[[1]]\n#> [1] \"0\" \"1\"\n#> \n#> [[3]]$dimnames[[2]]\n#> [1] \"4\" \"6\" \"8\"\n#> \n#> [[3]]$dimnames[[3]]\n#> [1] \"0\" \"1\"\n#> \n#> \n#> [[3]]$class\n#> [1] \"table\"\nf1 <- factor(letters)\nlevels(f1) <- rev(levels(f1))\nf1 <- factor(letters)\nf1\n#>  [1] a b c d e f g h i j k l m n o p q r s t u v w x y z\n#> 26 Levels: a b c d e f g h i j k l m n o p q r s t u ... z\nas.integer(f1)\n#>  [1]  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18\n#> [19] 19 20 21 22 23 24 25 26\n\nlevels(f1) <- rev(levels(f1))\nf1\n#>  [1] z y x w v u t s r q p o n m l k j i h g f e d c b a\n#> 26 Levels: z y x w v u t s r q p o n m l k j i h g f ... a\nas.integer(f1)\n#>  [1]  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18\n#> [19] 19 20 21 22 23 24 25 26\nf2 <- rev(factor(letters))\nf3 <- factor(letters, levels = rev(letters))\nf2 <- rev(factor(letters))\nf2\n#>  [1] z y x w v u t s r q p o n m l k j i h g f e d c b a\n#> 26 Levels: a b c d e f g h i j k l m n o p q r s t u ... z\nas.integer(f2)\n#>  [1] 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10  9\n#> [19]  8  7  6  5  4  3  2  1\nf3 <- factor(letters, levels = rev(letters))\nf3\n#>  [1] a b c d e f g h i j k l m n o p q r s t u v w x y z\n#> 26 Levels: z y x w v u t s r q p o n m l k j i h g f ... a\nas.integer(f3)\n#>  [1] 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10  9\n#> [19]  8  7  6  5  4  3  2  1"},{"path":"vectors.html","id":"lists-exercises-3.5.4","chapter":"3 Vectors","heading":"3.4 Lists (Exercises 3.5.4)","text":"Q1. List ways list differs atomic vector.A1. table comparison:Q2. need use unlist() convert list atomic vector? doesn’t .vector() work?A2. list already (generic) vector, .vector() going change anything, .atomic.vector. Thus, need use unlist().Q3. Compare contrast c() unlist() combining date date-time single vector.A3. Let’s first create date datetime objectAnd check attributes underlying double representation:Behavior c()Since S3 method c() dispatches first argument, resulting class vector going first argument. , attributes lost.Behavior unlist()removes attributes left underlying double representations objects.","code":"\nx <- list(a = 1, b = 2)\n\nis.vector(x)\n#> [1] TRUE\nis.atomic(x)\n#> [1] FALSE\n\n# still a list\nas.vector(x)\n#> $a\n#> [1] 1\n#> \n#> $b\n#> [1] 2\n\n# now a vector\nunlist(x)\n#> a b \n#> 1 2\ndate <- as.Date(\"1947-08-15\")\ndatetime <- as.POSIXct(\"1950-01-26 00:01\", tz = \"UTC\")\nattributes(date)\n#> $class\n#> [1] \"Date\"\nattributes(datetime)\n#> $class\n#> [1] \"POSIXct\" \"POSIXt\" \n#> \n#> $tzone\n#> [1] \"UTC\"\n\nas.double(date) # number of days since the Unix epoch 1970-01-01\n#> [1] -8175\nas.double(datetime) # number of seconds since then\n#> [1] -628991940\nc(date, datetime)\n#> [1] \"1947-08-15\" \"1950-01-26\"\n\nattributes(c(date, datetime))\n#> $class\n#> [1] \"Date\"\n\nc(datetime, date)\n#> [1] \"1950-01-26 00:01:00 UTC\" \"1947-08-15 00:00:00 UTC\"\n\nattributes(c(datetime, date))\n#> $class\n#> [1] \"POSIXct\" \"POSIXt\" \n#> \n#> $tzone\n#> [1] \"UTC\"\nunlist(list(date, datetime))\n#> [1]      -8175 -628991940\n\nunlist(list(datetime, date))\n#> [1] -628991940      -8175"},{"path":"vectors.html","id":"data-frames-and-tibbles-exercises-3.6.8","chapter":"3 Vectors","heading":"3.5 Data frames and tibbles (Exercises 3.6.8)","text":"Q1. Can data frame zero rows? zero columns?A1. Data frame 0 rows possible. basically list vector length 0.Data frame 0 columns also possible. empty list., finally, data frame 0 rows columns also possible:Although, might common create data frames, can results subsetting. example,Q2. happens attempt set rownames unique?A2. attempt set data frame rownames unique, work.Q3. df data frame, can say t(df), t(t(df))? Perform experiments, making sure try different column types.A3. Transposing data frame:transforms matrixcoerces elements typeQ4. .matrix() applied data frame columns different types? differ data.matrix()?A4. return type .matrix() depends data frame column types.docs .matrix() mention:method data frames return character matrix atomic columns non-(numeric/logical/complex) column, applying .vector factors format non-character columns. Otherwise usual coercion hierarchy (logical < integer < double < complex) used, e.g. -logical data frames coerced logical matrix, mixed logical-integer give integer matrix, etc.Let’s experiment:hand, data.matrix() always returns numeric matrix.documentation data.matrix():Return matrix obtained converting variables data frame numeric mode binding together columns matrix. Factors ordered factors replaced internal codes. […] Character columns first converted factors integers.Let’s experiment:","code":"\ndata.frame(x = numeric(0))\n#> [1] x\n#> <0 rows> (or 0-length row.names)\ndata.frame(row.names = 1)\n#> data frame with 0 columns and 1 row\ndata.frame()\n#> data frame with 0 columns and 0 rows\n\ndim(data.frame())\n#> [1] 0 0\nBOD[0, ]\n#> [1] Time   demand\n#> <0 rows> (or 0-length row.names)\n\nBOD[, 0]\n#> data frame with 0 columns and 6 rows\n\nBOD[0, 0]\n#> data frame with 0 columns and 0 rows\ndata.frame(row.names = c(1, 1))\n#> Error in data.frame(row.names = c(1, 1)): duplicate row.names: 1\n# original\n(df <- head(iris))\n#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species\n#> 1          5.1         3.5          1.4         0.2  setosa\n#> 2          4.9         3.0          1.4         0.2  setosa\n#> 3          4.7         3.2          1.3         0.2  setosa\n#> 4          4.6         3.1          1.5         0.2  setosa\n#> 5          5.0         3.6          1.4         0.2  setosa\n#> 6          5.4         3.9          1.7         0.4  setosa\n\n# transpose\nt(df)\n#>              1        2        3        4        5       \n#> Sepal.Length \"5.1\"    \"4.9\"    \"4.7\"    \"4.6\"    \"5.0\"   \n#> Sepal.Width  \"3.5\"    \"3.0\"    \"3.2\"    \"3.1\"    \"3.6\"   \n#> Petal.Length \"1.4\"    \"1.4\"    \"1.3\"    \"1.5\"    \"1.4\"   \n#> Petal.Width  \"0.2\"    \"0.2\"    \"0.2\"    \"0.2\"    \"0.2\"   \n#> Species      \"setosa\" \"setosa\" \"setosa\" \"setosa\" \"setosa\"\n#>              6       \n#> Sepal.Length \"5.4\"   \n#> Sepal.Width  \"3.9\"   \n#> Petal.Length \"1.7\"   \n#> Petal.Width  \"0.4\"   \n#> Species      \"setosa\"\n\n# transpose of a transpose\nt(t(df))\n#>   Sepal.Length Sepal.Width Petal.Length Petal.Width\n#> 1 \"5.1\"        \"3.5\"       \"1.4\"        \"0.2\"      \n#> 2 \"4.9\"        \"3.0\"       \"1.4\"        \"0.2\"      \n#> 3 \"4.7\"        \"3.2\"       \"1.3\"        \"0.2\"      \n#> 4 \"4.6\"        \"3.1\"       \"1.5\"        \"0.2\"      \n#> 5 \"5.0\"        \"3.6\"       \"1.4\"        \"0.2\"      \n#> 6 \"5.4\"        \"3.9\"       \"1.7\"        \"0.4\"      \n#>   Species \n#> 1 \"setosa\"\n#> 2 \"setosa\"\n#> 3 \"setosa\"\n#> 4 \"setosa\"\n#> 5 \"setosa\"\n#> 6 \"setosa\"\n\n# is it a dataframe?\nis.data.frame(df)\n#> [1] TRUE\nis.data.frame(t(df))\n#> [1] FALSE\nis.data.frame(t(t(df)))\n#> [1] FALSE\n\n# check type\ntypeof(df)\n#> [1] \"list\"\ntypeof(t(df))\n#> [1] \"character\"\ntypeof(t(t(df)))\n#> [1] \"character\"\n\n# check dimensions\ndim(df)\n#> [1] 6 5\ndim(t(df))\n#> [1] 5 6\ndim(t(t(df)))\n#> [1] 6 5\n# example with mixed types (coerced to character)\n(df <- head(iris))\n#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species\n#> 1          5.1         3.5          1.4         0.2  setosa\n#> 2          4.9         3.0          1.4         0.2  setosa\n#> 3          4.7         3.2          1.3         0.2  setosa\n#> 4          4.6         3.1          1.5         0.2  setosa\n#> 5          5.0         3.6          1.4         0.2  setosa\n#> 6          5.4         3.9          1.7         0.4  setosa\n\nas.matrix(df)\n#>   Sepal.Length Sepal.Width Petal.Length Petal.Width\n#> 1 \"5.1\"        \"3.5\"       \"1.4\"        \"0.2\"      \n#> 2 \"4.9\"        \"3.0\"       \"1.4\"        \"0.2\"      \n#> 3 \"4.7\"        \"3.2\"       \"1.3\"        \"0.2\"      \n#> 4 \"4.6\"        \"3.1\"       \"1.5\"        \"0.2\"      \n#> 5 \"5.0\"        \"3.6\"       \"1.4\"        \"0.2\"      \n#> 6 \"5.4\"        \"3.9\"       \"1.7\"        \"0.4\"      \n#>   Species \n#> 1 \"setosa\"\n#> 2 \"setosa\"\n#> 3 \"setosa\"\n#> 4 \"setosa\"\n#> 5 \"setosa\"\n#> 6 \"setosa\"\n\nstr(as.matrix(df))\n#>  chr [1:6, 1:5] \"5.1\" \"4.9\" \"4.7\" \"4.6\" \"5.0\" \"5.4\" ...\n#>  - attr(*, \"dimnames\")=List of 2\n#>   ..$ : chr [1:6] \"1\" \"2\" \"3\" \"4\" ...\n#>   ..$ : chr [1:5] \"Sepal.Length\" \"Sepal.Width\" \"Petal.Length\" \"Petal.Width\" ...\n\n# another example (no such coercion)\nBOD\n#>   Time demand\n#> 1    1    8.3\n#> 2    2   10.3\n#> 3    3   19.0\n#> 4    4   16.0\n#> 5    5   15.6\n#> 6    7   19.8\n\nas.matrix(BOD)\n#>      Time demand\n#> [1,]    1    8.3\n#> [2,]    2   10.3\n#> [3,]    3   19.0\n#> [4,]    4   16.0\n#> [5,]    5   15.6\n#> [6,]    7   19.8\ndata.matrix(df)\n#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species\n#> 1          5.1         3.5          1.4         0.2       1\n#> 2          4.9         3.0          1.4         0.2       1\n#> 3          4.7         3.2          1.3         0.2       1\n#> 4          4.6         3.1          1.5         0.2       1\n#> 5          5.0         3.6          1.4         0.2       1\n#> 6          5.4         3.9          1.7         0.4       1\n\nstr(data.matrix(df))\n#>  num [1:6, 1:5] 5.1 4.9 4.7 4.6 5 5.4 3.5 3 3.2 3.1 ...\n#>  - attr(*, \"dimnames\")=List of 2\n#>   ..$ : chr [1:6] \"1\" \"2\" \"3\" \"4\" ...\n#>   ..$ : chr [1:5] \"Sepal.Length\" \"Sepal.Width\" \"Petal.Length\" \"Petal.Width\" ..."},{"path":"vectors.html","id":"session-information-1","chapter":"3 Vectors","heading":"3.6 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  purrr         1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"subsetting.html","id":"subsetting","chapter":"4 Subsetting","heading":"4 Subsetting","text":"Attaching needed libraries:","code":"\nlibrary(tibble)"},{"path":"subsetting.html","id":"selecting-multiple-elements-exercises-4.2.6","chapter":"4 Subsetting","heading":"4.1 Selecting multiple elements (Exercises 4.2.6)","text":"Q1. Fix following common data frame subsetting errors:A1. Fixed versions commands:Q2. following code yield five missing values?A2. two reasons:default type NA R logical type.R recycles indexes match length vector.Q3. upper.tri() return? subsetting matrix work? need additional subsetting rules describe behaviour?A3. documentation upper.tri() states-Returns matrix logicals size given matrix entries TRUE upper triangleWhen used matrix subsetting, elements corresponding TRUE subsetting matrix selected. , instead matrix, returns vector:Q4. mtcars[1:20] return error? differ similar mtcars[1:20, ]?A4. indexed like list, data frame columns given indices selected.mtcars[1:20] doesn’t work 11 columns mtcars dataset.hand, mtcars[1:20, ] indexes dataframe like matrix, indeed 20 rows mtcars, columns rows selected.Q5. Implement function extracts diagonal entries matrix (behave like diag(x) x matrix).A5. can combine existing functions advantage:Q6. df[.na(df)] <- 0 ? work?A6. expression replaces every instance NA df 0..na(df) produces matrix logical values, provides way subsetting.","code":"mtcars[mtcars$cyl = 4, ]\nmtcars[-1:4, ]\nmtcars[mtcars$cyl <= 5]\nmtcars[mtcars$cyl == 4 | 6, ]\n# `==` instead of `=`\nmtcars[mtcars$cyl == 4, ]\n\n# `-(1:4)` instead of `-1:4`\nmtcars[-(1:4), ]\n\n# `,` was missing\nmtcars[mtcars$cyl <= 5, ]\n\n# correct subsetting syntax\nmtcars[mtcars$cyl == 4 | mtcars$cyl == 6, ]\nmtcars[mtcars$cyl %in% c(4, 6), ]\nx <- 1:5\nx[NA]\n#> [1] NA NA NA NA NA\ntypeof(NA)\n#> [1] \"logical\"\nx <- 1:5\nx[c(TRUE, FALSE)] # recycled to c(TRUE, FALSE, TRUE, FALSE, TRUE)\n#> [1] 1 3 5\nx <- outer(1:5, 1:5, FUN = \"*\")\nx[upper.tri(x)]\n(x <- outer(1:5, 1:5, FUN = \"*\"))\n#>      [,1] [,2] [,3] [,4] [,5]\n#> [1,]    1    2    3    4    5\n#> [2,]    2    4    6    8   10\n#> [3,]    3    6    9   12   15\n#> [4,]    4    8   12   16   20\n#> [5,]    5   10   15   20   25\n\nupper.tri(x)\n#>       [,1]  [,2]  [,3]  [,4]  [,5]\n#> [1,] FALSE  TRUE  TRUE  TRUE  TRUE\n#> [2,] FALSE FALSE  TRUE  TRUE  TRUE\n#> [3,] FALSE FALSE FALSE  TRUE  TRUE\n#> [4,] FALSE FALSE FALSE FALSE  TRUE\n#> [5,] FALSE FALSE FALSE FALSE FALSE\nx[upper.tri(x)]\n#>  [1]  2  3  6  4  8 12  5 10 15 20\nhead(mtcars[1:2])\n#>                    mpg cyl\n#> Mazda RX4         21.0   6\n#> Mazda RX4 Wag     21.0   6\n#> Datsun 710        22.8   4\n#> Hornet 4 Drive    21.4   6\n#> Hornet Sportabout 18.7   8\n#> Valiant           18.1   6\nnrow(mtcars[1:20, ])\n#> [1] 20\nx[!upper.tri(x) & !lower.tri(x)]\n#> [1]  1  4  9 16 25\n\ndiag(x)\n#> [1]  1  4  9 16 25\n(df <- tibble(x = c(1, 2, NA), y = c(NA, 5, NA)))\n#> # A tibble: 3 × 2\n#>       x     y\n#>   <dbl> <dbl>\n#> 1     1    NA\n#> 2     2     5\n#> 3    NA    NA\n\nis.na(df)\n#>          x     y\n#> [1,] FALSE  TRUE\n#> [2,] FALSE FALSE\n#> [3,]  TRUE  TRUE\n\nclass(is.na(df))\n#> [1] \"matrix\" \"array\""},{"path":"subsetting.html","id":"selecting-a-single-element-exercises-4.3.5","chapter":"4 Subsetting","heading":"4.2 Selecting a single element (Exercises 4.3.5)","text":"Q1. Brainstorm many ways possible extract third value cyl variable mtcars dataset.A1. Possible ways extract third value cyl variable mtcars dataset:Q2. Given linear model, e.g., mod <- lm(mpg ~ wt, data = mtcars), extract residual degrees freedom. extract R squared model summary (summary(mod))A2. Given objects class lm lists, can use subsetting operators extract elements want.extracting residual degrees freedomextracting R squared model summary","code":"\nmtcars[[\"cyl\"]][[3]]\n#> [1] 4\nmtcars[[c(2, 3)]]\n#> [1] 4\nmtcars[3, ][[\"cyl\"]]\n#> [1] 4\nmtcars[3, ]$cyl\n#> [1] 4\nmtcars[3, \"cyl\"]\n#> [1] 4\nmtcars[, \"cyl\"][[3]]\n#> [1] 4\nmtcars[3, 2]\n#> [1] 4\nmtcars$cyl[[3]]\n#> [1] 4\nmod <- lm(mpg ~ wt, data = mtcars)\nclass(mod)\n#> [1] \"lm\"\ntypeof(mod)\n#> [1] \"list\"\nmod$df.residual \n#> [1] 30\nmod[[\"df.residual\"]]\n#> [1] 30\nsummary(mod)$r.squared\n#> [1] 0.7528328\nsummary(mod)[[\"r.squared\"]]\n#> [1] 0.7528328"},{"path":"subsetting.html","id":"applications-exercises-4.5.9","chapter":"4 Subsetting","heading":"4.3 Applications (Exercises 4.5.9)","text":"Q1. randomly permute columns data frame? (important technique random forests.) Can simultaneously permute rows columns one step?A1. Let’s create small data frame work .randomly permute columns data frame, can combine [ sample() follows:randomly permute columnsrandomly permute rowsrandomly permute columns rowsQ2. select random sample m rows data frame? sample contiguous (.e., initial row, final row, every row )?A2. Let’s create small data frame work .select random sample m rows data frame, can combine [ sample() follows:random non-contiguous sample m rows data framerandom contiguous sample m rows data frameQ3. put columns data frame alphabetical order?A3. can sort columns data frame alphabetical order using [ order():","code":"\ndf <- head(mtcars)\n\n# original\ndf\n#>                    mpg cyl disp  hp drat    wt  qsec vs am\n#> Mazda RX4         21.0   6  160 110 3.90 2.620 16.46  0  1\n#> Mazda RX4 Wag     21.0   6  160 110 3.90 2.875 17.02  0  1\n#> Datsun 710        22.8   4  108  93 3.85 2.320 18.61  1  1\n#> Hornet 4 Drive    21.4   6  258 110 3.08 3.215 19.44  1  0\n#> Hornet Sportabout 18.7   8  360 175 3.15 3.440 17.02  0  0\n#> Valiant           18.1   6  225 105 2.76 3.460 20.22  1  0\n#>                   gear carb\n#> Mazda RX4            4    4\n#> Mazda RX4 Wag        4    4\n#> Datsun 710           4    1\n#> Hornet 4 Drive       3    1\n#> Hornet Sportabout    3    2\n#> Valiant              3    1\ndf[sample.int(ncol(df))]\n#>                   drat    wt carb am  qsec vs  hp  mpg disp\n#> Mazda RX4         3.90 2.620    4  1 16.46  0 110 21.0  160\n#> Mazda RX4 Wag     3.90 2.875    4  1 17.02  0 110 21.0  160\n#> Datsun 710        3.85 2.320    1  1 18.61  1  93 22.8  108\n#> Hornet 4 Drive    3.08 3.215    1  0 19.44  1 110 21.4  258\n#> Hornet Sportabout 3.15 3.440    2  0 17.02  0 175 18.7  360\n#> Valiant           2.76 3.460    1  0 20.22  1 105 18.1  225\n#>                   cyl gear\n#> Mazda RX4           6    4\n#> Mazda RX4 Wag       6    4\n#> Datsun 710          4    4\n#> Hornet 4 Drive      6    3\n#> Hornet Sportabout   8    3\n#> Valiant             6    3\ndf[sample.int(nrow(df)), ]\n#>                    mpg cyl disp  hp drat    wt  qsec vs am\n#> Datsun 710        22.8   4  108  93 3.85 2.320 18.61  1  1\n#> Mazda RX4 Wag     21.0   6  160 110 3.90 2.875 17.02  0  1\n#> Mazda RX4         21.0   6  160 110 3.90 2.620 16.46  0  1\n#> Hornet Sportabout 18.7   8  360 175 3.15 3.440 17.02  0  0\n#> Hornet 4 Drive    21.4   6  258 110 3.08 3.215 19.44  1  0\n#> Valiant           18.1   6  225 105 2.76 3.460 20.22  1  0\n#>                   gear carb\n#> Datsun 710           4    1\n#> Mazda RX4 Wag        4    4\n#> Mazda RX4            4    4\n#> Hornet Sportabout    3    2\n#> Hornet 4 Drive       3    1\n#> Valiant              3    1\ndf[sample.int(nrow(df)), sample.int(ncol(df))]\n#>                    qsec vs gear am    wt drat carb disp  hp\n#> Mazda RX4         16.46  0    4  1 2.620 3.90    4  160 110\n#> Hornet 4 Drive    19.44  1    3  0 3.215 3.08    1  258 110\n#> Datsun 710        18.61  1    4  1 2.320 3.85    1  108  93\n#> Mazda RX4 Wag     17.02  0    4  1 2.875 3.90    4  160 110\n#> Valiant           20.22  1    3  0 3.460 2.76    1  225 105\n#> Hornet Sportabout 17.02  0    3  0 3.440 3.15    2  360 175\n#>                    mpg cyl\n#> Mazda RX4         21.0   6\n#> Hornet 4 Drive    21.4   6\n#> Datsun 710        22.8   4\n#> Mazda RX4 Wag     21.0   6\n#> Valiant           18.1   6\n#> Hornet Sportabout 18.7   8\ndf <- head(mtcars)\n\n# original\ndf\n#>                    mpg cyl disp  hp drat    wt  qsec vs am\n#> Mazda RX4         21.0   6  160 110 3.90 2.620 16.46  0  1\n#> Mazda RX4 Wag     21.0   6  160 110 3.90 2.875 17.02  0  1\n#> Datsun 710        22.8   4  108  93 3.85 2.320 18.61  1  1\n#> Hornet 4 Drive    21.4   6  258 110 3.08 3.215 19.44  1  0\n#> Hornet Sportabout 18.7   8  360 175 3.15 3.440 17.02  0  0\n#> Valiant           18.1   6  225 105 2.76 3.460 20.22  1  0\n#>                   gear carb\n#> Mazda RX4            4    4\n#> Mazda RX4 Wag        4    4\n#> Datsun 710           4    1\n#> Hornet 4 Drive       3    1\n#> Hornet Sportabout    3    2\n#> Valiant              3    1\n\n# number of rows to sample\nm <- 2L\ndf[sample(nrow(df), m), ]\n#>                mpg cyl disp  hp drat    wt  qsec vs am gear\n#> Valiant       18.1   6  225 105 2.76 3.460 20.22  1  0    3\n#> Mazda RX4 Wag 21.0   6  160 110 3.90 2.875 17.02  0  1    4\n#>               carb\n#> Valiant          1\n#> Mazda RX4 Wag    4\n# select a random starting position from available number of rows\nstart_row <- sample(nrow(df) - m + 1, size = 1)\n\n# adjust ending position while avoiding off-by-one error\nend_row <- start_row + m - 1\n\ndf[start_row:end_row, ]\n#>               mpg cyl disp  hp drat    wt  qsec vs am gear\n#> Mazda RX4      21   6  160 110  3.9 2.620 16.46  0  1    4\n#> Mazda RX4 Wag  21   6  160 110  3.9 2.875 17.02  0  1    4\n#>               carb\n#> Mazda RX4        4\n#> Mazda RX4 Wag    4\n# columns in original order\nnames(mtcars)\n#>  [1] \"mpg\"  \"cyl\"  \"disp\" \"hp\"   \"drat\" \"wt\"   \"qsec\" \"vs\"  \n#>  [9] \"am\"   \"gear\" \"carb\"\n\n# columns in alphabetical order\nnames(mtcars[order(names(mtcars))])\n#>  [1] \"am\"   \"carb\" \"cyl\"  \"disp\" \"drat\" \"gear\" \"hp\"   \"mpg\" \n#>  [9] \"qsec\" \"vs\"   \"wt\""},{"path":"subsetting.html","id":"session-information-2","chapter":"4 Subsetting","heading":"4.4 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble      * 3.2.1   2023-03-20 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"control-flow.html","id":"control-flow","chapter":"5 Control flow","heading":"5 Control flow","text":"","code":""},{"path":"control-flow.html","id":"choices-exercises-5.2.4","chapter":"5 Control flow","heading":"5.1 Choices (Exercises 5.2.4)","text":"Q1. type vector following calls ifelse() return?Read documentation write rules words.A1. rules call ifelse() might return:type unstable, .e. type return depend type condition true (yes , .e.):works cases test argument evaluates logical type:test argument logical type, NA, return NA:test argument doesn’t resolve logical type, try coerce output logical type:also clarified docs function:vector length attributes (including dimensions \"class\") test data values values yes . mode answer coerced logical accommodate first values taken yes values taken .Q2. following code work?A2. code works conditional expressions () - even though numeric type - can successfully coerced logical type.","code":"\nifelse(TRUE, 1, \"no\")\nifelse(FALSE, 1, \"no\")\nifelse(NA, 1, \"no\")\nifelse(TRUE, 1, \"no\") # `numeric` returned\n#> [1] 1\nifelse(FALSE, 1, \"no\") # `character` returned\n#> [1] \"no\"\nifelse(NA_real_, 1, \"no\")\n#> [1] NA\nifelse(NaN, 1, \"no\")\n#> [1] NA\nifelse(NA, 1, \"no\")\n#> [1] NA\n# will work\nifelse(\"TRUE\", 1, \"no\")\n#> [1] 1\nifelse(\"false\", 1, \"no\")\n#> [1] \"no\"\n\n# won't work\nifelse(\"tRuE\", 1, \"no\")\n#> [1] NA\nifelse(NaN, 1, \"no\")\n#> [1] NA\nx <- 1:10\nif (length(x)) \"not empty\" else \"empty\"\n#> [1] \"not empty\"\n\nx <- numeric()\nif (length(x)) \"not empty\" else \"empty\"\n#> [1] \"empty\"\nas.logical(length(1:10))\n#> [1] TRUE\n\nas.logical(length(numeric()))\n#> [1] FALSE"},{"path":"control-flow.html","id":"loops-exercises-5.3.3","chapter":"5 Control flow","heading":"5.2 Loops (Exercises 5.3.3)","text":"Q1. code succeed without errors warnings?A1. works 1:length(x) works positive negative directions.case, since x length 0, go 1 0.Additionally, since --bound (OOB) value atomic vectors NA, related operations OOB values also produce NA.way avoid unintended behavior use seq_along() instead:Q2. following code evaluated, can say vector iterated?A2. iterator variable x initially takes values vector xs. can check printing x iteration:worth noting x updated iteration; otherwise, take increasingly bigger values xs, loop never end executing.Q3. following code tell index updated?A3. () loop index updated beginning iteration. Otherwise, encounter infinite loop.Also, worth contrasting behavior () loop () loop:","code":"\nx <- numeric()\nout <- vector(\"list\", length(x))\nfor (i in 1:length(x)) {\n  out[i] <- x[i]^2\n}\nout\n1:2\n#> [1] 1 2\n1:0\n#> [1] 1 0\n1:-3\n#> [1]  1  0 -1 -2 -3\nx <- numeric()\nout <- vector(\"list\", length(x))\n\nfor (i in 1:length(x)) {\n  print(paste(\"i:\", i, \", x[i]:\", x[i], \", out[i]:\", out[i]))\n\n  out[i] <- x[i]^2\n}\n#> [1] \"i: 1 , x[i]: NA , out[i]: NULL\"\n#> [1] \"i: 0 , x[i]:  , out[i]: \"\n\nout\n#> [[1]]\n#> [1] NA\nx <- numeric()\nout <- vector(\"list\", length(x))\n\nfor (i in seq_along(x)) {\n  out[i] <- x[i]^2\n}\n\nout\n#> list()\nxs <- c(1, 2, 3)\nfor (x in xs) {\n  xs <- c(xs, x * 2)\n}\nxs\n#> [1] 1 2 3 2 4 6\nxs <- c(1, 2, 3)\nfor (x in xs) {\n  cat(\"x:\", x, \"\\n\")\n  xs <- c(xs, x * 2)\n  cat(\"xs:\", paste(xs), \"\\n\")\n}\n#> x: 1 \n#> xs: 1 2 3 2 \n#> x: 2 \n#> xs: 1 2 3 2 4 \n#> x: 3 \n#> xs: 1 2 3 2 4 6\nfor (i in 1:3) {\n  i <- i * 2\n  print(i)\n}\n#> [1] 2\n#> [1] 4\n#> [1] 6\nfor (i in 1:3) {\n  cat(\"before: \", i, \"\\n\")\n  i <- i * 2\n  cat(\"after:  \", i, \"\\n\")\n}\n#> before:  1 \n#> after:   2 \n#> before:  2 \n#> after:   4 \n#> before:  3 \n#> after:   6\ni <- 1\nwhile (i < 4) {\n  cat(\"before: \", i, \"\\n\")\n  i <- i * 2\n  cat(\"after:  \", i, \"\\n\")\n}\n#> before:  1 \n#> after:   2 \n#> before:  2 \n#> after:   4"},{"path":"control-flow.html","id":"session-information-3","chapter":"5 Control flow","heading":"5.3 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"functions.html","id":"functions","chapter":"6 Functions","heading":"6 Functions","text":"Attaching needed libraries:","code":"\nlibrary(tidyverse, warn.conflicts = FALSE)"},{"path":"functions.html","id":"function-fundamentals-exercises-6.2.5","chapter":"6 Functions","heading":"6.1 Function fundamentals (Exercises 6.2.5)","text":"Q1. Given name, like \"mean\", match.fun() lets find function. Given function, can find name? doesn’t make sense R?A1. Given name, match.fun() lets find function., given function, doesn’t make sense find name can multiple names bound function.Q2. ’s possible (although typically useful) call anonymous function. two approaches correct? ?A2. first expression correct since function evaluate 3(), syntactically allowed since literals can’t treated like functions.correct way call anonymous function.Q3. good rule thumb anonymous function fit one line shouldn’t need use {}. Review code. used anonymous function instead named function? used named function instead anonymous function?A3. Self activity.Q4. function allows tell object function? function allows tell function primitive function?A4. Use .function() check object function:Use .primitive() check function primitive:Q5. code makes list functions base package.Use answer following questions:base function arguments?base function arguments?many base functions arguments? ’s special functions?many base functions arguments? ’s special functions?adapt code find primitive functions?adapt code find primitive functions?A5. provided code following:base function arguments?can use formals() extract number arguments, function returns NULL primitive functions.Therefore, focus non-primitive functions.scan() function arguments.many base functions arguments? ’s special functions?time writing, 47 base (non-primitive) functions arguments.adapt code find primitive functions?Q6. three important components function?A6. Except primitive functions, functions 3 important components:formals()body()environment()Q7. printing function show environment created ?A7. package functions print environment:two exceptions enclosing environment won’t printed:primitive functionsfunctions created global environment","code":"\nmatch.fun(\"mean\")\n#> function (x, ...) \n#> UseMethod(\"mean\")\n#> <bytecode: 0x5610be5e25a8>\n#> <environment: namespace:base>\nf1 <- function(x) mean(x)\nf2 <- f1\n\nmatch.fun(\"f1\")\n#> function(x) mean(x)\n\nmatch.fun(\"f2\")\n#> function(x) mean(x)\nfunction(x) 3()\n#> function(x) 3()\n(function(x) 3)()\n#> [1] 3\nf <- (function(x) 3())\nf\n#> function(x) 3()\nf()\n#> Error in f(): attempt to apply non-function\n\nrlang::is_syntactic_literal(3)\n#> [1] TRUE\ng <- (function(x) 3)\ng\n#> function(x) 3\ng()\n#> [1] 3\n# these are functions\nf <- function(x) 3\nis.function(mean)\n#> [1] TRUE\nis.function(f)\n#> [1] TRUE\n\n# these aren't\nis.function(\"x\")\n#> [1] FALSE\nis.function(new.env())\n#> [1] FALSE\n# primitive\nis.primitive(sum)\n#> [1] TRUE\nis.primitive(`+`)\n#> [1] TRUE\n\n# not primitive\nis.primitive(mean)\n#> [1] FALSE\nis.primitive(read.csv)\n#> [1] FALSE\nobjs <- mget(ls(\"package:base\", all = TRUE), inherits = TRUE)\nfuns <- Filter(is.function, objs)\nobjs <- mget(ls(\"package:base\", all = TRUE), inherits = TRUE)\nfuns <- Filter(is.function, objs)\nformals(\"!\")\n#> NULL\n\nlength(formals(\"!\"))\n#> [1] 0\nfuns <- purrr::discard(funs, is.primitive)\ndf_formals <- purrr::map_df(funs, ~ length(formals(.))) %>%\n  tidyr::pivot_longer(\n    cols = dplyr::everything(),\n    names_to = \"function\",\n    values_to = \"argumentCount\"\n  ) %>%\n  dplyr::arrange(desc(argumentCount))\n\ndf_formals\n#> # A tibble: 1,137 × 2\n#>    `function`       argumentCount\n#>    <chr>                    <int>\n#>  1 scan                        22\n#>  2 format.default              16\n#>  3 source                      16\n#>  4 formatC                     15\n#>  5 library                     13\n#>  6 merge.data.frame            13\n#>  7 prettyNum                   13\n#>  8 system2                     11\n#>  9 print.default               10\n#> 10 save                        10\n#> # ℹ 1,127 more rows\ndplyr::filter(df_formals, argumentCount == 0)\n#> # A tibble: 47 × 2\n#>    `function`               argumentCount\n#>    <chr>                            <int>\n#>  1 .First.sys                           0\n#>  2 .NotYetImplemented                   0\n#>  3 .OptRequireMethods                   0\n#>  4 .standard_regexps                    0\n#>  5 .tryResumeInterrupt                  0\n#>  6 closeAllConnections                  0\n#>  7 contributors                         0\n#>  8 Cstack_info                          0\n#>  9 default.stringsAsFactors             0\n#> 10 extSoftVersion                       0\n#> # ℹ 37 more rows\nobjs <- mget(ls(\"package:base\", all = TRUE), inherits = TRUE)\nfuns <- Filter(is.function, objs)\nprimitives <- Filter(is.primitive, funs)\n\nlength(primitives)\n#> [1] 205\n\nnames(primitives)\n#>   [1] \"-\"                    \":\"                   \n#>   [3] \"::\"                   \":::\"                 \n#>   [5] \"!\"                    \"!=\"                  \n#>   [7] \"...elt\"               \"...length\"           \n#>   [9] \"...names\"             \".C\"                  \n#>  [11] \".cache_class\"         \".Call\"               \n#>  [13] \".Call.graphics\"       \".class2\"             \n#>  [15] \".External\"            \".External.graphics\"  \n#>  [17] \".External2\"           \".Fortran\"            \n#>  [19] \".Internal\"            \".isMethodsDispatchOn\"\n#>  [21] \".Primitive\"           \".primTrace\"          \n#>  [23] \".primUntrace\"         \".subset\"             \n#>  [25] \".subset2\"             \"(\"                   \n#>  [27] \"[\"                    \"[[\"                  \n#>  [29] \"[[<-\"                 \"[<-\"                 \n#>  [31] \"{\"                    \"@\"                   \n#>  [33] \"@<-\"                  \"*\"                   \n#>  [35] \"/\"                    \"&\"                   \n#>  [37] \"&&\"                   \"%*%\"                 \n#>  [39] \"%/%\"                  \"%%\"                  \n#>  [41] \"^\"                    \"+\"                   \n#>  [43] \"<\"                    \"<-\"                  \n#>  [45] \"<<-\"                  \"<=\"                  \n#>  [47] \"=\"                    \"==\"                  \n#>  [49] \">\"                    \">=\"                  \n#>  [51] \"|\"                    \"||\"                  \n#>  [53] \"~\"                    \"$\"                   \n#>  [55] \"$<-\"                  \"abs\"                 \n#>  [57] \"acos\"                 \"acosh\"               \n#>  [59] \"all\"                  \"any\"                 \n#>  [61] \"anyNA\"                \"Arg\"                 \n#>  [63] \"as.call\"              \"as.character\"        \n#>  [65] \"as.complex\"           \"as.double\"           \n#>  [67] \"as.environment\"       \"as.integer\"          \n#>  [69] \"as.logical\"           \"as.numeric\"          \n#>  [71] \"as.raw\"               \"asin\"                \n#>  [73] \"asinh\"                \"atan\"                \n#>  [75] \"atanh\"                \"attr\"                \n#>  [77] \"attr<-\"               \"attributes\"          \n#>  [79] \"attributes<-\"         \"baseenv\"             \n#>  [81] \"break\"                \"browser\"             \n#>  [83] \"c\"                    \"call\"                \n#>  [85] \"ceiling\"              \"class\"               \n#>  [87] \"class<-\"              \"Conj\"                \n#>  [89] \"cos\"                  \"cosh\"                \n#>  [91] \"cospi\"                \"cummax\"              \n#>  [93] \"cummin\"               \"cumprod\"             \n#>  [95] \"cumsum\"               \"digamma\"             \n#>  [97] \"dim\"                  \"dim<-\"               \n#>  [99] \"dimnames\"             \"dimnames<-\"          \n#> [101] \"emptyenv\"             \"enc2native\"          \n#> [103] \"enc2utf8\"             \"environment<-\"       \n#> [105] \"exp\"                  \"expm1\"               \n#> [107] \"expression\"           \"floor\"               \n#> [109] \"for\"                  \"forceAndCall\"        \n#> [111] \"function\"             \"gamma\"               \n#> [113] \"gc.time\"              \"globalenv\"           \n#> [115] \"if\"                   \"Im\"                  \n#> [117] \"interactive\"          \"invisible\"           \n#> [119] \"is.array\"             \"is.atomic\"           \n#> [121] \"is.call\"              \"is.character\"        \n#> [123] \"is.complex\"           \"is.double\"           \n#> [125] \"is.environment\"       \"is.expression\"       \n#> [127] \"is.finite\"            \"is.function\"         \n#> [129] \"is.infinite\"          \"is.integer\"          \n#> [131] \"is.language\"          \"is.list\"             \n#> [133] \"is.logical\"           \"is.matrix\"           \n#> [135] \"is.na\"                \"is.name\"             \n#> [137] \"is.nan\"               \"is.null\"             \n#> [139] \"is.numeric\"           \"is.object\"           \n#> [141] \"is.pairlist\"          \"is.raw\"              \n#> [143] \"is.recursive\"         \"is.single\"           \n#> [145] \"is.symbol\"            \"isS4\"                \n#> [147] \"lazyLoadDBfetch\"      \"length\"              \n#> [149] \"length<-\"             \"levels<-\"            \n#> [151] \"lgamma\"               \"list\"                \n#> [153] \"log\"                  \"log10\"               \n#> [155] \"log1p\"                \"log2\"                \n#> [157] \"max\"                  \"min\"                 \n#> [159] \"missing\"              \"Mod\"                 \n#> [161] \"names\"                \"names<-\"             \n#> [163] \"nargs\"                \"next\"                \n#> [165] \"nzchar\"               \"oldClass\"            \n#> [167] \"oldClass<-\"           \"on.exit\"             \n#> [169] \"pos.to.env\"           \"proc.time\"           \n#> [171] \"prod\"                 \"quote\"               \n#> [173] \"range\"                \"Re\"                  \n#> [175] \"rep\"                  \"repeat\"              \n#> [177] \"retracemem\"           \"return\"              \n#> [179] \"round\"                \"seq_along\"           \n#> [181] \"seq_len\"              \"seq.int\"             \n#> [183] \"sign\"                 \"signif\"              \n#> [185] \"sin\"                  \"sinh\"                \n#> [187] \"sinpi\"                \"sqrt\"                \n#> [189] \"standardGeneric\"      \"storage.mode<-\"      \n#> [191] \"substitute\"           \"sum\"                 \n#> [193] \"switch\"               \"tan\"                 \n#> [195] \"tanh\"                 \"tanpi\"               \n#> [197] \"tracemem\"             \"trigamma\"            \n#> [199] \"trunc\"                \"unCfillPOSIXlt\"      \n#> [201] \"unclass\"              \"untracemem\"          \n#> [203] \"UseMethod\"            \"while\"               \n#> [205] \"xtfrm\"\n# base\nmean\n#> function (x, ...) \n#> UseMethod(\"mean\")\n#> <bytecode: 0x5610be5e25a8>\n#> <environment: namespace:base>\n\n# other package function\npurrr::map\n#> function (.x, .f, ..., .progress = FALSE) \n#> {\n#>     map_(\"list\", .x, .f, ..., .progress = .progress)\n#> }\n#> <bytecode: 0x5610c05c5a28>\n#> <environment: namespace:purrr>\nsum\n#> function (..., na.rm = FALSE)  .Primitive(\"sum\")\nf <- function(x) mean(x)\nf\n#> function(x) mean(x)"},{"path":"functions.html","id":"lexical-scoping-exercises-6.4.5","chapter":"6 Functions","heading":"6.2 Lexical scoping (Exercises 6.4.5)","text":"Q1. following code return? ? Describe three c’s interpreted.A1. c(c = c):first c interpreted function call c()second c name vector elementthird c variable value 10You can also see lexical analysis expression:Q2. four principles govern R looks values?A2. Principles govern R looks values:Name masking (names defined inside function mask names defined outside function)Name masking (names defined inside function mask names defined outside function)Functions vs. variables (rule also applies function names)Functions vs. variables (rule also applies function names)fresh start (every time function called, new environment created host execution)fresh start (every time function called, new environment created host execution)Dynamic look-(R looks values function run, function created)Dynamic look-(R looks values function run, function created)Q3. following function return? Make prediction running code .A3. Correctly predicted 😉Although multiple f() functions, order evaluation goes inside outside x^2 evaluated first f(x) * 2 evaluated last. results 202 (= ((10 ^ 2) + 1) * 2).","code":"\nc <- 10\nc(c = c)\nc <- 10\nc(c = c)\n#>  c \n#> 10\np_expr <- parse(text = \"c(c = c)\", keep.source = TRUE)\ngetParseData(p_expr) %>% select(token, text)\n#>                   token text\n#> 12                 expr     \n#> 1  SYMBOL_FUNCTION_CALL    c\n#> 3                  expr     \n#> 2                   '('    (\n#> 4            SYMBOL_SUB    c\n#> 5                EQ_SUB    =\n#> 6                SYMBOL    c\n#> 8                  expr     \n#> 7                   ')'    )\nf <- function(x) {\n  f <- function(x) {\n    f <- function() {\n      x^2\n    }\n    f() + 1\n  }\n  f(x) * 2\n}\nf(10)\nf <- function(x) {\n  f <- function(x) {\n    f <- function() {\n      x^2\n    }\n    f() + 1\n  }\n  f(x) * 2\n}\n\nf(10)\n#> [1] 202"},{"path":"functions.html","id":"lazy-evaluation-exercises-6.5.4","chapter":"6 Functions","heading":"6.3 Lazy evaluation (Exercises 6.5.4)","text":"Q1. important property && makes x_ok() work?different code? behaviour undesirable ?A1. && evaluates left right short-circuit evaluation, .e., first operand TRUE, R short-circuit even look second operand.Replacing && & undesirable performs element-wise logical comparisons returns vector values always useful decision (TRUE, FALSE, NA).Q2. function return? ? principle illustrate?A2. function returns 100 due lazy evaluation.function execution environment encounters x, evaluates argument x = z since name z already bound value 100 environment, x also bound value.can check looking memory addresses:Q3. function return? ? principle illustrate?A3. Let’s first look function returns:name masking. function call c(x, y), x accessed function environment, following promise evaluated function environment:, thus y gets assigned 1, x 2, since last value scope.Therefore, neither promise y = 0 global assignment y <- 10 ever consulted find value y.Q4. hist(), default value xlim range(breaks), default value breaks \"Sturges\", andExplain hist() works get correct xlim value.A4. xlim defines range histogram’s x-axis.default xlim = range(breaks) breaks = \"Sturges\" arguments reveal function uses Sturges’ algorithm compute number breaks.see implementation, run sloop::s3_get_method(\"hist.default\").hist() ensures chosen algorithm returns numeric vector containing least two unique elements xlim computed.Q5. Explain function works. confusing?A5. Let’s take step--step.function argument x missing function call. means stop(\"Error!\") evaluated function environment, global environment., due lazy evaluation, promise stop(\"Error!\") evaluated x accessed. happens print(x) called.print(x) leads x evaluated, evaluates stop function environment. , function environment, base::stop() masked locally defined stop() function, returns Sys.time() output.Q6. many arguments required calling library()?A6. Going solely signature,looks like following arguments required:, reality, one argument required: package. function internally checks arguments missing adjusts accordingly.better arguments NULL instead missing; avoid confusion.","code":"\nx_ok <- function(x) {\n  !is.null(x) && length(x) == 1 && x > 0\n}\n\nx_ok(NULL)\nx_ok(1)\nx_ok(1:3)\nx_ok <- function(x) {\n  !is.null(x) & length(x) == 1 & x > 0\n}\n\nx_ok(NULL)\nx_ok(1)\nx_ok(1:3)\nx_ok <- function(x) {\n  !is.null(x) && length(x) == 1 && x > 0\n}\n\nx_ok(NULL)\n#> [1] FALSE\n\nx_ok(1)\n#> [1] TRUE\n\nx_ok(1:3)\n#> [1] FALSE\nx_ok <- function(x) {\n  !is.null(x) & length(x) == 1 & x > 0\n}\n\nx_ok(NULL)\n#> logical(0)\n\nx_ok(1)\n#> [1] TRUE\n\nx_ok(1:3)\n#> [1] FALSE FALSE FALSE\nf2 <- function(x = z) {\n  z <- 100\n  x\n}\nf2()\nf2 <- function(x = z) {\n  z <- 100\n  print(lobstr::obj_addrs(list(x, z)))\n  x\n}\n\nf2()\n#> [1] \"0x5610bed32bc0\" \"0x5610bed32bc0\"\n#> [1] 100\ny <- 10\nf1 <- function(x =\n                 {\n                   y <- 1\n                   2\n                 },\n               y = 0) {\n  c(x, y)\n}\nf1()\ny\ny <- 10\nf1 <- function(x =\n                 {\n                   y <- 1\n                   2\n                 },\n               y = 0) {\n  c(x, y)\n}\nf1()\n#> [1] 2 1\ny\n#> [1] 10\nx <- {\n  y <- 1\n  2\n}\nrange(\"Sturges\")\n#> [1] \"Sturges\" \"Sturges\"\nhist(mtcars$wt, xlim = c(1, 6))\nnclass.Sturges(mtcars$wt)\n#> [1] 6\nshow_time <- function(x = stop(\"Error!\")) {\n  stop <- function(...) Sys.time()\n  print(x)\n}\n\nshow_time()\n#> [1] \"2023-12-31 00:42:55 UTC\"\nformals(library)\n#> $package\n#> \n#> \n#> $help\n#> \n#> \n#> $pos\n#> [1] 2\n#> \n#> $lib.loc\n#> NULL\n#> \n#> $character.only\n#> [1] FALSE\n#> \n#> $logical.return\n#> [1] FALSE\n#> \n#> $warn.conflicts\n#> \n#> \n#> $quietly\n#> [1] FALSE\n#> \n#> $verbose\n#> getOption(\"verbose\")\n#> \n#> $mask.ok\n#> \n#> \n#> $exclude\n#> \n#> \n#> $include.only\n#> \n#> \n#> $attach.required\n#> missing(include.only)\nformals(library) %>%\n  purrr::discard(is.null) %>%\n  purrr::map_lgl(~ .x == \"\") %>%\n  purrr::keep(~ isTRUE(.x)) %>%\n  names()\n#> [1] \"package\"        \"help\"           \"warn.conflicts\"\n#> [4] \"mask.ok\"        \"exclude\"        \"include.only\""},{"path":"functions.html","id":"dot-dot-dot-exercises-6.6.1","chapter":"6 Functions","heading":"6.4 ... (dot-dot-dot) (Exercises 6.6.1)","text":"Q1. Explain following results:A1. Let’s look arguments functions:can seen, sum() function doesn’t na.omit argument. , input na.omit = TRUE treated 1 (logical implicitly coerced numeric), thus results. , expression evaluates sum(1, 2, 3, 1).mean() function, one parameter (x) ’s matched first argument (1). , expression evaluates mean(1).Q2. Explain find documentation named arguments following function call:A2. Typing ?plot console, see documentation, also shows signature:Since ... passed par(), can look ?par docs:.docs parameters interest reside .Q3. plot(1:10, col = \"red\") colour points, axes labels? Read source code plot.default() find .A3. Source code can found .plot.default() passes ... localTitle(), passes title().title() four parts: main, sub, xlab, ylab.single argument col work ambiguous element apply argument .","code":"\nsum(1, 2, 3)\n#> [1] 6\nmean(1, 2, 3)\n#> [1] 1\n\nsum(1, 2, 3, na.omit = TRUE)\n#> [1] 7\nmean(1, 2, 3, na.omit = TRUE)\n#> [1] 1\nstr(sum)\n#> function (..., na.rm = FALSE)\nstr(mean)\n#> function (x, ...)\nplot(1:10, col = \"red\", pch = 20, xlab = \"x\", col.lab = \"blue\")#> function (x, y, ...)#> function (..., no.readonly = FALSE)\nlocalTitle <- function(..., col, bg, pch, cex, lty, lwd) title(...)\n\ntitle <- function(main = NULL, sub = NULL, xlab = NULL, ylab = NULL,\n                  line = NA, outer = FALSE, ...) {\n  main <- as.graphicsAnnot(main)\n  sub <- as.graphicsAnnot(sub)\n  xlab <- as.graphicsAnnot(xlab)\n  ylab <- as.graphicsAnnot(ylab)\n  .External.graphics(C_title, main, sub, xlab, ylab, line, outer, ...)\n  invisible()\n}"},{"path":"functions.html","id":"exiting-a-function-exercises-6.7.5","chapter":"6 Functions","heading":"6.5 Exiting a function (Exercises 6.7.5)","text":"Q1. load() return? don’t normally see values?A1. load() function reloads datasets saved using save() function:normally don’t see value function loads datasets invisibly.can change setting verbose = TRUE:Q2. write.table() return? useful?A2. write.table() writes data frame file returns NULL invisibly.helpful function invisibly returned actual object written file, used.Q3. chdir parameter source() compare with_dir()? might prefer one ?A3. chdir parameter source() described :TRUE file pathname, R working directory temporarily changed directory containing file evaluatingThat , chdir allows changing working directory temporarily directory containing file sourced:withr::with_dir() temporarily changes current working directory:importantly, parameters dir allows temporarily changing working directory directory.Q4. Write function opens graphics device, runs supplied code, closes graphics device (always, regardless whether plotting code works).A4. function opens graphics device, runs supplied code, closes graphics device:Q5. can use .exit() implement simple version capture.output().Compare capture.output() capture.output2(). functions differ? features removed make key ideas easier see? rewritten key ideas ’re easier understand?A5. capture.output() significantly complex, can seen definition:key differences:capture.output() uses print() function print console:capture.output() can capture messages well:capture.output() takes account visibility expression:","code":"\nsave(iris, file = \"my_iris.rda\")\nload(\"my_iris.rda\")\nload(\"my_iris.rda\", verbose = TRUE)\n#> Loading objects:\n#>   iris\n\n# cleanup\nunlink(\"my_iris.rda\")\nwrite.table(BOD, file = \"BOD.csv\")\n# cleanup\nunlink(\"BOD.csv\")\nwithr::with_dir\n#> function (new, code) \n#> {\n#>     old <- setwd(dir = new)\n#>     on.exit(setwd(old))\n#>     force(code)\n#> }\n#> <bytecode: 0x5610c7d2faa0>\n#> <environment: namespace:withr>\nwith_png_device <- function(filename, code, ...) {\n  grDevices::png(filename = filename, ...)\n  on.exit(grDevices::dev.off(), add = TRUE)\n\n  force(code)\n}\ncapture.output2 <- function(code) {\n  temp <- tempfile()\n  on.exit(file.remove(temp), add = TRUE, after = TRUE)\n\n  sink(temp)\n  on.exit(sink(), add = TRUE, after = TRUE)\n\n  force(code)\n  readLines(temp)\n}\n\ncapture.output2(cat(\"a\", \"b\", \"c\", sep = \"\\n\"))\n#> [1] \"a\" \"b\" \"c\"\ncapture.output\n#> function (..., file = NULL, append = FALSE, type = c(\"output\", \n#>     \"message\"), split = FALSE) \n#> {\n#>     type <- match.arg(type)\n#>     rval <- NULL\n#>     closeit <- TRUE\n#>     if (is.null(file)) \n#>         file <- textConnection(\"rval\", \"w\", local = TRUE)\n#>     else if (is.character(file)) \n#>         file <- file(file, if (append) \n#>             \"a\"\n#>         else \"w\")\n#>     else if (inherits(file, \"connection\")) {\n#>         if (!isOpen(file)) \n#>             open(file, if (append) \n#>                 \"a\"\n#>             else \"w\")\n#>         else closeit <- FALSE\n#>     }\n#>     else stop(\"'file' must be NULL, a character string or a connection\")\n#>     sink(file, type = type, split = split)\n#>     on.exit({\n#>         sink(type = type, split = split)\n#>         if (closeit) close(file)\n#>     })\n#>     for (i in seq_len(...length())) {\n#>         out <- withVisible(...elt(i))\n#>         if (out$visible) \n#>             print(out$value)\n#>     }\n#>     on.exit()\n#>     sink(type = type, split = split)\n#>     if (closeit) \n#>         close(file)\n#>     if (is.null(rval)) \n#>         invisible(NULL)\n#>     else rval\n#> }\n#> <bytecode: 0x5610c81bfaa0>\n#> <environment: namespace:utils>\ncapture.output(1)\n#> [1] \"[1] 1\"\n\ncapture.output2(1)\n#> character(0)\ncapture.output(message(\"Hi there!\"), \"a\", type = \"message\")\n#> Hi there!\n#> [1] \"a\"\n#> character(0)\ncapture.output(1, invisible(2), 3)\n#> [1] \"[1] 1\" \"[1] 3\""},{"path":"functions.html","id":"function-forms-exercises-6.8.6","chapter":"6 Functions","heading":"6.6 Function forms (Exercises 6.8.6)","text":"Q1. Rewrite following code snippets prefix form:A1. Prefix forms code snippets:Q2. Clarify following list odd function calls:A2. functions don’t dots (...) parameters, argument matching takes place following steps:exact matching named argumentspartial matchingposition-basedQ3. Explain following code fails:A3. provided book, replacement function defined :Let’s re-write provided code prefix format understand doesn’t work:Although works:following doesn’t code evaluates :get<- function R.Q4. Create replacement function modifies random location vector.A4. replacement function modifies random location vector:Let’s try :Q5. Write version + pastes inputs together character vectors behaves usual otherwise. words, make code work:A5. Infix operator re-create desired output:Q6. Create list replacement functions found base package. ones primitive functions? (Hint: use apropos().)A6. Replacement functions always <- end names., using apropos(), can find replacement functions search paths filter ones don’t belong {base} package:primitive replacement functions can listed using .primitive():Q7. valid names user-created infix functions?A7. mentioned respective section book:names infix functions flexible regular R functions: can contain sequence characters except %.Q8. Create infix xor() operator.A8. Exclusive logical operation TRUE arguments differ (one TRUE, FALSE).can create infix operator exclusive like :function vectorized inputs underlying logical operators vectorized.Q9. Create infix versions set functions intersect(), union(), setdiff(). might call %n%, %u%, %/% match conventions mathematics.A9. required infix operators can created following:can check outputs agree underlying functions:","code":"\n1 + 2 + 3\n\n1 + (2 + 3)\n\nif (length(x) <= 5) x[[5]] else x[[n]]\n# The binary `+`  operator has left to right associative property.\n`+`(`+`(1, 2), 3)\n\n`+`(1, `(`(`+`(2, 3)))\n\n`if`(cond = `<=`(length(x), 5), cons.expr = `[[`(x, 5), alt.expr = `[[`(x, n))\nx <- sample(replace = TRUE, 20, x = c(1:10, NA))\ny <- runif(min = 0, max = 1, 20)\ncor(m = \"k\", y = y, u = \"p\", x = x)\nmodify(get(\"x\"), 1) <- 10\n#> Error: target of assignment expands to non-language object\n`modify<-` <- function(x, position, value) {\n  x[position] <- value\n  x\n}\nget(\"x\") <- `modify<-`(x = get(\"x\"), position = 1, value = 10)\nx <- 5\n`modify<-`(x = get(\"x\"), position = 1, value = 10)\n#> [1] 10\n`get<-`(\"x\", 10)\n#> Error in `get<-`(\"x\", 10): could not find function \"get<-\"\n`random_modify<-` <- function(x, value) {\n  random_index <- sample(seq_along(x), size = 1)\n  x[random_index] <- value\n  return(x)\n}\nx1 <- rep(\"a\", 10)\nrandom_modify(x1) <- \"X\"\nx1\n#>  [1] \"a\" \"a\" \"a\" \"a\" \"X\" \"a\" \"a\" \"a\" \"a\" \"a\"\n\nx2 <- rep(\"a\", 10)\nrandom_modify(x2) <- \"Y\"\nx2\n#>  [1] \"a\" \"a\" \"a\" \"a\" \"a\" \"Y\" \"a\" \"a\" \"a\" \"a\"\n\nx3 <- rep(0, 15)\nrandom_modify(x3) <- -4\nx3\n#>  [1]  0  0  0  0 -4  0  0  0  0  0  0  0  0  0  0\n\nx4 <- rep(0, 15)\nrandom_modify(x4) <- -1\nx4\n#>  [1]  0  0  0  0  0  0  0  0  0  0  0  0 -1  0  0\n1 + 2\n#> [1] 3\n\n\"a\" + \"b\"\n#> [1] \"ab\"\n`+` <- function(x, y) {\n  if (is.character(x) || is.character(y)) {\n    paste0(x, y)\n  } else {\n    base::`+`(x, y)\n  }\n}\n\n1 + 2\n#> [1] 3\n\n\"a\" + \"b\"\n#> [1] \"ab\"\n\nrm(\"+\", envir = .GlobalEnv)\nls_replacement <- apropos(\"<-\", where = TRUE, mode = \"function\")\n\nbase_index <- which(grepl(\"base\", searchpaths()))\n\nls_replacement <- ls_replacement[which(names(ls_replacement) == as.character(base_index))]\n\nunname(ls_replacement)\n#>  [1] \".rowNamesDF<-\"           \"[[<-\"                   \n#>  [3] \"[[<-.data.frame\"         \"[[<-.factor\"            \n#>  [5] \"[[<-.numeric_version\"    \"[[<-.POSIXlt\"           \n#>  [7] \"[<-\"                     \"[<-.data.frame\"         \n#>  [9] \"[<-.Date\"                \"[<-.difftime\"           \n#> [11] \"[<-.factor\"              \"[<-.numeric_version\"    \n#> [13] \"[<-.POSIXct\"             \"[<-.POSIXlt\"            \n#> [15] \"@<-\"                     \"<-\"                     \n#> [17] \"<<-\"                     \"$<-\"                    \n#> [19] \"$<-.data.frame\"          \"$<-.POSIXlt\"            \n#> [21] \"attr<-\"                  \"attributes<-\"           \n#> [23] \"body<-\"                  \"class<-\"                \n#> [25] \"colnames<-\"              \"comment<-\"              \n#> [27] \"diag<-\"                  \"dim<-\"                  \n#> [29] \"dimnames<-\"              \"dimnames<-.data.frame\"  \n#> [31] \"Encoding<-\"              \"environment<-\"          \n#> [33] \"formals<-\"               \"is.na<-\"                \n#> [35] \"is.na<-.default\"         \"is.na<-.factor\"         \n#> [37] \"is.na<-.numeric_version\" \"length<-\"               \n#> [39] \"length<-.Date\"           \"length<-.difftime\"      \n#> [41] \"length<-.factor\"         \"length<-.POSIXct\"       \n#> [43] \"length<-.POSIXlt\"        \"levels<-\"               \n#> [45] \"levels<-.factor\"         \"mode<-\"                 \n#> [47] \"mostattributes<-\"        \"names<-\"                \n#> [49] \"names<-.POSIXlt\"         \"oldClass<-\"             \n#> [51] \"parent.env<-\"            \"regmatches<-\"           \n#> [53] \"row.names<-\"             \"row.names<-.data.frame\" \n#> [55] \"row.names<-.default\"     \"rownames<-\"             \n#> [57] \"split<-\"                 \"split<-.data.frame\"     \n#> [59] \"split<-.default\"         \"storage.mode<-\"         \n#> [61] \"substr<-\"                \"substring<-\"            \n#> [63] \"units<-\"                 \"units<-.difftime\"\nmget(ls_replacement, envir = baseenv()) %>%\n  purrr::keep(is.primitive) %>%\n  names()\n#>  [1] \"[[<-\"           \"[<-\"            \"@<-\"           \n#>  [4] \"<-\"             \"<<-\"            \"$<-\"           \n#>  [7] \"attr<-\"         \"attributes<-\"   \"class<-\"       \n#> [10] \"dim<-\"          \"dimnames<-\"     \"environment<-\" \n#> [13] \"length<-\"       \"levels<-\"       \"names<-\"       \n#> [16] \"oldClass<-\"     \"storage.mode<-\"\nlv1 <- c(TRUE, FALSE, TRUE, FALSE)\nlv2 <- c(TRUE, TRUE, FALSE, FALSE)\n\nxor(lv1, lv2)\n#> [1] FALSE  TRUE  TRUE FALSE\n`%xor%` <- function(x, y) {\n  !((x & y) | !(x | y))\n}\n\nlv1 %xor% lv2\n#> [1] FALSE  TRUE  TRUE FALSE\n\nTRUE %xor% TRUE\n#> [1] FALSE\n`%n%` <- function(x, y) {\n  intersect(x, y)\n}\n\n`%u%` <- function(x, y) {\n  union(x, y)\n}\n\n`%/%` <- function(x, y) {\n  setdiff(x, y)\n}\n(x <- c(sort(sample(1:20, 9)), NA))\n#>  [1]  4  7  8  9 11 13 15 16 20 NA\n(y <- c(sort(sample(3:23, 7)), NA))\n#> [1]  9 10 13 15 17 19 20 NA\n\nidentical(intersect(x, y), x %n% y)\n#> [1] TRUE\nidentical(union(x, y), x %u% y)\n#> [1] TRUE\nidentical(setdiff(x, y), x %/% y)\n#> [1] TRUE"},{"path":"functions.html","id":"session-information-4","chapter":"6 Functions","heading":"6.7 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  colorspace    2.1-0   2023-01-23 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr       * 1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  forcats     * 1.0.0   2023-01-29 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  ggplot2     * 3.4.4   2023-10-12 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  grid          4.3.2   2023-11-01 [3] local\n#>  gtable        0.3.4   2023-08-21 [1] RSPM\n#>  highr         0.10    2022-12-22 [1] RSPM\n#>  hms           1.1.3   2023-03-21 [1] RSPM\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  lobstr        1.1.2   2022-06-22 [1] RSPM\n#>  lubridate   * 1.9.3   2023-09-27 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  munsell       0.5.0   2018-06-12 [1] RSPM\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  purrr       * 1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  readr       * 2.1.4   2023-02-10 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  scales        1.3.0   2023-11-28 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  stringi       1.8.3   2023-12-11 [1] RSPM\n#>  stringr     * 1.5.1   2023-11-14 [1] RSPM\n#>  tibble      * 3.2.1   2023-03-20 [1] RSPM\n#>  tidyr       * 1.3.0   2023-01-24 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tidyverse   * 2.0.0   2023-02-22 [1] RSPM\n#>  timechange    0.2.0   2023-01-11 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  tzdb          0.4.0   2023-05-12 [1] RSPM\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"environments.html","id":"environments","chapter":"7 Environments","heading":"7 Environments","text":"Loading needed libraries:","code":"\nlibrary(rlang, warn.conflicts = FALSE)"},{"path":"environments.html","id":"environment-basics-exercises-7.2.7","chapter":"7 Environments","heading":"7.1 Environment basics (Exercises 7.2.7)","text":"Q1. List three ways environment differs list.A1. mentioned book, ways environments differ lists:Q2. Create environment illustrated picture.A2. Creating environment illustrated picture:binding loop memory address environment e:Q3. Create pair environments illustrated picture.A3. Creating specified environment:Q4. Explain e[[1]] e[c(\"\", \"b\")] don’t make sense e environment.A4. environment non-linear data structure, concept ordered elements. Therefore, indexing (e.g. e[[1]]) doesn’t make sense.Subsetting list vector returns subset underlying data structure. example, subsetting vector returns another vector. ’s unclear subsetting environment (e.g. e[c(\"\", \"b\")]) return data structure contain returns. can’t another environment since environments reference semantics.Q5. Create version env_poke() bind new names, never re-bind old names. programming languages , known single assignment languages.A5. Create version env_poke() doesn’t allow re-binding old names:Making sure behaves expected:Contrast behavior following:Q6. function ? differ <<- might prefer ?A6. downside <<- create new binding doesn’t exist given environment, something may wish:rebind() function let us know binding doesn’t exist, much safer:","code":"\nlibrary(rlang)\n\ne <- env()\ne$loop <- e\nenv_print(e)\n#> <environment: 0x558a72b692e8>\n#> Parent: <environment: global>\n#> Bindings:\n#> • loop: <env>\nlobstr::ref(e$loop)\n#> █ [1:0x558a72b692e8] <env> \n#> └─loop = [1:0x558a72b692e8]\ne1 <- env()\ne2 <- env()\n\ne1$loop <- e2\ne2$deloop <- e1\n\n# following should be the same\nlobstr::obj_addrs(list(e1, e2$deloop))\n#> [1] \"0x558a751c8be8\" \"0x558a751c8be8\"\nlobstr::obj_addrs(list(e2, e1$loop))\n#> [1] \"0x558a75216348\" \"0x558a75216348\"\nenv_poke2 <- function(env, nm, value) {\n  if (env_has(env, nm)) {\n    abort(\"Can't re-bind existing names.\")\n  }\n\n  env_poke(env, nm, value)\n}\ne <- env(a = 1, b = 2, c = 3)\n\n# re-binding old names not allowed\nenv_poke2(e, \"b\", 4)\n#> Error in `env_poke2()`:\n#> ! Can't re-bind existing names.\n\n# binding new names allowed\nenv_poke2(e, \"d\", 8)\ne$d\n#> [1] 8\ne <- env(a = 1, b = 2, c = 3)\n\ne$b\n#> [1] 2\n\n# re-binding old names allowed\nenv_poke(e, \"b\", 4)\ne$b\n#> [1] 4\nrebind <- function(name, value, env = caller_env()) {\n  if (identical(env, empty_env())) {\n    stop(\"Can't find `\", name, \"`\", call. = FALSE)\n  } else if (env_has(env, name)) {\n    env_poke(env, name, value)\n  } else {\n    rebind(name, value, env_parent(env))\n  }\n}\nrebind(\"a\", 10)\n#> Error: Can't find `a`\na <- 5\nrebind(\"a\", 10)\na\n#> [1] 10\n# `x` doesn't exist\nexists(\"x\")\n#> [1] FALSE\n\n# so `<<-` will create one for us\n{\n  x <<- 5\n}\n\n# in the global environment\nenv_has(global_env(), \"x\")\n#>    x \n#> TRUE\nx\n#> [1] 5\nrebind <- function(name, value, env = caller_env()) {\n  if (identical(env, empty_env())) {\n    stop(\"Can't find `\", name, \"`\", call. = FALSE)\n  } else if (env_has(env, name)) {\n    env_poke(env, name, value)\n  } else {\n    rebind(name, value, env_parent(env))\n  }\n}\n\n# doesn't exist\nexists(\"abc\")\n#> [1] FALSE\n\n# so function will produce an error instead of creating it for us\nrebind(\"abc\", 10)\n#> Error: Can't find `abc`\n\n# but it will work as expected when the variable already exists\nabc <- 5\nrebind(\"abc\", 10)\nabc\n#> [1] 10"},{"path":"environments.html","id":"recursing-over-environments-exercises-7.3.1","chapter":"7 Environments","heading":"7.2 Recursing over environments (Exercises 7.3.1)","text":"Q1. Modify () return environments contain binding name. Carefully think type object function need return.A1. modified version () returns environments contain binding name.Since anticipate one environment, dynamically update list time environment specified binding found. important initialize empty list since signifies given binding found environments.Let’s try :Q2. Write function called fget() finds function objects. two arguments, name env, obey regular scoping rules functions: ’s object matching name ’s function, look parent. added challenge, also add inherits argument controls whether function recurses parents looks one environment.A2. function recursively looks function objects:Let’s try :","code":"\nwhere <- function(name, env = caller_env()) {\n  env_list <- list()\n\n  while (!identical(env, empty_env())) {\n    if (env_has(env, name)) {\n      env_list <- append(env_list, env)\n    }\n\n    env <- env_parent(env)\n  }\n\n  return(env_list)\n}\nwhere(\"yyy\")\n#> list()\n\nx <- 5\nwhere(\"x\")\n#> [[1]]\n#> <environment: R_GlobalEnv>\n\nwhere(\"mean\")\n#> [[1]]\n#> <environment: base>\n\nlibrary(dplyr, warn.conflicts = FALSE)\nwhere(\"filter\")\n#> [[1]]\n#> <environment: package:dplyr>\n#> attr(,\"name\")\n#> [1] \"package:dplyr\"\n#> attr(,\"path\")\n#> [1] \"/home/runner/work/_temp/Library/dplyr\"\n#> \n#> [[2]]\n#> <environment: package:stats>\n#> attr(,\"name\")\n#> [1] \"package:stats\"\n#> attr(,\"path\")\n#> [1] \"/opt/R/4.3.2/lib/R/library/stats\"\ndetach(\"package:dplyr\")\nfget <- function(name, env = caller_env(), inherits = FALSE) {\n  # we need only function objects\n  f_value <- mget(name,\n    envir = env,\n    mode = \"function\",\n    inherits = FALSE, # since we have our custom argument\n    ifnotfound = list(NULL)\n  )\n\n  if (!is.null(f_value[[1]])) {\n    # success case\n    f_value[[1]]\n  } else {\n    if (inherits && !identical(env, empty_env())) {\n      # recursive case\n      env <- env_parent(env)\n      fget(name, env, inherits = TRUE)\n    } else {\n      # base case\n      stop(\"No function objects with matching name was found.\", call. = FALSE)\n    }\n  }\n}\nfget(\"mean\", inherits = FALSE)\n#> Error: No function objects with matching name was found.\n\nfget(\"mean\", inherits = TRUE)\n#> function (x, ...) \n#> UseMethod(\"mean\")\n#> <bytecode: 0x558a7059e5a8>\n#> <environment: namespace:base>\n\nmean <- 5\nfget(\"mean\", inherits = FALSE)\n#> Error: No function objects with matching name was found.\n\nmean <- function() NULL\nfget(\"mean\", inherits = FALSE)\n#> function() NULL\nrm(\"mean\")"},{"path":"environments.html","id":"special-environments-exercises-7.4.5","chapter":"7 Environments","heading":"7.3 Special environments (Exercises 7.4.5)","text":"Q1. search_envs() different env_parents(global_env())?A1. search_envs() lists chain environments currently attached search path contains exported functions packages. search path always ends {base} package environment. search path also includes global environment.env_parents() lists parent environments empty environment. course, global environment included list.Q2. Draw diagram shows enclosing environments function:A2. don’t access graphics software used create diagrams book, linking diagram official solutions manual, also find detailed description figure:Q3. Write enhanced version str() provides information functions. Show function found environment defined .A3. write required function, can first re-purpose fget() function wrote return environment found enclosing environment:Let’s try :can now write new version str() wrapper around function. need foresee users might enter function name either symbol string.Let’s first try base::mean():variant present global environment:","code":"\nsearch_envs()\n#>  [[1]] $ <env: global>\n#>  [[2]] $ <env: package:rlang>\n#>  [[3]] $ <env: package:magrittr>\n#>  [[4]] $ <env: package:stats>\n#>  [[5]] $ <env: package:graphics>\n#>  [[6]] $ <env: package:grDevices>\n#>  [[7]] $ <env: package:utils>\n#>  [[8]] $ <env: package:datasets>\n#>  [[9]] $ <env: package:methods>\n#> [[10]] $ <env: Autoloads>\n#> [[11]] $ <env: package:base>\nenv_parents(global_env())\n#>  [[1]] $ <env: package:rlang>\n#>  [[2]] $ <env: package:magrittr>\n#>  [[3]] $ <env: package:stats>\n#>  [[4]] $ <env: package:graphics>\n#>  [[5]] $ <env: package:grDevices>\n#>  [[6]] $ <env: package:utils>\n#>  [[7]] $ <env: package:datasets>\n#>  [[8]] $ <env: package:methods>\n#>  [[9]] $ <env: Autoloads>\n#> [[10]] $ <env: package:base>\n#> [[11]] $ <env: empty>\nf1 <- function(x1) {\n  f2 <- function(x2) {\n    f3 <- function(x3) {\n      x1 + x2 + x3\n    }\n    f3(3)\n  }\n  f2(2)\n}\nf1(1)\nfget2 <- function(name, env = caller_env()) {\n  # we need only function objects\n  f_value <- mget(name,\n    envir = env,\n    mode = \"function\",\n    inherits = FALSE,\n    ifnotfound = list(NULL)\n  )\n\n  if (!is.null(f_value[[1]])) {\n    # success case\n    list(\n      \"where\" = env,\n      \"enclosing\" = fn_env(f_value[[1]])\n    )\n  } else {\n    if (!identical(env, empty_env())) {\n      # recursive case\n      env <- env_parent(env)\n      fget2(name, env)\n    } else {\n      # base case\n      stop(\"No function objects with matching name was found.\", call. = FALSE)\n    }\n  }\n}\nfget2(\"mean\")\n#> $where\n#> <environment: base>\n#> \n#> $enclosing\n#> <environment: namespace:base>\n\nmean <- function() NULL\nfget2(\"mean\")\n#> $where\n#> <environment: R_GlobalEnv>\n#> \n#> $enclosing\n#> <environment: R_GlobalEnv>\nrm(\"mean\")\nstr_function <- function(.f) {\n  fget2(as_string(ensym(.f)))\n}\nstr_function(mean)\n#> $where\n#> <environment: base>\n#> \n#> $enclosing\n#> <environment: namespace:base>\n\nstr_function(\"mean\")\n#> $where\n#> <environment: base>\n#> \n#> $enclosing\n#> <environment: namespace:base>\nmean <- function() NULL\n\nstr_function(mean)\n#> $where\n#> <environment: R_GlobalEnv>\n#> \n#> $enclosing\n#> <environment: R_GlobalEnv>\n\nstr_function(\"mean\")\n#> $where\n#> <environment: R_GlobalEnv>\n#> \n#> $enclosing\n#> <environment: R_GlobalEnv>\n\nrm(\"mean\")"},{"path":"environments.html","id":"call-stacks-exercises-7.5.5","chapter":"7 Environments","heading":"7.4 Call stacks (Exercises 7.5.5)","text":"Q1. Write function lists variables defined environment called. return results ls().A1. function lists variables defined environment called:workhorse rlang::caller_env(), let’s also look definition:Let’s try :global environment:function environment:","code":"\n# let's first remove everything that exists in the global environment right now\n# to test with only newly defined objects\nrm(list = ls())\nrm(.Random.seed, envir = globalenv())\n\nls_env <- function(env = rlang::caller_env()) {\n  sort(rlang::env_names(env))\n}\nrlang::caller_env\n#> function (n = 1) \n#> {\n#>     parent.frame(n + 1)\n#> }\n#> <bytecode: 0x558a710bd8d0>\n#> <environment: namespace:rlang>\nx <- \"a\"\ny <- 1\n\nls_env()\n#> [1] \"ls_env\" \"x\"      \"y\"\n\nls()\n#> [1] \"ls_env\" \"x\"      \"y\"\nfoo <- function() {\n  a <- \"x\"\n  b <- 2\n\n  print(ls_env())\n\n  print(ls())\n}\n\nfoo()\n#> [1] \"a\" \"b\"\n#> [1] \"a\" \"b\""},{"path":"environments.html","id":"session-information-5","chapter":"7 Environments","heading":"7.5 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  crayon        1.5.2   2022-09-29 [1] RSPM\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr         1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  highr         0.10    2022-12-22 [1] RSPM\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  lobstr        1.1.2   2022-06-22 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  png           0.1-8   2022-11-29 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"conditions.html","id":"conditions","chapter":"8 Conditions","heading":"8 Conditions","text":"Attaching needed libraries:","code":"\nlibrary(rlang, warn.conflicts = FALSE)\nlibrary(testthat, warn.conflicts = FALSE)"},{"path":"conditions.html","id":"signalling-conditions-exercises-8.2.4","chapter":"8 Conditions","heading":"8.1 Signalling conditions (Exercises 8.2.4)","text":"Q1. Write wrapper around file.remove() throws error file deleted exist.A1. Let’s first create wrapper function around file.remove() throws error file deleted exist.Let’s first create file can delete immediately.function fail files provided don’t exist:work expected file exists:Q2. appendLF argument message() ? related cat()?A2. mentioned docs message(), appendLF argument decides:messages given character string newline appended?TRUE (default value), final newline regarded part message:FALSE, messages concatenated:hand, cat() converts arguments character vectors concatenates single character vector default:order get message()-like default behavior outputs, can set sep = \"\\n\":","code":"\nfileRemove <- function(...) {\n  existing_files <- fs::file_exists(...)\n\n  if (!all(existing_files)) {\n    stop(\n      cat(\n        \"The following files to be deleted don't exist:\",\n        names(existing_files[!existing_files]),\n        sep = \"\\n\"\n      ),\n      call. = FALSE\n    )\n  }\n\n  file.remove(...)\n}\nfs::file_create(\"random.R\")\nfileRemove(c(\"random.R\", \"XYZ.csv\"))\n#> The following files to be deleted don't exist:\n#> XYZ.csv\n#> Error:\nfileRemove(\"random.R\")\n#> [1] TRUE\nfoo <- function(appendLF) {\n  message(\"Beetle\", appendLF = appendLF)\n  message(\"Juice\", appendLF = appendLF)\n}\n\nfoo(appendLF = TRUE)\n#> Beetle\n#> Juice\nfoo <- function(appendLF) {\n  message(\"Beetle\", appendLF = appendLF)\n  message(\"Juice\", appendLF = appendLF)\n}\n\nfoo(appendLF = FALSE)\n#> BeetleJuice\nfoo <- function() {\n  cat(\"Beetle\")\n  cat(\"Juice\")\n}\n\nfoo()\n#> BeetleJuice\nfoo <- function() {\n  cat(\"Beetle\", sep = \"\\n\")\n  cat(\"Juice\", sep = \"\\n\")\n}\n\nfoo()\n#> Beetle\n#> Juice"},{"path":"conditions.html","id":"handling-conditions-exercises-8.4.5","chapter":"8 Conditions","heading":"8.2 Handling conditions (Exercises 8.4.5)","text":"Q1. extra information condition generated abort() contain compared condition generated stop() .e. ’s difference two objects? Read help ?abort learn .A1. Compared base::stop(), rlang::abort() contains two additional pieces information:trace: traceback capturing sequence calls lead current functionparent: Information another condition used parent create chained condition.Q2. Predict results evaluating following codeA2. Correctly predicted 😉first three pieces code straightforward:last piece code challenging one illustrates tryCatch() works. docs:several handlers supplied single tryCatch first one considered recent second.Q3. Explain results running code:A3. surprising part output b last c.happens inner calling handler doesn’t handle message, bubbles outer calling handler.Q4. Read source code catch_cnd() explain works.A4. Let’s look source code catch_cnd():mentioned function docs:small wrapper around tryCatch() captures condition signalled evaluating argument.classes argument allows character vector condition classes catch, complex tidy evaluation generates necessary condition (; otherwise NULL).Q5. rewrite show_condition() use single handler?A5. source code rlang::catch_cond() gives us clue can .Conditions also class attribute, can use determine handler match condition.Let’s try new version examples used original version:","code":"\ncatch_cnd(stop(\"An error\"))\ncatch_cnd(abort(\"An error\"))\nlibrary(rlang)\n\nstopInfo <- catch_cnd(stop(\"An error\"))\nabortInfo <- catch_cnd(abort(\"An error\"))\n\nstr(stopInfo)\n#> List of 2\n#>  $ message: chr \"An error\"\n#>  $ call   : language force(expr)\n#>  - attr(*, \"class\")= chr [1:3] \"simpleError\" \"error\" \"condition\"\n\nstr(abortInfo)\n#> List of 5\n#>  $ message: chr \"An error\"\n#>  $ trace  :Classes 'rlang_trace', 'rlib_trace', 'tbl' and 'data.frame':  8 obs. of  6 variables:\n#>   ..$ call       :List of 8\n#>   .. ..$ : language catch_cnd(abort(\"An error\"))\n#>   .. ..$ : language eval_bare(rlang::expr(tryCatch(!!!handlers, {     force(expr) ...\n#>   .. ..$ : language tryCatch(condition = `<fn>`, {     force(expr) ...\n#>   .. ..$ : language tryCatchList(expr, classes, parentenv, handlers)\n#>   .. ..$ : language tryCatchOne(expr, names, parentenv, handlers[[1L]])\n#>   .. ..$ : language doTryCatch(return(expr), name, parentenv, handler)\n#>   .. ..$ : language force(expr)\n#>   .. ..$ : language abort(\"An error\")\n#>   ..$ parent     : int [1:8] 0 1 1 3 4 5 1 0\n#>   ..$ visible    : logi [1:8] FALSE FALSE FALSE FALSE FALSE FALSE ...\n#>   ..$ namespace  : chr [1:8] \"rlang\" \"rlang\" \"base\" \"base\" ...\n#>   ..$ scope      : chr [1:8] \"::\" \"::\" \"::\" \"local\" ...\n#>   ..$ error_frame: logi [1:8] FALSE FALSE FALSE FALSE FALSE FALSE ...\n#>   ..- attr(*, \"version\")= int 2\n#>  $ parent : NULL\n#>  $ rlang  :List of 1\n#>   ..$ inherit: logi TRUE\n#>  $ call   : NULL\n#>  - attr(*, \"class\")= chr [1:3] \"rlang_error\" \"error\" \"condition\"\nshow_condition <- function(code) {\n  tryCatch(\n    error = function(cnd) \"error\",\n    warning = function(cnd) \"warning\",\n    message = function(cnd) \"message\",\n    {\n      code\n      NULL\n    }\n  )\n}\n\nshow_condition(stop(\"!\"))\nshow_condition(10)\nshow_condition(warning(\"?!\"))\nshow_condition({\n  10\n  message(\"?\")\n  warning(\"?!\")\n})\nshow_condition <- function(code) {\n  tryCatch(\n    error = function(cnd) \"error\",\n    warning = function(cnd) \"warning\",\n    message = function(cnd) \"message\",\n    {\n      code\n      NULL\n    }\n  )\n}\n\nshow_condition(stop(\"!\"))\n#> [1] \"error\"\nshow_condition(10)\n#> NULL\nshow_condition(warning(\"?!\"))\n#> [1] \"warning\"\nshow_condition({\n  10\n  message(\"?\")\n  warning(\"?!\")\n})\n#> [1] \"message\"\nwithCallingHandlers(\n  message = function(cnd) message(\"b\"),\n  withCallingHandlers(\n    message = function(cnd) message(\"a\"),\n    message(\"c\")\n  )\n)\n#> b\n#> a\n#> b\n#> c\nrlang::catch_cnd\n#> function (expr, classes = \"condition\") \n#> {\n#>     stopifnot(is_character(classes))\n#>     handlers <- rep_named(classes, list(identity))\n#>     eval_bare(rlang::expr(tryCatch(!!!handlers, {\n#>         force(expr)\n#>         return(NULL)\n#>     })))\n#> }\n#> <bytecode: 0x564fa21357a8>\n#> <environment: namespace:rlang>\ncatch_cnd(10)\n#> NULL\n\ncatch_cnd(abort(message = \"an error\", class = \"class1\"))\n#> <error/class1>\n#> Error:\n#> ! an error\n#> ---\n#> Backtrace:\n#> ▆\nshow_condition2 <- function(code) {\n  tryCatch(\n    condition = function(cnd) {\n      if (inherits(cnd, \"error\")) {\n        return(\"error\")\n      }\n      if (inherits(cnd, \"warning\")) {\n        return(\"warning\")\n      }\n      if (inherits(cnd, \"message\")) {\n        return(\"message\")\n      }\n    },\n    {\n      code\n      NULL\n    }\n  )\n}\nshow_condition2(stop(\"!\"))\n#> [1] \"error\"\nshow_condition2(10)\n#> NULL\nshow_condition2(warning(\"?!\"))\n#> [1] \"warning\"\nshow_condition2({\n  10\n  message(\"?\")\n  warning(\"?!\")\n})\n#> [1] \"message\""},{"path":"conditions.html","id":"custom-conditions-exercises-8.5.4","chapter":"8 Conditions","heading":"8.3 Custom conditions (Exercises 8.5.4)","text":"Q1. Inside package, ’s occasionally useful check package installed using . Write function checks package installed (requireNamespace(\"pkg\", quietly = FALSE)) , throws custom condition includes package name metadata.A1. desired function:reference, also see source code following functions:rlang::is_installed()insight::check_if_installed()Q2. Inside package often need stop error something right. packages depend package might tempted check errors unit tests. help packages avoid relying error message part user interface rather API might change without notice?A2. example, let’s say another package developer wanted use check_if_pkg_installed() function just wrote.developer using package can write unit test like :dissuade developers rely error messages check errors, can instead provide custom condition, can used unit testing instead:unit test :test wouldn’t fail even decided change exact message.","code":"\nabort_missing_package <- function(pkg) {\n  msg <- glue::glue(\"Problem loading `{pkg}` package, which is missing and must be installed.\")\n\n  abort(\"error_missing_package\",\n    message = msg,\n    pkg = pkg\n  )\n}\n\ncheck_if_pkg_installed <- function(pkg) {\n  if (!requireNamespace(pkg, quietly = TRUE)) {\n    abort_missing_package(pkg)\n  }\n\n  TRUE\n}\n\ncheck_if_pkg_installed(\"xyz123\")\n#> Error in `abort_missing_package()`:\n#> ! Problem loading `xyz123` package, which is missing and must be installed.\ncheck_if_pkg_installed(\"dplyr\")\n#> [1] TRUE\nexpect_error(\n  check_if_pkg_installed(\"xyz123\"),\n  \"Problem loading `xyz123` package, which is missing and must be installed.\"\n)\ne <- catch_cnd(check_if_pkg_installed(\"xyz123\"))\n\ninherits(e, \"error_missing_package\")\n#> [1] TRUE\nexpect_s3_class(e, \"error_missing_package\")"},{"path":"conditions.html","id":"applications-exercises-8.6.6","chapter":"8 Conditions","heading":"8.4 Applications (Exercises 8.6.6)","text":"Q1. Create suppressConditions() works like suppressMessages() suppressWarnings() suppresses everything. Think carefully handle errors.A1. create desired suppressConditions(), just need create equivalent suppressWarnings() suppressMessages() errors. suppress error message, can handle errors within tryCatch() return error object invisibly:Let’s try see works expected:condition messages now suppressed, note assign error object variable, can still extract useful information debugging:Q2. Compare following two implementations message2error(). main advantage withCallingHandlers() scenario? (Hint: look carefully traceback.)A2. withCallingHandlers(), condition handler called signaling function , , therefore, provides detailed call stack.tryCatch(), signalling function terminates condition raised, doesn’t provide detailed call stack.Q3. modify catch_cnds() definition wanted recreate original intermingling warnings messages?A3. Actually, won’t modify anything function defined chapter, since supports box.nothing additional here5! 😅Q4. catching interrupts dangerous? Run code find .A4. function catches interrupt way stop bottles_of_beer(), way usually stop using interrupt!RStudio IDE, can snap loop terminating R session.shows catching interrupt dangerous can result poor user experience.","code":"\nsuppressErrors <- function(expr) {\n  tryCatch(\n    error = function(cnd) invisible(cnd),\n    expr\n  )\n}\n\nsuppressConditions <- function(expr) {\n  suppressErrors(suppressWarnings(suppressMessages(expr)))\n}\nsuppressConditions(1)\n#> [1] 1\n\nsuppressConditions({\n  message(\"I'm messaging you\")\n  warning(\"I'm warning you\")\n})\n\nsuppressConditions({\n  stop(\"I'm stopping this\")\n})\ne <- suppressConditions({\n  stop(\"I'm stopping this\")\n})\n\ne\n#> <simpleError in withCallingHandlers(expr, message = function(c) if (inherits(c,     classes)) tryInvokeRestart(\"muffleMessage\")): I'm stopping this>\nmessage2error <- function(code) {\n  withCallingHandlers(code, message = function(e) stop(e))\n}\nmessage2error <- function(code) {\n  tryCatch(code, message = function(e) stop(e))\n}\nmessage2error1 <- function(code) {\n  withCallingHandlers(code, message = function(e) stop(\"error\"))\n}\n\nmessage2error1({\n  1\n  message(\"hidden error\")\n  NULL\n})\n#> Error in (function (e) : error\n\ntraceback()\n#> 9: stop(\"error\") at #2\n#> 8: (function (e)\n#>    stop(\"error\"))(list(message = \"hidden error\\n\",\n#>      call = message(\"hidden error\")))\n#> 7: signalCondition(cond)\n#> 6: doWithOneRestart(return(expr), restart)\n#> 5: withOneRestart(expr, restarts[[1L]])\n#> 4: withRestarts({\n#>        signalCondition(cond)\n#>        defaultHandler(cond)\n#>    }, muffleMessage = function() NULL)\n#> 3: message(\"hidden error\") at #1\n#> 2: withCallingHandlers(code,\n#>      message = function(e) stop(\"error\")) at #2\n#> 1: message2error1({\n#>        1\n#>        message(\"hidden error\")\n#>        NULL\n#>    })\nmessage2error2 <- function(code) {\n  tryCatch(code, message = function(e) (stop(\"error\")))\n}\n\nmessage2error2({\n  1\n  stop(\"hidden error\")\n  NULL\n})\n#> Error in value[[3L]](cond) : error\n\ntraceback()\n#> 6: stop(\"error\") at #2\n#> 5: value[[3L]](cond)\n#> 4: tryCatchOne(expr, names, parentenv, handlers[[1L]])\n#> 3: tryCatchList(expr, classes, parentenv, handlers)\n#> 2: tryCatch(code, message = function(e) (stop(\"error\"))) at #2\n#> 1: message2error2({\n#>        1\n#>        message(\"hidden error\")\n#>        NULL\n#>    })\ncatch_cnds <- function(expr) {\n  conds <- list()\n  add_cond <- function(cnd) {\n    conds <<- append(conds, list(cnd))\n    cnd_muffle(cnd)\n  }\n\n  withCallingHandlers(\n    message = add_cond,\n    warning = add_cond,\n    expr\n  )\n\n  conds\n}\n\ncatch_cnds({\n  inform(\"a\")\n  warn(\"b\")\n  inform(\"c\")\n})\n#> [[1]]\n#> <message/rlang_message>\n#> Message:\n#> a\n#> \n#> [[2]]\n#> <warning/rlang_warning>\n#> Warning:\n#> b\n#> \n#> [[3]]\n#> <message/rlang_message>\n#> Message:\n#> c\nbottles_of_beer <- function(i = 99) {\n  message(\n    \"There are \", i, \" bottles of beer on the wall, \",\n    i, \" bottles of beer.\"\n  )\n  while (i > 0) {\n    tryCatch(\n      Sys.sleep(1),\n      interrupt = function(err) {\n        i <<- i - 1\n        if (i > 0) {\n          message(\n            \"Take one down, pass it around, \", i,\n            \" bottle\", if (i > 1) \"s\", \" of beer on the wall.\"\n          )\n        }\n      }\n    )\n  }\n  message(\n    \"No more bottles of beer on the wall, \",\n    \"no more bottles of beer.\"\n  )\n}\nbottles_of_beer()\n#> There are 99 bottles of beer on the wall, 99 bottles of beer.\n#> Take one down, pass it around, 98 bottles of beer on the wall.\n#> Take one down, pass it around, 97 bottles of beer on the wall.\n#> Take one down, pass it around, 96 bottles of beer on the wall.\n#> Take one down, pass it around, 95 bottles of beer on the wall.\n#> Take one down, pass it around, 94 bottles of beer on the wall.\n#> Take one down, pass it around, 93 bottles of beer on the wall.\n#> Take one down, pass it around, 92 bottles of beer on the wall.\n#> Take one down, pass it around, 91 bottles of beer on the wall.\n#> ..."},{"path":"conditions.html","id":"session-information-6","chapter":"8 Conditions","heading":"8.5 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  brio          1.1.4   2023-12-10 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  desc          1.4.3   2023-12-10 [1] RSPM\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr         1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  pkgload       1.3.3   2023-09-22 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  rprojroot     2.0.4   2023-11-05 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  testthat    * 3.2.1   2023-12-02 [1] RSPM\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"functionals.html","id":"functionals","chapter":"9 Functionals","heading":"9 Functionals","text":"Attaching needed libraries:","code":"\nlibrary(purrr, warn.conflicts = FALSE)"},{"path":"functionals.html","id":"my-first-functional-map-exercises-9.2.6","chapter":"9 Functionals","heading":"9.1 My first functional: map() (Exercises 9.2.6)","text":"Q1. Use as_mapper() explore purrr generates anonymous functions integer, character, list helpers. helper allows extract attributes? Read documentation find .A1. Let’s handle two parts question separately.as_mapper() purrr-generated anonymous functions:Looking experimentation map() as_mapper(), can see , depending type input, as_mapper() creates extractor function using pluck().can extract attributes using purrr::attr_getter():Q2. map(1:3, ~ runif(2)) useful pattern generating random numbers, map(1:3, runif(2)) . ? Can explain returns result ?A2. shown as_mapper() outputs , second call appropriate generating random numbers translates pluck() function indices plucking taken randomly generated numbers, valid accessors get NULLs return.Q3. Use appropriate map() function :A3. Using appropriate map() function :Compute standard deviation every column numeric data frame:Compute standard deviation every numeric column mixed data frame:Compute number levels every factor data frame:Q4. following code simulates performance t-test non-normal data. Extract p-value test, visualise.A4.Extract p-value test:Visualise extracted p-values:Q5. following code uses map nested inside another map apply function every element nested list. fail, need make work?A5. function fails call effectively evaluates following:triple() one parameter (x), execution fails.fixed version:Q6. Use map() fit linear models mtcars dataset using formulas stored list:A6. Fitting linear models mtcars dataset using provided formulas:Q7. Fit model mpg ~ disp bootstrap replicates mtcars list , extract \\(R^2\\) model fit (Hint: can compute \\(R^2\\) summary().)A7. can done using map_dbl():","code":"\n# mapping by position -----------------------\n\nx <- list(1, list(2, 3, list(1, 2)))\n\nmap(x, 1)\n#> [[1]]\n#> [1] 1\n#> \n#> [[2]]\n#> [1] 2\nas_mapper(1)\n#> function (x, ...) \n#> pluck_raw(x, list(1), .default = NULL)\n#> <environment: 0x555ecae48430>\n\nmap(x, list(2, 1))\n#> [[1]]\n#> NULL\n#> \n#> [[2]]\n#> [1] 3\nas_mapper(list(2, 1))\n#> function (x, ...) \n#> pluck_raw(x, list(2, 1), .default = NULL)\n#> <environment: 0x555ecaf2f158>\n\n# mapping by name -----------------------\n\ny <- list(\n  list(m = \"a\", list(1, m = \"mo\")),\n  list(n = \"b\", list(2, n = \"no\"))\n)\n\nmap(y, \"m\")\n#> [[1]]\n#> [1] \"a\"\n#> \n#> [[2]]\n#> NULL\nas_mapper(\"m\")\n#> function (x, ...) \n#> pluck_raw(x, list(\"m\"), .default = NULL)\n#> <environment: 0x555ecb07d330>\n\n# mixing position and name\nmap(y, list(2, \"m\"))\n#> [[1]]\n#> [1] \"mo\"\n#> \n#> [[2]]\n#> NULL\nas_mapper(list(2, \"m\"))\n#> function (x, ...) \n#> pluck_raw(x, list(2, \"m\"), .default = NULL)\n#> <environment: 0x555ecb163178>\n\n# compact functions ----------------------------\n\nmap(y, ~ length(.x))\n#> [[1]]\n#> [1] 2\n#> \n#> [[2]]\n#> [1] 2\nas_mapper(~ length(.x))\n#> <lambda>\n#> function (..., .x = ..1, .y = ..2, . = ..1) \n#> length(.x)\n#> attr(,\"class\")\n#> [1] \"rlang_lambda_function\" \"function\"\npluck(Titanic, attr_getter(\"class\"))\n#> [1] \"table\"\nmap(1:3, ~ runif(2))\n#> [[1]]\n#> [1] 0.2180892 0.9876342\n#> \n#> [[2]]\n#> [1] 0.3484619 0.3810470\n#> \n#> [[3]]\n#> [1] 0.02098596 0.74972687\nas_mapper(~ runif(2))\n#> <lambda>\n#> function (..., .x = ..1, .y = ..2, . = ..1) \n#> runif(2)\n#> attr(,\"class\")\n#> [1] \"rlang_lambda_function\" \"function\"\n\nmap(1:3, runif(2))\n#> [[1]]\n#> [1] 1\n#> \n#> [[2]]\n#> [1] 2\n#> \n#> [[3]]\n#> [1] 3\nas_mapper(runif(2))\n#> function (x, ...) \n#> pluck_raw(x, list(0.597890264587477, 0.587997315218672), .default = NULL)\n#> <environment: 0x555ecc0f46f8>a) Compute the standard deviation of every column in a numeric data frame.\n\na) Compute the standard deviation of every numeric column in a mixed data frame. (Hint: you'll need to do it in two steps.)\n\na) Compute the number of levels for every factor in a data frame.\nmap_dbl(mtcars, sd)\n#>         mpg         cyl        disp          hp        drat \n#>   6.0269481   1.7859216 123.9386938  68.5628685   0.5346787 \n#>          wt        qsec          vs          am        gear \n#>   0.9784574   1.7869432   0.5040161   0.4989909   0.7378041 \n#>        carb \n#>   1.6152000\nkeep(iris, is.numeric) %>%\n  map_dbl(sd)\n#> Sepal.Length  Sepal.Width Petal.Length  Petal.Width \n#>    0.8280661    0.4358663    1.7652982    0.7622377\nmodify_if(dplyr::starwars, is.character, as.factor) %>%\n  keep(is.factor) %>%\n  map_int(~ length(levels(.)))\n#>       name hair_color skin_color  eye_color        sex \n#>         87         11         31         15          4 \n#>     gender  homeworld    species \n#>          2         48         37\ntrials <- map(1:100, ~ t.test(rpois(10, 10), rpois(7, 10)))\ntrials <- map(1:100, ~ t.test(rpois(10, 10), rpois(7, 10)))\n\n(p <- map_dbl(trials, \"p.value\"))\n#>   [1] 0.81695628 0.53177360 0.94750819 0.41026769 0.34655294\n#>   [6] 0.05300287 0.56479901 0.85936864 0.77517391 0.64321161\n#>  [11] 0.84462914 0.54144946 0.63070476 0.20325827 0.39824435\n#>  [16] 0.67052432 0.39932663 0.44437632 0.51645941 0.96578745\n#>  [21] 0.70219557 0.69931716 0.23946786 0.55100566 0.76028958\n#>  [26] 0.38105366 0.64544126 0.15379307 0.86945196 0.09965658\n#>  [31] 0.96425489 0.54239108 0.38985789 0.59019282 0.96247907\n#>  [36] 0.54997487 0.66111391 0.30961551 0.10897334 0.55049635\n#>  [41] 0.93882405 0.14836866 0.44307287 0.61583610 0.37284284\n#>  [46] 0.38559622 0.42935767 0.26059293 0.07831619 0.93768396\n#>  [51] 0.48459268 0.73571291 0.30288560 0.68521609 0.06374636\n#>  [56] 0.11007808 0.98758443 0.17831882 0.94471538 0.19711729\n#>  [61] 0.02094185 0.12370745 0.23247837 0.93842382 0.19160550\n#>  [66] 0.49005550 0.98146240 0.09034183 0.94912080 0.55857523\n#>  [71] 0.24692070 0.63658206 0.14290966 0.10309770 0.89516449\n#>  [76] 0.25660092 0.16943034 0.41199780 0.82721280 0.74017418\n#>  [81] 0.43724631 0.55944024 0.93615100 0.68788872 0.01416627\n#>  [86] 0.60120497 0.54125910 0.91581929 0.78949327 0.57887371\n#>  [91] 0.83217542 0.90108906 0.97474727 0.99129282 0.54436155\n#>  [96] 0.74159859 0.06534957 0.10834529 0.19737786 0.93750342\nplot(p)\n\nhist(p)\nx <- list(\n  list(1, c(3, 9)),\n  list(c(3, 6), 7, c(4, 7, 6))\n)\n\ntriple <- function(x) x * 3\nmap(x, map, .f = triple)\n#> Error in `map()`:\n#> ℹ In index: 1.\n#> Caused by error in `.f()`:\n#> ! unused argument (function (.x, .f, ..., .progress = FALSE) \n#> {\n#>     map_(\"list\", .x, .f, ..., .progress = .progress)\n#> })\nmap(.x = x, .f = ~ triple(x = .x, map))\nx <- list(\n  list(1, c(3, 9)),\n  list(c(3, 6), 7, c(4, 7, 6))\n)\n\ntriple <- function(x) x * 3\nmap(x, .f = ~ map(.x, ~ triple(.x)))\n#> [[1]]\n#> [[1]][[1]]\n#> [1] 3\n#> \n#> [[1]][[2]]\n#> [1]  9 27\n#> \n#> \n#> [[2]]\n#> [[2]][[1]]\n#> [1]  9 18\n#> \n#> [[2]][[2]]\n#> [1] 21\n#> \n#> [[2]][[3]]\n#> [1] 12 21 18\nformulas <- list(\n  mpg ~ disp,\n  mpg ~ I(1 / disp),\n  mpg ~ disp + wt,\n  mpg ~ I(1 / disp) + wt\n)\nformulas <- list(\n  mpg ~ disp,\n  mpg ~ I(1 / disp),\n  mpg ~ disp + wt,\n  mpg ~ I(1 / disp) + wt\n)\n\nmap(formulas, ~ lm(formula = ., data = mtcars))\n#> [[1]]\n#> \n#> Call:\n#> lm(formula = ., data = mtcars)\n#> \n#> Coefficients:\n#> (Intercept)         disp  \n#>    29.59985     -0.04122  \n#> \n#> \n#> [[2]]\n#> \n#> Call:\n#> lm(formula = ., data = mtcars)\n#> \n#> Coefficients:\n#> (Intercept)    I(1/disp)  \n#>       10.75      1557.67  \n#> \n#> \n#> [[3]]\n#> \n#> Call:\n#> lm(formula = ., data = mtcars)\n#> \n#> Coefficients:\n#> (Intercept)         disp           wt  \n#>    34.96055     -0.01772     -3.35083  \n#> \n#> \n#> [[4]]\n#> \n#> Call:\n#> lm(formula = ., data = mtcars)\n#> \n#> Coefficients:\n#> (Intercept)    I(1/disp)           wt  \n#>      19.024     1142.560       -1.798\nbootstrap <- function(df) {\n  df[sample(nrow(df), replace = TRUE), , drop = FALSE]\n}\n\nbootstraps <- map(1:10, ~ bootstrap(mtcars))\nbootstrap <- function(df) {\n  df[sample(nrow(df), replace = TRUE), , drop = FALSE]\n}\n\nbootstraps <- map(1:10, ~ bootstrap(mtcars))\n\nbootstraps %>%\n  map(~ lm(mpg ~ disp, data = .x)) %>%\n  map(summary) %>%\n  map_dbl(\"r.squared\")\n#>  [1] 0.7864562 0.8110818 0.7956331 0.7632399 0.7967824\n#>  [6] 0.7364226 0.7203027 0.6653252 0.7732780 0.6753329"},{"path":"functionals.html","id":"map-variants-exercises-9.4.6","chapter":"9 Functionals","heading":"9.2 Map variants (Exercises 9.4.6)","text":"Q1. Explain results modify(mtcars, 1).A1. modify() returns object type input. Since input data frame certain dimensions .f = 1 translates plucking first element column, returns data frame dimensions plucked element recycled across rows.Q2. Rewrite following code use iwalk() instead walk2(). advantages disadvantages?A2. Let’s first rewrite provided code using iwalk():advantage using iwalk() need now deal single variable (cyls) instead two (cyls paths).disadvantage code difficult reason :\nwalk2(), ’s explicit .x (= cyls) .y (= paths) correspond , iwalk() (.e., .x = cyls .y = names(cyls)) .y argument “invisible”.Q3. Explain following code transforms data frame using functions stored list.Compare contrast map2() approach map() approach:A3. map2() supplies functions stored trans anonymous functions via placeholder f, names columns specified mtcars[nm] supplied var argument anonymous function. Note function iterating indices vectors transformations column names.map() approach, function iterating indices vectors column names.latter approach can’t afford passing arguments placeholders anonymous function.Q4. write.csv() return, .e. happens use map2() instead walk2()?A4. use map2(), work, print NULLs console every list element.","code":"\nhead(modify(mtcars, 1))\n#>   mpg cyl disp  hp drat   wt  qsec vs am gear carb\n#> 1  21   6  160 110  3.9 2.62 16.46  0  1    4    4\n#> 2  21   6  160 110  3.9 2.62 16.46  0  1    4    4\n#> 3  21   6  160 110  3.9 2.62 16.46  0  1    4    4\n#> 4  21   6  160 110  3.9 2.62 16.46  0  1    4    4\n#> 5  21   6  160 110  3.9 2.62 16.46  0  1    4    4\n#> 6  21   6  160 110  3.9 2.62 16.46  0  1    4    4\ncyls <- split(mtcars, mtcars$cyl)\npaths <- file.path(temp, paste0(\"cyl-\", names(cyls), \".csv\"))\nwalk2(cyls, paths, write.csv)\ncyls <- split(mtcars, mtcars$cyl)\nnames(cyls) <- file.path(temp, paste0(\"cyl-\", names(cyls), \".csv\"))\niwalk(cyls, ~ write.csv(.x, .y))\ntrans <- list(\n  disp = function(x) x * 0.0163871,\n  am = function(x) factor(x, labels = c(\"auto\", \"manual\"))\n)\n\nnm <- names(trans)\nmtcars[nm] <- map2(trans, mtcars[nm], function(f, var) f(var))\nmtcars[nm] <- map(nm, ~ trans[[.x]](mtcars[[.x]]))\ntrans <- list(\n  disp = function(x) x * 0.0163871,\n  am = function(x) factor(x, labels = c(\"auto\", \"manual\"))\n)\n\nnm <- names(trans)\nmtcars[nm] <- map2(trans, mtcars[nm], function(f, var) f(var))\nmtcars[nm] <- map(nm, ~ trans[[.x]](mtcars[[.x]]))\nwithr::with_tempdir(\n  code = {\n    ls <- split(mtcars, mtcars$cyl)\n    nm <- names(ls)\n    map2(ls, nm, write.csv)\n  }\n)\n#> $`4`\n#> NULL\n#> \n#> $`6`\n#> NULL\n#> \n#> $`8`\n#> NULL"},{"path":"functionals.html","id":"predicate-functionals-exercises-9.6.3","chapter":"9 Functionals","heading":"9.3 Predicate functionals (Exercises 9.6.3)","text":"Q1. isn’t .na() predicate function? base R function closest predicate version .na()?A1. mentioned docs:predicate function returns single TRUE FALSE..na() function return logical scalar, instead returns vector thus isn’t predicate function.closest equivalent predicate function base-R anyNA() function.Q2. simple_reduce() problem x length 0 length 1. Describe source problem might go fixing .A2. supplied function struggles inputs length 0 1 function tries subscript --bound values.problem can solved adding init argument, supplies default initial value:Let’s try :Depending function, can provide different init argument:Q3. Implement span() function Haskell: given list x predicate function f, span(x, f) returns location longest sequential run elements predicate true. (Hint: might find rle() helpful.)A3. Implementation span():Testing :Testing twice:Q4. Implement arg_max(). take function vector inputs, return elements input function returns highest value. example, arg_max(-10:5, function(x) x ^ 2) return -10. arg_max(-5:5, function(x) x ^ 2) return c(-5, 5). Also implement matching arg_min() function.A4. implementations specified functions:Implementing arg_max()Implementing arg_min()Q5. function scales vector falls range [0, 1]. apply every column data frame? apply every numeric column data frame?A5. use purrr package apply function. Key thing keep mind data frame list atomic vectors equal length.Applying function every column data frame: use anscombe example since numeric columns.Applying function every numeric column data frame: use iris example since columns numeric type.","code":"\n# contrast the following behavior of predicate functions\nis.character(c(\"x\", 2))\n#> [1] TRUE\nis.null(c(3, NULL))\n#> [1] FALSE\n\n# with this behavior\nis.na(c(NA, 1))\n#> [1]  TRUE FALSE\nanyNA(c(NA, 1))\n#> [1] TRUE\nsimple_reduce <- function(x, f) {\n  out <- x[[1]]\n  for (i in seq(2, length(x))) {\n    out <- f(out, x[[i]])\n  }\n  out\n}\nsimple_reduce(numeric(), sum)\n#> Error in x[[1]]: subscript out of bounds\nsimple_reduce(1, sum)\n#> Error in x[[i]]: subscript out of bounds\nsimple_reduce(1:3, sum)\n#> [1] 6\nsimple_reduce2 <- function(x, f, init = 0) {\n  # initializer will become the first value\n  if (length(x) == 0L) {\n    return(init)\n  }\n\n  if (length(x) == 1L) {\n    return(x[[1L]])\n  }\n\n  out <- x[[1]]\n\n  for (i in seq(2, length(x))) {\n    out <- f(out, x[[i]])\n  }\n\n  out\n}\nsimple_reduce2(numeric(), sum)\n#> [1] 0\nsimple_reduce2(1, sum)\n#> [1] 1\nsimple_reduce2(1:3, sum)\n#> [1] 6\nsimple_reduce2(numeric(), `*`, init = 1)\n#> [1] 1\nsimple_reduce2(1, `*`, init = 1)\n#> [1] 1\nsimple_reduce2(1:3, `*`, init = 1)\n#> [1] 6\nspan <- function(x, f) {\n  running_lengths <- purrr::map_lgl(x, ~ f(.x)) %>% rle()\n\n  df <- dplyr::tibble(\n    \"lengths\" = running_lengths$lengths,\n    \"values\" = running_lengths$values\n  ) %>%\n    dplyr::mutate(rowid = dplyr::row_number()) %>%\n    dplyr::filter(values)\n\n  # no sequence where condition is `TRUE`\n  if (nrow(df) == 0L) {\n    return(integer())\n  }\n\n  # only single sequence where condition is `TRUE`\n  if (nrow(df) == 1L) {\n    return((df$rowid):(df$lengths - 1 + df$rowid))\n  }\n\n  # multiple sequences where condition is `TRUE`; select max one\n  if (nrow(df) > 1L) {\n    df <- dplyr::filter(df, lengths == max(lengths))\n    return((df$rowid):(df$lengths - 1 + df$rowid))\n  }\n}\nspan(c(0, 0, 0, 0, 0), is.na)\n#> integer(0)\nspan(c(NA, 0, NA, NA, NA), is.na)\n#> [1] 3 4 5\nspan(c(NA, 0, 0, 0, 0), is.na)\n#> [1] 1\nspan(c(NA, NA, 0, 0, 0), is.na)\n#> [1] 1 2\nspan(c(3, 1, 2, 4, 5, 6), function(x) x > 3)\n#> [1] 2 3 4\nspan(c(3, 1, 2, 4, 5, 6), function(x) x > 9)\n#> integer(0)\nspan(c(3, 1, 2, 4, 5, 6), function(x) x == 3)\n#> [1] 1\nspan(c(3, 1, 2, 4, 5, 6), function(x) x %in% c(2, 4))\n#> [1] 2 3\narg_max <- function(.x, .f) {\n  df <- dplyr::tibble(\n    original = .x,\n    transformed = purrr::map_dbl(.x, .f)\n  )\n\n  dplyr::filter(df, transformed == max(transformed))[[\"original\"]]\n}\n\narg_max(-10:5, function(x) x^2)\n#> [1] -10\narg_max(-5:5, function(x) x^2)\n#> [1] -5  5\narg_min <- function(.x, .f) {\n  df <- dplyr::tibble(\n    original = .x,\n    transformed = purrr::map_dbl(.x, .f)\n  )\n\n  dplyr::filter(df, transformed == min(transformed))[[\"original\"]]\n}\n\narg_min(-10:5, function(x) x^2)\n#> [1] 0\narg_min(-5:5, function(x) x^2)\n#> [1] 0\nscale01 <- function(x) {\n  rng <- range(x, na.rm = TRUE)\n  (x - rng[1]) / (rng[2] - rng[1])\n}\npurrr::map_df(head(anscombe), .f = scale01)\n#> # A tibble: 6 × 8\n#>      x1    x2    x3    x4    y1     y2     y3    y4\n#>   <dbl> <dbl> <dbl> <dbl> <dbl>  <dbl>  <dbl> <dbl>\n#> 1 0.333 0.333 0.333   NaN 0.362 0.897  0.116  0.266\n#> 2 0     0     0       NaN 0     0.0345 0      0    \n#> 3 0.833 0.833 0.833   NaN 0.209 0.552  1      0.633\n#> 4 0.167 0.167 0.167   NaN 0.618 0.578  0.0570 1    \n#> 5 0.5   0.5   0.5     NaN 0.458 1      0.174  0.880\n#> 6 1     1     1       NaN 1     0      0.347  0.416\npurrr::modify_if(head(iris), .p = is.numeric, .f = scale01)\n#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species\n#> 1        0.625   0.5555556         0.25           0  setosa\n#> 2        0.375   0.0000000         0.25           0  setosa\n#> 3        0.125   0.2222222         0.00           0  setosa\n#> 4        0.000   0.1111111         0.50           0  setosa\n#> 5        0.500   0.6666667         0.25           0  setosa\n#> 6        1.000   1.0000000         1.00           1  setosa"},{"path":"functionals.html","id":"base-functionals-exercises-9.7.3","chapter":"9 Functionals","heading":"9.4 Base functionals (Exercises 9.7.3)","text":"Q1. apply() arrange output? Read documentation perform experiments.A1. Let’s prepare array apply function different margins:can seen, apply() returns outputs organised first margins operated , results.Q2. eapply() rapply() ? purrr equivalents?A2. Let’s consider one--one.eapply()mentioned documentation:eapply() applies FUN named values environment returns results list.example:purrr doesn’t function iterate environments.rapply()rapply() recursive version lapply flexibility result structured (= “..”).example:purrr something similar modify_tree().Q3. Challenge: read fixed point algorithm. Complete exercises using R.A3. mentioned suggested reading material:number \\(x\\) called fixed point function \\(f\\) \\(x\\) satisfies equation \\(f(x) = x\\). functions \\(f\\) can locate fixed point beginning initial guess applying \\(f\\) repeatedly, \\(f(x), f(f(x)), f(f(f(x))), ...\\) value change much. Using idea, can devise procedure fixed-point takes inputs function initial guess produces approximation fixed point function.Let’s first implement fixed-point algorithm:Let’s check works expected:solve one exercise reading material. Rest beyond scope solution manual.Show golden ratio \\(\\phi\\) fixed point transformation \\(x \\mapsto 1 + 1/x\\), use fact compute \\(\\phi\\) means fixed-point procedure.","code":"\n(m <- as.array(table(mtcars$cyl, mtcars$am, mtcars$vs)))\n#> , ,  = 0\n#> \n#>    \n#>     auto manual\n#>   4    0      1\n#>   6    0      3\n#>   8   12      2\n#> \n#> , ,  = 1\n#> \n#>    \n#>     auto manual\n#>   4    3      7\n#>   6    4      0\n#>   8    0      0\n\n# rows\napply(m, 1, function(x) x^2)\n#>       \n#>         4  6   8\n#>   [1,]  0  0 144\n#>   [2,]  1  9   4\n#>   [3,]  9 16   0\n#>   [4,] 49  0   0\n\n# columns\napply(m, 2, function(x) x^2)\n#>       \n#>        auto manual\n#>   [1,]    0      1\n#>   [2,]    0      9\n#>   [3,]  144      4\n#>   [4,]    9     49\n#>   [5,]   16      0\n#>   [6,]    0      0\n\n# rows and columns\napply(m, c(1, 2), function(x) x^2)\n#> , ,  = auto\n#> \n#>    \n#>     4  6   8\n#>   0 0  0 144\n#>   1 9 16   0\n#> \n#> , ,  = manual\n#> \n#>    \n#>      4 6 8\n#>   0  1 9 4\n#>   1 49 0 0\nlibrary(rlang)\n#> \n#> Attaching package: 'rlang'\n#> The following objects are masked from 'package:purrr':\n#> \n#>     %@%, flatten, flatten_chr, flatten_dbl,\n#>     flatten_int, flatten_lgl, flatten_raw, invoke,\n#>     splice\n#> The following object is masked from 'package:magrittr':\n#> \n#>     set_names\n\ne <- env(\"x\" = 1, \"y\" = 2)\nrlang::env_print(e)\n#> <environment: 0x555ec937cf20>\n#> Parent: <environment: global>\n#> Bindings:\n#> • x: <dbl>\n#> • y: <dbl>\n\neapply(e, as.character)\n#> $x\n#> [1] \"1\"\n#> \n#> $y\n#> [1] \"2\"\nX <- list(list(a = TRUE, b = list(c = c(4L, 3.2))), d = 9.0)\n\nrapply(X, as.character, classes = \"numeric\", how = \"replace\")\n#> [[1]]\n#> [[1]]$a\n#> [1] TRUE\n#> \n#> [[1]]$b\n#> [[1]]$b$c\n#> [1] \"4\"   \"3.2\"\n#> \n#> \n#> \n#> $d\n#> [1] \"9\"\nX <- list(list(a = TRUE, b = list(c = c(4L, 3.2))), d = 9.0)\n\npurrr::modify_tree(X, leaf = length)\n#> [[1]]\n#> [[1]]$a\n#> [1] 1\n#> \n#> [[1]]$b\n#> [[1]]$b$c\n#> [1] 2\n#> \n#> \n#> \n#> $d\n#> [1] 1\nclose_enough <- function(x1, x2, tolerance = 0.001) {\n  if (abs(x1 - x2) < tolerance) {\n    return(TRUE)\n  } else {\n    return(FALSE)\n  }\n}\n\nfind_fixed_point <- function(.f, .guess, tolerance = 0.001) {\n  .next <- .f(.guess)\n  is_close_enough <- close_enough(.next, .guess, tol = tolerance)\n\n  if (is_close_enough) {\n    return(.next)\n  } else {\n    find_fixed_point(.f, .next, tolerance)\n  }\n}\nfind_fixed_point(cos, 1.0)\n#> [1] 0.7387603\n\n# cos(x) = x\ncos(find_fixed_point(cos, 1.0))\n#> [1] 0.7393039\ngolden_ratio_f <- function(x) 1 + (1 / x)\n\nfind_fixed_point(golden_ratio_f, 1.0)\n#> [1] 1.618182"},{"path":"functionals.html","id":"session-information-7","chapter":"9 Functionals","heading":"9.5 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr         1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  highr         0.10    2022-12-22 [1] RSPM\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  purrr       * 1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"function-factories.html","id":"function-factories","chapter":"10 Function factories","heading":"10 Function factories","text":"Attaching needed libraries:","code":"\nlibrary(rlang, warn.conflicts = FALSE)\nlibrary(ggplot2, warn.conflicts = FALSE)"},{"path":"function-factories.html","id":"factory-fundamentals-exercises-10.2.6","chapter":"10 Function factories","heading":"10.1 Factory fundamentals (Exercises 10.2.6)","text":"Q1. definition force() simple:better force(x) instead just x?A1. Due lazy evaluation, argument function won’t evaluated value needed. sometimes may want eager evaluation, using force() makes intent clearer.Q2. Base R contains two function factories, approxfun() ecdf(). Read documentation experiment figure functions return.A2. two function factories-approxfun()function factory returns function performing linear (constant) interpolation.ecdf()function factory computes empirical cumulative distribution function.Q3. Create function pick() takes index, , argument returns function argument x subsets x .A3. write desired function, just need make sure argument eagerly evaluated.Testing specified test cases:Q4. Create function creates functions compute ithcentral moment numeric vector. can test running following code:A4. following function satisfied specified requirements:Testing specified test cases:Q5. happens don’t use closure? Make predictions, verify code .A5. case closures used context, counts stored global variable, can modified processes even deleted.Q6. happens use <- instead <<-? Make predictions, verify code .A6. case, function always return 1.","code":"\nforce\n#> function (x) \n#> x\n#> <bytecode: 0x556cfd712ab0>\n#> <environment: namespace:base>\nx <- 1:10\ny <- rnorm(10)\nf <- approxfun(x, y)\nf\n#> function (v) \n#> .approxfun(x, y, v, method, yleft, yright, f, na.rm)\n#> <bytecode: 0x556d01983cb0>\n#> <environment: 0x556d01982ba0>\nf(x)\n#>  [1] -0.7786629 -0.3894764 -2.0337983 -0.9823731  0.2478901\n#>  [6] -2.1038646 -0.3814180  2.0749198  1.0271384  0.4730142\ncurve(f(x), 0, 11)\nx <- rnorm(12)\nf <- ecdf(x)\nf\n#> Empirical CDF \n#> Call: ecdf(x)\n#>  x[1:12] = -1.8793, -1.3221, -1.2392,  ..., 1.1604, 1.7956\nf(seq(-2, 2, by = 0.1))\n#>  [1] 0.00000000 0.00000000 0.08333333 0.08333333 0.08333333\n#>  [6] 0.08333333 0.08333333 0.16666667 0.25000000 0.25000000\n#> [11] 0.33333333 0.33333333 0.33333333 0.41666667 0.41666667\n#> [16] 0.41666667 0.41666667 0.50000000 0.58333333 0.58333333\n#> [21] 0.66666667 0.75000000 0.75000000 0.75000000 0.75000000\n#> [26] 0.75000000 0.75000000 0.75000000 0.75000000 0.83333333\n#> [31] 0.83333333 0.83333333 0.91666667 0.91666667 0.91666667\n#> [36] 0.91666667 0.91666667 0.91666667 1.00000000 1.00000000\n#> [41] 1.00000000\npick(1)(x)\n# should be equivalent to\nx[[1]]\n\nlapply(mtcars, pick(5))\n# should be equivalent to\nlapply(mtcars, function(x) x[[5]])\npick <- function(i) {\n  force(i)\n  function(x) x[[i]]\n}\nx <- list(\"a\", \"b\", \"c\")\nidentical(x[[1]], pick(1)(x))\n#> [1] TRUE\n\nidentical(\n  lapply(mtcars, pick(5)),\n  lapply(mtcars, function(x) x[[5]])\n)\n#> [1] TRUE\nm1 <- moment(1)\nm2 <- moment(2)\nx <- runif(100)\nstopifnot(all.equal(m1(x), 0))\nstopifnot(all.equal(m2(x), var(x) * 99 / 100))\nmoment <- function(k) {\n  force(k)\n\n  function(x) (sum((x - mean(x))^k)) / length(x)\n}\nm1 <- moment(1)\nm2 <- moment(2)\nx <- runif(100)\n\nstopifnot(all.equal(m1(x), 0))\nstopifnot(all.equal(m2(x), var(x) * 99 / 100))\ni <- 0\nnew_counter2 <- function() {\n  i <<- i + 1\n  i\n}\nnew_counter2()\n#> [1] 1\n\nnew_counter2()\n#> [1] 2\n\nnew_counter2()\n#> [1] 3\n\ni <- 20\nnew_counter2()\n#> [1] 21\nnew_counter3 <- function() {\n  i <- 0\n  function() {\n    i <- i + 1\n    i\n  }\n}\nnew_counter3()\n#> function() {\n#>     i <- i + 1\n#>     i\n#>   }\n#> <environment: 0x556d01b28e68>\n\nnew_counter3()\n#> function() {\n#>     i <- i + 1\n#>     i\n#>   }\n#> <bytecode: 0x556d01ced600>\n#> <environment: 0x556d01ba0c80>"},{"path":"function-factories.html","id":"graphical-factories-exercises-10.3.4","chapter":"10 Function factories","heading":"10.2 Graphical factories (Exercises 10.3.4)","text":"Q1. Compare contrast ggplot2::label_bquote() scales::number_format().A1. compare contrast, let’s first look source code functions:ggplot2::label_bquote()scales::number_format()functions return formatting functions used style facets labels labels desired format ggplot2 plots.example, using plotmath expression facet label:display axes labels desired format:ggplot2::label_bquote() adds additional class returned function.scales::number_format() function simple pass-method forces evaluation parameters passes underlying scales::number() function.","code":"\nggplot2::label_bquote\n#> function (rows = NULL, cols = NULL, default) \n#> {\n#>     cols_quoted <- substitute(cols)\n#>     rows_quoted <- substitute(rows)\n#>     call_env <- env_parent()\n#>     fun <- function(labels) {\n#>         quoted <- resolve_labeller(rows_quoted, cols_quoted, \n#>             labels)\n#>         if (is.null(quoted)) {\n#>             return(label_value(labels))\n#>         }\n#>         evaluate <- function(...) {\n#>             params <- list(...)\n#>             params <- as_environment(params, call_env)\n#>             eval(substitute(bquote(expr, params), list(expr = quoted)))\n#>         }\n#>         list(inject(mapply(evaluate, !!!labels, SIMPLIFY = FALSE)))\n#>     }\n#>     structure(fun, class = \"labeller\")\n#> }\n#> <bytecode: 0x556d01e7f0b0>\n#> <environment: namespace:ggplot2>\nscales::number_format\n#> function (accuracy = NULL, scale = 1, prefix = \"\", suffix = \"\", \n#>     big.mark = \" \", decimal.mark = \".\", style_positive = c(\"none\", \n#>         \"plus\", \"space\"), style_negative = c(\"hyphen\", \"minus\", \n#>         \"parens\"), scale_cut = NULL, trim = TRUE, ...) \n#> {\n#>     force_all(accuracy, scale, prefix, suffix, big.mark, decimal.mark, \n#>         style_positive, style_negative, scale_cut, trim, ...)\n#>     function(x) {\n#>         number(x, accuracy = accuracy, scale = scale, prefix = prefix, \n#>             suffix = suffix, big.mark = big.mark, decimal.mark = decimal.mark, \n#>             style_positive = style_positive, style_negative = style_negative, \n#>             scale_cut = scale_cut, trim = trim, ...)\n#>     }\n#> }\n#> <bytecode: 0x556d020cfbc8>\n#> <environment: namespace:scales>\nlibrary(ggplot2)\n\np <- ggplot(mtcars, aes(wt, mpg)) +\n  geom_point()\np + facet_grid(. ~ vs, labeller = label_bquote(cols = alpha^.(vs)))\nlibrary(scales)\n\nggplot(mtcars, aes(wt, mpg)) +\n  geom_point() +\n  scale_y_continuous(labels = number_format(accuracy = 0.01, decimal.mark = \",\"))"},{"path":"function-factories.html","id":"statistical-factories-exercises-10.4.4","chapter":"10 Function factories","heading":"10.3 Statistical factories (Exercises 10.4.4)","text":"Q1. boot_model(), don’t need force evaluation df model?A1. don’t need force evaluation df model arguments automatically evaluated lm():Q2. might formulate Box-Cox transformation like ?A2. see formulate transformation like , can compare one mentioned book:Let’s look one example :can seen:boxcox2(), can vary x value lambda, whilein boxcox3(), can vary lambda vector.Thus, boxcox3() can handy exploring different transformations across inputs.Q3. don’t need worry boot_permute() stores copy data inside function generates?A3. look source code generated function factory, notice exact data frame (mtcars) referenced:don’t need worry copy made df function environment points memory address data frame. can confirm comparing memory addresses:can also check values bindings entered function factory:Q4. much time ll_poisson2() save compared ll_poisson1()? Use bench::mark() see much faster optimisation occurs. changing length x change results?A4. Let’s first compare performance functions example book:can seen, second version much faster first version.can also vary length vector confirm across wide range vector lengths, performance advantage observed.","code":"\nboot_model <- function(df, formula) {\n  mod <- lm(formula, data = df)\n  fitted <- unname(fitted(mod))\n  resid <- unname(resid(mod))\n  rm(mod)\n\n  function() {\n    fitted + sample(resid)\n  }\n}\nboxcox3 <- function(x) {\n  function(lambda) {\n    if (lambda == 0) {\n      log(x)\n    } else {\n      (x^lambda - 1) / lambda\n    }\n  }\n}\nboxcox2 <- function(lambda) {\n  if (lambda == 0) {\n    function(x) log(x)\n  } else {\n    function(x) (x^lambda - 1) / lambda\n  }\n}\nboxcox2(1)\n#> function(x) (x^lambda - 1) / lambda\n#> <environment: 0x556cfdac76a0>\n\nboxcox3(mtcars$wt)\n#> function(lambda) {\n#>     if (lambda == 0) {\n#>       log(x)\n#>     } else {\n#>       (x^lambda - 1) / lambda\n#>     }\n#>   }\n#> <environment: 0x556cfefae4d0>\nboot_permute <- function(df, var) {\n  n <- nrow(df)\n  force(var)\n\n  function() {\n    col <- df[[var]]\n    col[sample(n, replace = TRUE)]\n  }\n}\n\nboot_permute(mtcars, \"mpg\")\n#> function() {\n#>     col <- df[[var]]\n#>     col[sample(n, replace = TRUE)]\n#>   }\n#> <environment: 0x556d00bef850>\nboot_permute_env <- rlang::fn_env(boot_permute(mtcars, \"mpg\"))\nrlang::env_print(boot_permute_env)\n#> <environment: 0x556d0181e450>\n#> Parent: <environment: global>\n#> Bindings:\n#> • n: <int>\n#> • df: <df[,11]>\n#> • var: <chr>\n\nidentical(\n  lobstr::obj_addr(boot_permute_env$df),\n  lobstr::obj_addr(mtcars)\n)\n#> [1] TRUE\nidentical(boot_permute_env$df, mtcars)\n#> [1] TRUE\nidentical(boot_permute_env$var, \"mpg\")\n#> [1] TRUE\nll_poisson1 <- function(x) {\n  n <- length(x)\n\n  function(lambda) {\n    log(lambda) * sum(x) - n * lambda - sum(lfactorial(x))\n  }\n}\n\nll_poisson2 <- function(x) {\n  n <- length(x)\n  sum_x <- sum(x)\n  c <- sum(lfactorial(x))\n\n  function(lambda) {\n    log(lambda) * sum_x - n * lambda - c\n  }\n}\n\nx1 <- c(41, 30, 31, 38, 29, 24, 30, 29, 31, 38)\n\nbench::mark(\n  \"LL1\" = optimise(ll_poisson1(x1), c(0, 100), maximum = TRUE),\n  \"LL2\" = optimise(ll_poisson2(x1), c(0, 100), maximum = TRUE)\n)\n#> # A tibble: 2 × 6\n#>   expression      min   median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1 LL1          27.5µs   29.3µs    33174.    12.9KB     43.2\n#> 2 LL2          14.6µs   15.3µs    62143.        0B     49.8\ngenerate_ll_benches <- function(n) {\n  x_vec <- sample.int(n, n)\n\n  bench::mark(\n    \"LL1\" = optimise(ll_poisson1(x_vec), c(0, 100), maximum = TRUE),\n    \"LL2\" = optimise(ll_poisson2(x_vec), c(0, 100), maximum = TRUE)\n  )[1:4] %>%\n    dplyr::mutate(length = n, .before = expression)\n}\n\n(df_bench <- purrr::map_dfr(\n  .x = c(10, 20, 50, 100, 1000),\n  .f = ~ generate_ll_benches(n = .x)\n))\n#> # A tibble: 10 × 5\n#>    length expression      min   median `itr/sec`\n#>     <dbl> <bch:expr> <bch:tm> <bch:tm>     <dbl>\n#>  1     10 LL1          38.9µs   40.9µs    23209.\n#>  2     10 LL2          17.2µs     18µs    54710.\n#>  3     20 LL1          42.1µs     44µs    22427.\n#>  4     20 LL2          16.9µs   17.7µs    55356.\n#>  5     50 LL1          46.2µs   47.9µs    20558.\n#>  6     50 LL2          16.2µs   17.2µs    57361.\n#>  7    100 LL1          61.9µs   63.6µs    15410.\n#>  8    100 LL2          17.6µs   18.4µs    53603.\n#>  9   1000 LL1         854.3µs  872.3µs     1092.\n#> 10   1000 LL2          54.6µs   55.9µs    17653.\n\nggplot(\n  df_bench,\n  aes(\n    x = as.numeric(length),\n    y = median,\n    group = as.character(expression),\n    color = as.character(expression)\n  )\n) +\n  geom_point() +\n  geom_line() +\n  labs(\n    x = \"Vector length\",\n    y = \"Median Execution Time\",\n    colour = \"Function used\"\n  )"},{"path":"function-factories.html","id":"function-factories-functionals-exercises-10.5.1","chapter":"10 Function factories","heading":"10.4 Function factories + functionals (Exercises 10.5.1)","text":"Q1. following commands equivalent (x, f(z))?A1. depends whether () used data frame list.Q2. Compare contrast effects env_bind() vs. attach() following code.A2. Let’s compare contrast effects env_bind() vs. attach().attach() adds funs search path. Since functions names functions {base} package, attached names mask ones {base} package.env_bind() adds functions funs global environment, instead masking names {base} package.Note \"funs\" output.","code":"(a) `x$f(x$z)`.\n(b) `f(x$z)`.\n(c) `x$f(z)`.\n(d) `f(z)`.\n(e) It depends.\nf <- mean\nz <- 1\nx <- list(f = mean, z = 1)\n\nidentical(with(x, f(z)), x$f(x$z))\n#> [1] TRUE\n\nidentical(with(x, f(z)), f(x$z))\n#> [1] TRUE\n\nidentical(with(x, f(z)), x$f(z))\n#> [1] TRUE\n\nidentical(with(x, f(z)), f(z))\n#> [1] TRUE\nfuns <- list(\n  mean = function(x) mean(x, na.rm = TRUE),\n  sum = function(x) sum(x, na.rm = TRUE)\n)\n\nattach(funs)\n#> The following objects are masked from package:base:\n#> \n#>     mean, sum\n\nmean\n#> function(x) mean(x, na.rm = TRUE)\nhead(search())\n#> [1] \".GlobalEnv\"       \"funs\"             \"package:scales\"  \n#> [4] \"package:ggplot2\"  \"package:rlang\"    \"package:magrittr\"\n\nmean <- function(x) stop(\"Hi!\")\nmean\n#> function(x) stop(\"Hi!\")\nhead(search())\n#> [1] \".GlobalEnv\"       \"funs\"             \"package:scales\"  \n#> [4] \"package:ggplot2\"  \"package:rlang\"    \"package:magrittr\"\n\ndetach(funs)\nenv_bind(globalenv(), !!!funs)\nmean\n#> function(x) mean(x, na.rm = TRUE)\n\nmean <- function(x) stop(\"Hi!\")\nmean\n#> function(x) stop(\"Hi!\")\nenv_unbind(globalenv(), names(funs))"},{"path":"function-factories.html","id":"session-information-8","chapter":"10 Function factories","heading":"10.5 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bench         1.1.3   2023-05-04 [1] RSPM\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  colorspace    2.1-0   2023-01-23 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr         1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  farver        2.1.1   2022-07-06 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  ggplot2     * 3.4.4   2023-10-12 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  grid          4.3.2   2023-11-01 [3] local\n#>  gtable        0.3.4   2023-08-21 [1] RSPM\n#>  highr         0.10    2022-12-22 [1] RSPM\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  labeling      0.4.3   2023-08-29 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  lobstr        1.1.2   2022-06-22 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  munsell       0.5.0   2018-06-12 [1] RSPM\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  profmem       0.6.0   2020-12-13 [1] RSPM\n#>  purrr         1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  scales      * 1.3.0   2023-11-28 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"function-operators.html","id":"function-operators","chapter":"11 Function operators","heading":"11 Function operators","text":"Attaching needed libraries:","code":"\nlibrary(purrr, warn.conflicts = FALSE)"},{"path":"function-operators.html","id":"existing-function-operators-exercises-11.2.3","chapter":"11 Function operators","heading":"11.1 Existing function operators (Exercises 11.2.3)","text":"Q1. Base R provides function operator form Vectorize(). ? might use ?A1. Vectorize() function creates function vectorizes action provided function specified arguments (.e., acts element vector). see utility solving problem otherwise difficult solve.problem find indices matching numeric values given threshold creating hybrid following functions:%% (doesn’t provide way provide tolerance comparing numeric values),dplyr::near() (vectorized element-wise thus expects two vectors equal length)Let’s use :Note needed create new function neither existing functions want.solved complex task using Vectorize() function!Q2. Read source code possibly(). work?A2. Let’s look source code function:Looking code, can see possibly():uses tryCatch() error handlinghas parameter otherwise specify default value case error occurshas parameter quiet suppress error message (needed)Q3. Read source code safely(). work?A3. Let’s look source code function:Looking code, can see safely():uses list save results (function executes successfully) error (fails)uses tryCatch() error handlinghas parameter otherwise specify default value case error occurshas parameter quiet suppress error message (needed)","code":"\nwhich_near <- function(x, y, tolerance) {\n  # Vectorize `dplyr::near()` function only over the `y` argument.\n  # `Vectorize()` is a function operator and will return a function.\n  customNear <- Vectorize(dplyr::near, vectorize.args = c(\"y\"), SIMPLIFY = FALSE)\n\n  # Apply the vectorized function to vector arguments and then check where the\n  # comparisons are equal (i.e. `TRUE`) using `which()`.\n  #\n  # Use `compact()` to remove empty elements from the resulting list.\n  index_list <- purrr::compact(purrr::map(customNear(x, y, tol = tolerance), which))\n\n  # If there are any matches, return the indices as an atomic vector of integers.\n  if (length(index_list) > 0L) {\n    index_vector <- purrr::simplify(index_list, \"integer\")\n    return(index_vector)\n  }\n\n  # If there are no matches\n  return(integer(0L))\n}\nx1 <- c(2.1, 3.3, 8.45, 8, 6)\nx2 <- c(6, 8.40, 3)\n\nwhich_near(x1, x2, tolerance = 0.1)\n#> [1] 5 3\nwhich(x1 %in% x2)\n#> [1] 5\n\nwhich(dplyr::near(x1, x2, tol = 0.1))\n#> Warning in x - y: longer object length is not a multiple of\n#> shorter object length\n#> integer(0)\npossibly\n#> function (.f, otherwise = NULL, quiet = TRUE) \n#> {\n#>     .f <- as_mapper(.f)\n#>     force(otherwise)\n#>     check_bool(quiet)\n#>     function(...) {\n#>         tryCatch(.f(...), error = function(e) {\n#>             if (!quiet) \n#>                 message(\"Error: \", conditionMessage(e))\n#>             otherwise\n#>         })\n#>     }\n#> }\n#> <bytecode: 0x55d90290c0b8>\n#> <environment: namespace:purrr>\nsafely\n#> function (.f, otherwise = NULL, quiet = TRUE) \n#> {\n#>     .f <- as_mapper(.f)\n#>     force(otherwise)\n#>     check_bool(quiet)\n#>     function(...) capture_error(.f(...), otherwise, quiet)\n#> }\n#> <bytecode: 0x55d902aa9948>\n#> <environment: namespace:purrr>\n\npurrr:::capture_error\n#> function (code, otherwise = NULL, quiet = TRUE) \n#> {\n#>     tryCatch(list(result = code, error = NULL), error = function(e) {\n#>         if (!quiet) \n#>             message(\"Error: \", conditionMessage(e))\n#>         list(result = otherwise, error = e)\n#>     })\n#> }\n#> <bytecode: 0x55d902b04180>\n#> <environment: namespace:purrr>"},{"path":"function-operators.html","id":"case-study-creating-your-own-function-operators-exercises-11.3.1","chapter":"11 Function operators","heading":"11.2 Case study: Creating your own function operators (Exercises 11.3.1)","text":"Q1. Weigh pros cons download.file %>% dot_every(10) %>% delay_by(0.1) versus download.file %>% delay_by(0.1) %>% dot_every(10).A1. Although chains piped operations produce number dots need amount time, subtle difference .download.file %>% dot_every(10) %>% delay_by(0.1), printing dot also delayed, first dot printed 10th URL download starts.download.file %>% delay_by(0.1) %>% dot_every(10), first dot printed 9th download finished, 10th download starts short delay.Q2. memoise download.file()? ?A2. Since download.file() meant download files Internet, memoising recommended following reasons:Memoization helpful giving input function returns output. necessarily case webpages since constantly change, may continue “download” outdated version webpage.Memoization helpful giving input function returns output. necessarily case webpages since constantly change, may continue “download” outdated version webpage.Memoization works caching results, can take significant amount memory.Memoization works caching results, can take significant amount memory.Q3. Create function operator reports whenever file created deleted working directory, using dir() setdiff(). global function effects might want track?A3. First, let’s create helper functions compare print added removed filenames:can write function operator use create functions necessary tracking:Let’s try :global function effects might want track:working directoryenvironment variablesconnectionslibrary pathsgraphics devicesetc.Q4. Write function operator logs timestamp message file every time function run.A4. following function operator logs timestamp message file every time function run:Q5. Modify delay_by() instead delaying fixed amount time, ensures certain amount time elapsed since function last called. , called g <- delay_by(1, f); g(); Sys.sleep(2); g() shouldn’t extra delay.A5. Modified version function meeting specified requirements:","code":"\nprint_multiple_entries <- function(header, entries) {\n  message(paste0(header, \":\\n\"), paste0(entries, collapse = \"\\n\"))\n}\n\nfile_comparator <- function(old, new) {\n  if (setequal(old, new)) {\n    return()\n  }\n\n  removed <- setdiff(old, new)\n  added <- setdiff(new, old)\n\n  if (length(removed) > 0L) print_multiple_entries(\"- File removed\", removed)\n  if (length(added) > 0L) print_multiple_entries(\"- File added\", added)\n}\ndir_tracker <- function(f) {\n  force(f)\n  function(...) {\n    old_files <- dir()\n    on.exit(file_comparator(old_files, dir()), add = TRUE)\n\n    f(...)\n  }\n}\n\nfile_creation_tracker <- dir_tracker(file.create)\nfile_deletion_tracker <- dir_tracker(file.remove)\nfile_creation_tracker(c(\"a.txt\", \"b.txt\"))\n#> - File added:\n#> a.txt\n#> b.txt\n#> [1] TRUE TRUE\n\nfile_deletion_tracker(c(\"a.txt\", \"b.txt\"))\n#> - File removed:\n#> a.txt\n#> b.txt\n#> [1] TRUE TRUE\n# helper function to write to a file connection\nwrite_line <- function(filepath, ...) {\n  cat(..., \"\\n\", sep = \"\", file = filepath, append = TRUE)\n}\n\n# function operator\nlogger <- function(f, filepath) {\n  force(f)\n  force(filepath)\n\n  write_line(filepath, \"Function created at: \", as.character(Sys.time()))\n\n  function(...) {\n    write_line(filepath, \"Function called at:  \", as.character(Sys.time()))\n    f(...)\n  }\n}\n\n# check that the function works as expected with a tempfile\nwithr::with_tempfile(\"logfile\", code = {\n  logged_runif <- logger(runif, logfile)\n\n  Sys.sleep(sample.int(10, 1))\n  logged_runif(1)\n\n  Sys.sleep(sample.int(10, 1))\n  logged_runif(2)\n\n  Sys.sleep(sample.int(10, 1))\n  logged_runif(3)\n\n  cat(readLines(logfile), sep = \"\\n\")\n})\n#> Function created at: 2023-12-31 00:43:08.117083\n#> Function called at:  2023-12-31 00:43:13.123988\n#> Function called at:  2023-12-31 00:43:18.129344\n#> Function called at:  2023-12-31 00:43:26.137767\ndelay_by_atleast <- function(f, amount) {\n  force(f)\n  force(amount)\n\n  # the last time the function was run\n  last_time <- NULL\n\n  function(...) {\n    if (!is.null(last_time)) {\n      wait <- (last_time - Sys.time()) + amount\n      if (wait > 0) Sys.sleep(wait)\n    }\n\n    # update the time in the parent frame for the next run when the function finishes\n    on.exit(last_time <<- Sys.time())\n\n    f(...)\n  }\n}"},{"path":"function-operators.html","id":"session-information-9","chapter":"11 Function operators","heading":"11.3 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr         1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  purrr       * 1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"base-types.html","id":"base-types","chapter":"12 Base Types","heading":"12 Base Types","text":"exercises.","code":""},{"path":"s3.html","id":"s3","chapter":"13 S3","heading":"13 S3","text":"Attaching needed libraries:","code":"\nlibrary(sloop, warn.conflicts = FALSE)\nlibrary(dplyr, warn.conflicts = FALSE)\nlibrary(purrr, warn.conflicts = FALSE)"},{"path":"s3.html","id":"basics-exercises-13.2.1","chapter":"13 S3","heading":"13.1 Basics (Exercises 13.2.1)","text":"Q1. Describe difference t.test() t.data.frame(). function called?A1. difference t.test() t.data.frame() following:t.test() generic function perform t-test.t.test() generic function perform t-test.t.data.frame() method generic t() (matrix transform function) dispatched data.frame objects.t.data.frame() method generic t() (matrix transform function) dispatched data.frame objects.can also confirm function types using ftype():Q2. Make list commonly used base R functions contain . name S3 methods.A2. common R functions . S3 methods:.equal().* functions (like .data.frame(), .numeric(), etc.)install.packages().exit()\netc.example,Q3. .data.frame.data.frame() method ? confusing? avoid confusion code?A3. ’s S3 method generic .data.frame().can seen methods supported generic:Given number .s name, quite confusing figure name generic name class.Q4. Describe difference behaviour two calls.A4. difference behaviour specified calls.unclassing, mean generic dispatches .Date method:unclassing, mean generic dispatches .numeric method:Q5. class object following code return? base type built ? attributes use?A5. object based base type closure6, type function.class ecdf, superclasses.Apart class, following attributes:Q6. class object following code return? base type built ? attributes use?A6. object based base type integer.class table.Apart class, following attributes:","code":"\nftype(t.test)\n#> [1] \"S3\"      \"generic\"\nftype(t.data.frame)\n#> [1] \"S3\"     \"method\"\nftype(as.data.frame)\n#> [1] \"S3\"      \"generic\"\nftype(on.exit)\n#> [1] \"primitive\"\nftype(as.data.frame.data.frame)\n#> [1] \"S3\"     \"method\"\ns3_methods_generic(\"as.data.frame\") %>%\n  dplyr::filter(class == \"data.frame\")\n#> # A tibble: 1 × 4\n#>   generic       class      visible source\n#>   <chr>         <chr>      <lgl>   <chr> \n#> 1 as.data.frame data.frame TRUE    base\nset.seed(1014)\nsome_days <- as.Date(\"2017-01-31\") + sample(10, 5)\nmean(some_days)\n#> [1] \"2017-02-06\"\nmean(unclass(some_days))\n#> [1] 17203.4\nsome_days <- as.Date(\"2017-01-31\") + sample(10, 5)\n\nsome_days\n#> [1] \"2017-02-06\" \"2017-02-09\" \"2017-02-05\" \"2017-02-08\"\n#> [5] \"2017-02-07\"\n\ns3_dispatch(mean(some_days))\n#> => mean.Date\n#>  * mean.default\n\nmean(some_days)\n#> [1] \"2017-02-07\"\nunclass(some_days)\n#> [1] 17203 17206 17202 17205 17204\n\nmean(unclass(some_days))\n#> [1] 17204\n\ns3_dispatch(mean(unclass(some_days)))\n#>    mean.double\n#>    mean.numeric\n#> => mean.default\nx <- ecdf(rpois(100, 10))\nx\nx <- ecdf(rpois(100, 10))\nx\n#> Empirical CDF \n#> Call: ecdf(rpois(100, 10))\n#>  x[1:18] =      2,      3,      4,  ...,     18,     19\n\notype(x)\n#> [1] \"S3\"\ntypeof(x)\n#> [1] \"closure\"\ns3_class(x)\n#> [1] \"ecdf\"     \"stepfun\"  \"function\"\nattributes(x)\n#> $class\n#> [1] \"ecdf\"     \"stepfun\"  \"function\"\n#> \n#> $call\n#> ecdf(rpois(100, 10))\nx <- table(rpois(100, 5))\nx\nx <- table(rpois(100, 5))\nx\n#> \n#>  1  2  3  4  5  6  7  8  9 10 \n#>  7  7 18 13 14 14 16  4  4  3\n\notype(x)\n#> [1] \"S3\"\ntypeof(x)\n#> [1] \"integer\"\ns3_class(x)\n#> [1] \"table\"\nattributes(x)\n#> $dim\n#> [1] 10\n#> \n#> $dimnames\n#> $dimnames[[1]]\n#>  [1] \"1\"  \"2\"  \"3\"  \"4\"  \"5\"  \"6\"  \"7\"  \"8\"  \"9\"  \"10\"\n#> \n#> \n#> $class\n#> [1] \"table\""},{"path":"s3.html","id":"classes-exercises-13.3.4","chapter":"13 S3","heading":"13.2 Classes (Exercises 13.3.4)","text":"Q1. Write constructor data.frame objects. base type data frame built ? attributes use? restrictions placed individual elements? names?A1. data frame built top named list atomic vectors attributes row names:restriction imposed individual elements need length. Additionally, names need syntactically valid unique.Let’s try :Q2. Enhance factor() helper better behaviour one values found levels. base::factor() situation?A2. one values found levels, values converted NA base::factor():new constructor, can throw error inform user:Let’s try :Q3. Carefully read source code factor(). constructor ?A3. source code factor() can read .number ways base version flexible.allows labeling values:checks levels duplicated.levels argument can NULL.Q4. Factors optional “contrasts” attribute. Read help C(), briefly describe purpose attribute. type ? Rewrite new_factor() constructor include attribute.A4. Categorical variables typically encoded dummy variables regression models default level compared first factor level. Contrats provide flexible way comparisons.can set \"contrasts\" attribute factor using stats::C().Alternatively, can set \"contrasts\" attribute using matrix (?contrasts):[Contrasts] can matrix one row level factor suitable function like contr.poly character string giving name functionThe constructor provided book:can updated also support contrasts:Q5. Read documentation utils::.roman(). write constructor class? need validator? might helper ?A5. utils::.roman() converts Indo-Arabic numerals Roman numerals. Removing class also reveals implemented using base type integer:Therefore, can create simple constructor create new instance class:docs mention following:numbers 1 3899 unique representation roman numbers, hence others result .roman(NA).Therefore, can warn user return NA validator function:helper function can coerce entered input integer type convenience:Let’s try :","code":"\nunclass(data.frame())\n#> named list()\n#> attr(,\"row.names\")\n#> integer(0)\nnew_data_frame <- function(x = list(), row.names = character()) {\n  # row names should be character\n  if (!all(is.character(row.names))) {\n    stop(\"Row name should be of `chracter` type.\", call. = FALSE)\n  }\n\n  # all elements should have the same length\n  unique_element_lengths <- unique(purrr::map_int(x, length))\n  if (length(unique_element_lengths) > 1L) {\n    stop(\"All list elements in `x` should have same length.\", call. = FALSE)\n  }\n\n  # if not provided, generate row names\n  # this is necessary if there is at least one element in the list\n  if (length(x) > 0L && length(row.names) == 0L) {\n    row.names <- .set_row_names(unique_element_lengths)\n  }\n\n  structure(x, class = \"data.frame\", row.names = row.names)\n}\nnew_data_frame(list(\"x\" = 1, \"y\" = c(2, 3)))\n#> Error: All list elements in `x` should have same length.\n\nnew_data_frame(list(\"x\" = 1, \"y\" = c(2)), row.names = 1L)\n#> Error: Row name should be of `chracter` type.\n\nnew_data_frame(list())\n#> data frame with 0 columns and 0 rows\n\nnew_data_frame(list(\"x\" = 1, \"y\" = 2))\n#>   x y\n#> 1 1 2\n\nnew_data_frame(list(\"x\" = 1, \"y\" = 2), row.names = \"row-1\")\n#>       x y\n#> row-1 1 2\nbase::factor(c(\"a\", \"b\", \"c\"), levels = c(\"a\", \"c\"))\n#> [1] a    <NA> c   \n#> Levels: a c\nnew_factor <- function(x = integer(), levels = character()) {\n  stopifnot(is.integer(x))\n  stopifnot(is.character(levels))\n\n  structure(\n    x,\n    levels = levels,\n    class = \"factor\"\n  )\n}\n\nvalidate_factor <- function(x) {\n  values <- unclass(x)\n  levels <- attr(x, \"levels\")\n\n  if (!all(!is.na(values) & values > 0)) {\n    stop(\n      \"All `x` values must be non-missing and greater than zero\",\n      call. = FALSE\n    )\n  }\n\n  if (length(levels) < max(values)) {\n    stop(\n      \"There must be at least as many `levels` as possible values in `x`\",\n      call. = FALSE\n    )\n  }\n\n  x\n}\n\ncreate_factor <- function(x = character(), levels = unique(x)) {\n  ind <- match(x, levels)\n\n  if (any(is.na(ind))) {\n    missing_values <- x[which(is.na(match(x, levels)))]\n\n    stop(\n      paste0(\n        \"Following values from `x` are not present in `levels`:\\n\",\n        paste0(missing_values, collapse = \"\\n\")\n      ),\n      call. = FALSE\n    )\n  }\n\n  validate_factor(new_factor(ind, levels))\n}\ncreate_factor(c(\"a\", \"b\", \"c\"), levels = c(\"a\", \"c\"))\n#> Error: Following values from `x` are not present in `levels`:\n#> b\n\ncreate_factor(c(\"a\", \"b\", \"c\"), levels = c(\"a\", \"b\", \"c\"))\n#> [1] a b c\n#> Levels: a b c\nx <- c(\"a\", \"b\", \"b\")\nlevels <- c(\"a\", \"b\", \"c\")\nlabels <- c(\"one\", \"two\", \"three\")\n\nfactor(x, levels = levels, labels = labels)\n#> [1] one two two\n#> Levels: one two three\nx <- c(\"a\", \"b\", \"b\")\nlevels <- c(\"a\", \"b\", \"b\")\n\nfactor(x, levels = levels)\n#> Error in `levels<-`(`*tmp*`, value = as.character(levels)): factor level [3] is duplicated\n\ncreate_factor(x, levels = levels)\n#> [1] a b b\n#> Levels: a b b\n#> Warning in print.factor(x): duplicated level [3] in factor\nx <- c(\"a\", \"b\", \"b\")\n\nfactor(x, levels = NULL)\n#> [1] <NA> <NA> <NA>\n#> Levels:\n\ncreate_factor(x, levels = NULL)\n#> Error: Following values from `x` are not present in `levels`:\n#> a\n#> b\n#> b\nnew_factor <- function(x = integer(), levels = character()) {\n  stopifnot(is.integer(x))\n  stopifnot(is.character(levels))\n\n  structure(\n    x,\n    levels = levels,\n    class = \"factor\"\n  )\n}\nnew_factor <- function(x = integer(),\n                       levels = character(),\n                       contrasts = NULL) {\n  stopifnot(is.integer(x))\n  stopifnot(is.character(levels))\n\n  if (!is.null(contrasts)) {\n    stopifnot(is.matrix(contrasts) && is.numeric(contrasts))\n  }\n\n  structure(\n    x,\n    levels = levels,\n    class = \"factor\",\n    contrasts = contrasts\n  )\n}\nas.roman(1)\n#> [1] I\n\ntypeof(unclass(as.roman(1)))\n#> [1] \"integer\"\nnew_roman <- function(x = integer()) {\n  stopifnot(is.integer(x))\n\n  structure(x, class = \"roman\")\n}\nas.roman(10000)\n#> [1] <NA>\nvalidate_new_roman <- function(x) {\n  int_values <- unclass(x)\n\n  if (any(int_values < 1L | int_values > 3899L)) {\n    warning(\n      \"Integer should be between 1 and 3899. Returning `NA` otherwise.\",\n      call. = FALSE\n    )\n  }\n\n  x\n}\nroman <- function(x = integer()) {\n  x <- as.integer(x)\n\n  validate_new_roman(new_roman(x))\n}\nroman(1)\n#> [1] I\n\nroman(c(5, 20, 100, 150, 100000))\n#> Warning: Integer should be between 1 and 3899. Returning\n#> `NA` otherwise.\n#> [1] V    XX   C    CL   <NA>"},{"path":"s3.html","id":"generics-and-methods-exercises-13.4.4","chapter":"13 S3","heading":"13.3 Generics and methods (Exercises 13.4.4)","text":"Q1. Read source code t() t.test() confirm t.test() S3 generic S3 method. happens create object class test call t() ? ?A1. Looking source code functions, can see generic, can confirm using sloop:Looking S3 dispatch, can see since R can’t find S3 method test class generic function t(), dispatches default method, converts structure matrix:behaviour can observed vector:Q2. generics table class methods ?A2. table class methods following generics:Q3. generics ecdf class methods ?A3. ecdf class methods following generics:Q4. base generic greatest number defined methods?A4. answer question, first, let’s list functions base retain generics.Now ’s simple matter counting number methods per generic ordering data frame descending order count:reveals base generic function methods print().Q5. Carefully read documentation UseMethod() explain following code returns results . two usual rules function evaluation UseMethod() violate?A5. called directly, g.default() method takes x value argument y global environment:, g() function called, takes x argument, comes function environment:docs ?UseMethod() clarify case:local variables defined call UseMethod retainedThat , UseMethod() calls g.default(), variables defined inside generic also available g.default() method. arguments supplied function passed , however, affected code inside generic.Two rules function evaluation violated UseMethod():Name maskingA fresh startQ6. arguments [? hard question answer?A6. difficult say many formal arguments subsetting [ operator generic function methods vectors, matrices, arrays, lists, etc., different methods different number arguments:can sample see wide variation number formal arguments:","code":"\nx <- structure(1:10, class = \"test\")\nt(x)\nt\n#> function (x) \n#> UseMethod(\"t\")\n#> <bytecode: 0x56246846b798>\n#> <environment: namespace:base>\nsloop::is_s3_generic(\"t\")\n#> [1] TRUE\n\nt.test\n#> function (x, ...) \n#> UseMethod(\"t.test\")\n#> <bytecode: 0x562463507140>\n#> <environment: namespace:stats>\nsloop::is_s3_generic(\"t.test\")\n#> [1] TRUE\nx <- structure(1:10, class = \"test\")\nt(x)\n#>      [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10]\n#> [1,]    1    2    3    4    5    6    7    8    9    10\n#> attr(,\"class\")\n#> [1] \"test\"\ns3_dispatch(t(x))\n#> => t.test\n#>  * t.default\nt(1:10)\n#>      [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10]\n#> [1,]    1    2    3    4    5    6    7    8    9    10\ns3_methods_class(\"table\")\n#> # A tibble: 11 × 4\n#>    generic       class visible source             \n#>    <chr>         <chr> <lgl>   <chr>              \n#>  1 [             table TRUE    base               \n#>  2 aperm         table TRUE    base               \n#>  3 as_tibble     table FALSE   registered S3method\n#>  4 as.data.frame table TRUE    base               \n#>  5 Axis          table FALSE   registered S3method\n#>  6 lines         table FALSE   registered S3method\n#>  7 plot          table FALSE   registered S3method\n#>  8 points        table FALSE   registered S3method\n#>  9 print         table TRUE    base               \n#> 10 summary       table TRUE    base               \n#> 11 tail          table FALSE   registered S3method\ns3_methods_class(\"ecdf\")\n#> # A tibble: 4 × 4\n#>   generic  class visible source             \n#>   <chr>    <chr> <lgl>   <chr>              \n#> 1 plot     ecdf  TRUE    stats              \n#> 2 print    ecdf  FALSE   registered S3method\n#> 3 quantile ecdf  FALSE   registered S3method\n#> 4 summary  ecdf  FALSE   registered S3method\n# getting all functions names\nobjs <- mget(ls(\"package:base\", all = TRUE), inherits = TRUE)\nfuns <- Filter(is.function, objs)\n\n# extracting only generics\ngenFuns <- names(funs) %>%\n  purrr::keep(~ sloop::is_s3_generic(.x))\npurrr::map_dfr(\n  genFuns,\n  ~ s3_methods_generic(.)\n) %>%\n  dplyr::group_by(generic) %>%\n  dplyr::tally() %>%\n  dplyr::arrange(desc(n))\n#> # A tibble: 122 × 2\n#>    generic           n\n#>    <chr>         <int>\n#>  1 print           289\n#>  2 format          130\n#>  3 [                53\n#>  4 summary          39\n#>  5 as.character     38\n#>  6 as.data.frame    32\n#>  7 plot             30\n#>  8 [[               26\n#>  9 [<-              17\n#> 10 $                15\n#> # ℹ 112 more rows\ng <- function(x) {\n  x <- 10\n  y <- 10\n  UseMethod(\"g\")\n}\ng.default <- function(x) c(x = x, y = y)\nx <- 1\ny <- 1\ng(x)\n#>  x  y \n#>  1 10\ng.default(x)\n#> x y \n#> 1 1\ng(x)\n#>  x  y \n#>  1 10\ns3_methods_generic(\"[\") %>%\n  dplyr::filter(source == \"base\")\n#> # A tibble: 17 × 4\n#>    generic class           visible source\n#>    <chr>   <chr>           <lgl>   <chr> \n#>  1 [       AsIs            TRUE    base  \n#>  2 [       data.frame      TRUE    base  \n#>  3 [       Date            TRUE    base  \n#>  4 [       difftime        TRUE    base  \n#>  5 [       Dlist           TRUE    base  \n#>  6 [       DLLInfoList     TRUE    base  \n#>  7 [       factor          TRUE    base  \n#>  8 [       hexmode         TRUE    base  \n#>  9 [       listof          TRUE    base  \n#> 10 [       noquote         TRUE    base  \n#> 11 [       numeric_version TRUE    base  \n#> 12 [       octmode         TRUE    base  \n#> 13 [       POSIXct         TRUE    base  \n#> 14 [       POSIXlt         TRUE    base  \n#> 15 [       simple.list     TRUE    base  \n#> 16 [       table           TRUE    base  \n#> 17 [       warnings        TRUE    base\n# table\nnames(formals(`[.table`))\n#> [1] \"x\"    \"i\"    \"j\"    \"...\"  \"drop\"\n\n# Date\nnames(formals(`[.Date`))\n#> [1] \"x\"    \"...\"  \"drop\"\n\n# data frame\nnames(formals(`[.data.frame`))\n#> [1] \"x\"    \"i\"    \"j\"    \"drop\"\n\n# etc."},{"path":"s3.html","id":"object-styles-exercises-13.5.1","chapter":"13 S3","heading":"13.4 Object styles (Exercises 13.5.1)","text":"Q1. Categorise objects returned lm(), factor(), table(), .Date(), .POSIXct() ecdf(), ordered(), () styles described .A1. Objects returned functions can categorized follows:Vector style objects (length represents . observations)factor()table().Date().POSIXct()ordered()Record style objects (equi-length vectors represent object components)None.Dataframe style objects (Record style two-dimensions)None.Scalar objects (list represent single thing)lm() (represent one regression model)ecdf() (represents one distribution)() special:\njust adds new class object indicate treated .Therefore, object style superclass’ object style.Q2. constructor function lm objects, new_lm(), look like? Use ?lm experimentation figure required fields types.A2. lm object scalar object, .e. object contains named list atomic vectors varying lengths types represent single thing (regression model).Based information, can write new constructor object:","code":"\nfactor_obj <- factor(c(\"a\", \"b\"))\nlength(factor_obj)\n#> [1] 2\nlength(unclass(factor_obj))\n#> [1] 2\ntab_object <- table(mtcars$am)\nlength(tab_object)\n#> [1] 2\nlength(unlist(tab_object))\n#> [1] 2\ndate_obj <- as.Date(\"02/27/92\", \"%m/%d/%y\")\nlength(date_obj)\n#> [1] 1\nlength(unclass(date_obj))\n#> [1] 1\nposix_obj <- as.POSIXct(1472562988, origin = \"1960-01-01\")\nlength(posix_obj)\n#> [1] 1\nlength(unclass(posix_obj))\n#> [1] 1\nordered_obj <- ordered(factor(c(\"a\", \"b\")))\nlength(ordered_obj)\n#> [1] 2\nlength(unclass(ordered_obj))\n#> [1] 2\nlm_obj <- lm(wt ~ mpg, mtcars)\nlength(lm_obj)\n#> [1] 12\nlength(unclass(lm_obj))\n#> [1] 12\necdf_obj <- ecdf(rnorm(12))\nlength(ecdf_obj)\n#> [1] 1\nlength(unclass(ecdf_obj))\n#> [1] 1\nx <- ecdf(rnorm(12))\nclass(x)\n#> [1] \"ecdf\"     \"stepfun\"  \"function\"\nclass(I(x))\n#> [1] \"AsIs\"     \"ecdf\"     \"stepfun\"  \"function\"\nmod <- lm(wt ~ mpg, mtcars)\n\ntypeof(mod)\n#> [1] \"list\"\n\nattributes(mod)\n#> $names\n#>  [1] \"coefficients\"  \"residuals\"     \"effects\"      \n#>  [4] \"rank\"          \"fitted.values\" \"assign\"       \n#>  [7] \"qr\"            \"df.residual\"   \"xlevels\"      \n#> [10] \"call\"          \"terms\"         \"model\"        \n#> \n#> $class\n#> [1] \"lm\"\n\npurrr::map_chr(unclass(mod), typeof)\n#>  coefficients     residuals       effects          rank \n#>      \"double\"      \"double\"      \"double\"     \"integer\" \n#> fitted.values        assign            qr   df.residual \n#>      \"double\"     \"integer\"        \"list\"     \"integer\" \n#>       xlevels          call         terms         model \n#>        \"list\"    \"language\"    \"language\"        \"list\"\n\npurrr::map_int(unclass(mod), length)\n#>  coefficients     residuals       effects          rank \n#>             2            32            32             1 \n#> fitted.values        assign            qr   df.residual \n#>            32             2             5             1 \n#>       xlevels          call         terms         model \n#>             0             3             3             2\nnew_lm <- function(coefficients,\n                   residuals,\n                   effects,\n                   rank,\n                   fitted.values,\n                   assign,\n                   qr,\n                   df.residual,\n                   xlevels,\n                   call,\n                   terms,\n                   model) {\n  stopifnot(\n    is.double(coefficients),\n    is.double(residuals),\n    is.double(effects),\n    is.integer(rank),\n    is.double(fitted.values),\n    is.integer(assign),\n    is.list(qr),\n    is.integer(df.residual),\n    is.list(xlevels),\n    is.language(call),\n    is.language(terms),\n    is.list(model)\n  )\n\n  structure(\n    list(\n      coefficients  = coefficients,\n      residuals     = residuals,\n      effects       = effects,\n      rank          = rank,\n      fitted.values = fitted.values,\n      assign        = assign,\n      qr            = qr,\n      df.residual   = df.residual,\n      xlevels       = xlevels,\n      call          = call,\n      terms         = terms,\n      model         = model\n    ),\n    class = \"lm\"\n  )\n}"},{"path":"s3.html","id":"inheritance-exercises-13.6.3","chapter":"13 S3","heading":"13.5 Inheritance (Exercises 13.6.3)","text":"Q1. [.Date support subclasses? fail support subclasses?A1. [.Date method defined follows:.Date function looks like :, oldClass class().Therefore, reading code, can surmise :[.Date supports subclasses preserving class input.[.Date fails support subclasses preserving attributes input.example,Q2. R two classes representing date time data, POSIXct POSIXlt, inherit POSIXt. generics different behaviours two classes? generics share behaviour?A2. First, let’s demonstrate POSIXct POSIXlt indeed subclasses POSIXt superclass.Remember way S3 method dispatch works, generic method superclass, method also inherited subclass.can extract vector generics supported sub- super-classes:Methods specific subclasses:Let’s see example:Methods inherited subclasses superclass:Let’s see one example generic:Q3. expect code return? actually return? ?A3. Naively, expect code return \"a1\", actually returns \"a2\":S3 dispatch explains :mentioned book, UseMethod() functiontracks list potential next methods special variable, means modifying object ’s dispatched upon impact method gets called next.special variable .Class:.Class character vector classes used find next method. NextMethod adds attribute “previous” .Class giving .Class last used dispatch, shifts .Class along used dispatch., can print .Class confirm adding new class x indeed doesn’t change .Class, therefore dispatch occurs \"a2\" class:","code":"\nsloop::s3_get_method(\"[.Date\")\n#> function (x, ..., drop = TRUE) \n#> {\n#>     .Date(NextMethod(\"[\"), oldClass(x))\n#> }\n#> <bytecode: 0x5624687fc3c0>\n#> <environment: namespace:base>\n.Date\n#> function (xx, cl = \"Date\") \n#> `class<-`(xx, cl)\n#> <bytecode: 0x562466349c10>\n#> <environment: namespace:base>\nx <- structure(Sys.Date(), name = \"myName\", class = c(\"subDate\", \"Date\"))\n\n# `$name` is gone\nattributes(x[1])\n#> $class\n#> [1] \"subDate\" \"Date\"\n\nx[1]\n#> [1] \"2023-12-31\"\ndt_lt <- as.POSIXlt(Sys.time(), \"GMT\")\nclass(dt_lt)\n#> [1] \"POSIXlt\" \"POSIXt\"\n\ndt_ct <- as.POSIXct(Sys.time(), \"GMT\")\nclass(dt_ct)\n#> [1] \"POSIXct\" \"POSIXt\"\n\ndt_t <- structure(dt_ct, class = \"POSIXt\")\nclass(dt_t)\n#> [1] \"POSIXt\"\n(t_generics <- s3_methods_class(\"POSIXt\")$generic)\n#>  [1] \"-\"            \"+\"            \"all.equal\"   \n#>  [4] \"as.character\" \"Axis\"         \"cut\"         \n#>  [7] \"diff\"         \"hist\"         \"is.numeric\"  \n#> [10] \"julian\"       \"Math\"         \"months\"      \n#> [13] \"Ops\"          \"pretty\"       \"quantile\"    \n#> [16] \"quarters\"     \"round\"        \"seq\"         \n#> [19] \"str\"          \"trunc\"        \"weekdays\"\n\n(lt_generics <- s3_methods_class(\"POSIXlt\")$generic)\n#>  [1] \"[\"             \"[[\"            \"[[<-\"         \n#>  [4] \"[<-\"           \"$<-\"           \"anyNA\"        \n#>  [7] \"as.data.frame\" \"as.Date\"       \"as.double\"    \n#> [10] \"as.list\"       \"as.matrix\"     \"as.POSIXct\"   \n#> [13] \"as.vector\"     \"c\"             \"duplicated\"   \n#> [16] \"format\"        \"is.finite\"     \"is.infinite\"  \n#> [19] \"is.na\"         \"is.nan\"        \"length\"       \n#> [22] \"length<-\"      \"mean\"          \"names\"        \n#> [25] \"names<-\"       \"print\"         \"rep\"          \n#> [28] \"sort\"          \"summary\"       \"Summary\"      \n#> [31] \"unique\"        \"weighted.mean\" \"xtfrm\"\n\n(ct_generics <- s3_methods_class(\"POSIXct\")$generic)\n#>  [1] \"[\"             \"[[\"            \"[<-\"          \n#>  [4] \"as.data.frame\" \"as.Date\"       \"as.list\"      \n#>  [7] \"as.POSIXlt\"    \"c\"             \"format\"       \n#> [10] \"length<-\"      \"mean\"          \"print\"        \n#> [13] \"rep\"           \"split\"         \"summary\"      \n#> [16] \"Summary\"       \"weighted.mean\" \"xtfrm\"\nunion(lt_generics, ct_generics)\n#>  [1] \"[\"             \"[[\"            \"[[<-\"         \n#>  [4] \"[<-\"           \"$<-\"           \"anyNA\"        \n#>  [7] \"as.data.frame\" \"as.Date\"       \"as.double\"    \n#> [10] \"as.list\"       \"as.matrix\"     \"as.POSIXct\"   \n#> [13] \"as.vector\"     \"c\"             \"duplicated\"   \n#> [16] \"format\"        \"is.finite\"     \"is.infinite\"  \n#> [19] \"is.na\"         \"is.nan\"        \"length\"       \n#> [22] \"length<-\"      \"mean\"          \"names\"        \n#> [25] \"names<-\"       \"print\"         \"rep\"          \n#> [28] \"sort\"          \"summary\"       \"Summary\"      \n#> [31] \"unique\"        \"weighted.mean\" \"xtfrm\"        \n#> [34] \"as.POSIXlt\"    \"split\"\ns3_dispatch(is.na(dt_lt))\n#> => is.na.POSIXlt\n#>    is.na.POSIXt\n#>    is.na.default\n#>  * is.na (internal)\n\ns3_dispatch(is.na(dt_ct))\n#>    is.na.POSIXct\n#>    is.na.POSIXt\n#>    is.na.default\n#> => is.na (internal)\n\ns3_dispatch(is.na(dt_t))\n#>    is.na.POSIXt\n#>    is.na.default\n#> => is.na (internal)\nsetdiff(t_generics, union(lt_generics, ct_generics))\n#>  [1] \"-\"            \"+\"            \"all.equal\"   \n#>  [4] \"as.character\" \"Axis\"         \"cut\"         \n#>  [7] \"diff\"         \"hist\"         \"is.numeric\"  \n#> [10] \"julian\"       \"Math\"         \"months\"      \n#> [13] \"Ops\"          \"pretty\"       \"quantile\"    \n#> [16] \"quarters\"     \"round\"        \"seq\"         \n#> [19] \"str\"          \"trunc\"        \"weekdays\"\ns3_dispatch(is.numeric(dt_lt))\n#>    is.numeric.POSIXlt\n#> => is.numeric.POSIXt\n#>    is.numeric.default\n#>  * is.numeric (internal)\n\ns3_dispatch(is.numeric(dt_ct))\n#>    is.numeric.POSIXct\n#> => is.numeric.POSIXt\n#>    is.numeric.default\n#>  * is.numeric (internal)\n\ns3_dispatch(is.numeric(dt_t))\n#> => is.numeric.POSIXt\n#>    is.numeric.default\n#>  * is.numeric (internal)\ngeneric2 <- function(x) UseMethod(\"generic2\")\ngeneric2.a1 <- function(x) \"a1\"\ngeneric2.a2 <- function(x) \"a2\"\ngeneric2.b <- function(x) {\n  class(x) <- \"a1\"\n  NextMethod()\n}\n\ngeneric2(structure(list(), class = c(\"b\", \"a2\")))\ngeneric2 <- function(x) UseMethod(\"generic2\")\ngeneric2.a1 <- function(x) \"a1\"\ngeneric2.a2 <- function(x) \"a2\"\ngeneric2.b <- function(x) {\n  class(x) <- \"a1\"\n  NextMethod()\n}\n\ngeneric2(structure(list(), class = c(\"b\", \"a2\")))\n#> [1] \"a2\"\nsloop::s3_dispatch(generic2(structure(list(), class = c(\"b\", \"a2\"))))\n#> => generic2.b\n#> -> generic2.a2\n#>    generic2.default\ngeneric2.b <- function(x) {\n  message(paste0(\"before: \", paste0(.Class, collapse = \", \")))\n  class(x) <- \"a1\"\n  message(paste0(\"after: \", paste0(.Class, collapse = \", \")))\n\n  NextMethod()\n}\n\ninvisible(generic2(structure(list(), class = c(\"b\", \"a2\"))))\n#> before: b, a2\n#> after: b, a2"},{"path":"s3.html","id":"dispatch-details-exercises-13.7.5","chapter":"13 S3","heading":"13.6 Dispatch details (Exercises 13.7.5)","text":"Q1. Explain differences dispatch :A1. differences dispatch due classes arguments:x1 implicit class integer inherits numeric, x2 explicitly assigned class integer.Q2. classes method Math group generic base R? Read source code. methods work?A2. following classes method Math group generic base R:Reading source code methods:Math.factor() Math.Date() provide error message:Math.data.frame() defined follows (except first line code, deliberately added):can surmised code: method checks elements expected type., applies generic (tracked via environment variable .Generic) element list atomic vectors makes data frame:, produces error:Q3. Math.difftime() complicated described. ?A3. Math.difftime() source code looks like following:group generic bit complicated produces error generics, works others.","code":"\nlength.integer <- function(x) 10\n\nx1 <- 1:5\nclass(x1)\n#> [1] \"integer\"\ns3_dispatch(length(x1))\n#>  * length.integer\n#>    length.numeric\n#>    length.default\n#> => length (internal)\n\nx2 <- structure(x1, class = \"integer\")\nclass(x2)\n#> [1] \"integer\"\ns3_dispatch(length(x2))\n#> => length.integer\n#>    length.default\n#>  * length (internal)\ns3_class(x1)\n#> [1] \"integer\" \"numeric\"\n\ns3_class(x2)\n#> [1] \"integer\"\ns3_methods_generic(\"Math\") %>%\n  dplyr::filter(source == \"base\")\n#> # A tibble: 5 × 4\n#>   generic class      visible source\n#>   <chr>   <chr>      <lgl>   <chr> \n#> 1 Math    data.frame TRUE    base  \n#> 2 Math    Date       TRUE    base  \n#> 3 Math    difftime   TRUE    base  \n#> 4 Math    factor     TRUE    base  \n#> 5 Math    POSIXt     TRUE    base\nMath.factor <- function(x, ...) {\n  stop(gettextf(\"%s not meaningful for factors\", sQuote(.Generic)))\n}\n\nMath.Date <- function(x, ...) {\n  stop(gettextf(\"%s not defined for \\\"Date\\\" objects\", .Generic),\n    domain = NA\n  )\n}\nMath.data.frame <- function(x, ...) {\n  message(paste0(\"Environment variable `.Generic` set to: \", .Generic))\n\n  mode.ok <- vapply(x, function(x) {\n    is.numeric(x) || is.logical(x) || is.complex(x)\n  }, NA)\n\n  if (all(mode.ok)) {\n    x[] <- lapply(X = x, FUN = .Generic, ...)\n    return(x)\n  } else {\n    vnames <- names(x)\n    if (is.null(vnames)) vnames <- seq_along(x)\n    stop(\n      \"non-numeric-alike variable(s) in data frame: \",\n      paste(vnames[!mode.ok], collapse = \", \")\n    )\n  }\n}\ndf1 <- data.frame(x = 1:2, y = 3:4)\nsqrt(df1)\n#> Environment variable `.Generic` set to: sqrt\n#>          x        y\n#> 1 1.000000 1.732051\n#> 2 1.414214 2.000000\ndf2 <- data.frame(x = c(TRUE, FALSE), y = c(\"a\", \"b\"))\nabs(df2)\n#> Environment variable `.Generic` set to: abs\n#> Error in Math.data.frame(df2): non-numeric-alike variable(s) in data frame: y\nMath.difftime <- function(x, ...) {\n  switch(.Generic,\n    \"abs\" = ,\n    \"sign\" = ,\n    \"floor\" = ,\n    \"ceiling\" = ,\n    \"trunc\" = ,\n    \"round\" = ,\n    \"signif\" = {\n      units <- attr(x, \"units\")\n      .difftime(NextMethod(), units)\n    },\n    ### otherwise :\n    stop(gettextf(\"'%s' not defined for \\\"difftime\\\" objects\", .Generic),\n      domain = NA\n    )\n  )\n}"},{"path":"s3.html","id":"session-information-10","chapter":"13 S3","heading":"13.7 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  codetools     0.2-19  2023-02-01 [3] CRAN (R 4.3.2)\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  crayon        1.5.2   2022-09-29 [1] RSPM\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr       * 1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  purrr       * 1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  sloop       * 1.0.1   2019-02-17 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"r6.html","id":"r6","chapter":"14 R6","heading":"14 R6","text":"Loading needed libraries:","code":"\nlibrary(R6)"},{"path":"r6.html","id":"classes-and-methods-exercises-14.2.6","chapter":"14 R6","heading":"14.1 Classes and methods (Exercises 14.2.6)","text":"Q1. Create bank account R6 class stores balance allows deposit withdraw money. Create subclass throws error attempt go overdraft. Create another subclass allows go overdraft, charges fee. Create superclass make sure works expected.A1. First, let’s create bank account R6 class stores balance allows deposit withdraw money:Let’s try :Create subclass errors attempt overdraw:Let’s try :Now let’s create subclass charges fee account overdrawn:Let’s try :Q2. Create R6 class represents shuffled deck cards. able draw cards deck $draw(n), return cards deck reshuffle $reshuffle(). Use following code make vector cards.A2. Let’s create needed class represents shuffled deck cards:Let’s try :Q3. can’t model bank account deck cards S3 class?A3. can’t model bank account deck cards S3 class instances classes immutable.hand, R6 classes encapsulate data represent state, can change course object’s lifecycle. words, objects mutable well-suited model bank account.Q4. Create R6 class allows get set current time zone. can access current time zone Sys.timezone() set Sys.setenv(TZ = \"newtimezone\"). setting time zone, make sure new time zone list provided OlsonNames().A4. R6 class manages current time zone:Let’s try :Q5. Create R6 class manages current working directory. $get() $set() methods.A5. R6 class manages current working directory:Let’s create instance class check methods work expected:Q6. can’t model time zone current working directory S3 class?A6. answer Q3:Objects represent real-life entities need mutable S3 class instances mutable.Q7. base type R6 objects built top ? attributes ?A7. Let’s create example class create instance class:R6 objects built top environment:class attribute, character vector \"R6\" last element superclasses elements:","code":"\nlibrary(R6)\n\nbankAccount <- R6::R6Class(\n  \"bankAccount\",\n  public = list(\n    balance = 0,\n    initialize = function(balance) {\n      self$balance <- balance\n    },\n    deposit = function(amount) {\n      self$balance <- self$balance + amount\n      message(paste0(\"Current balance is: \", self$balance))\n      invisible(self)\n    },\n    withdraw = function(amount) {\n      self$balance <- self$balance - amount\n      message(paste0(\"Current balance is: \", self$balance))\n      invisible(self)\n    }\n  )\n)\nindra <- bankAccount$new(balance = 100)\n\nindra$deposit(20)\n#> Current balance is: 120\n\nindra$withdraw(10)\n#> Current balance is: 110\nbankAccountStrict <- R6::R6Class(\n  \"bankAccountStrict\",\n  inherit = bankAccount,\n  public = list(\n    withdraw = function(amount) {\n      if (self$balance - amount < 0) {\n        stop(\n          paste0(\"Can't withdraw more than your current balance: \", self$balance),\n          call. = FALSE\n        )\n      }\n\n      super$withdraw(amount)\n    }\n  )\n)\nPritesh <- bankAccountStrict$new(balance = 100)\n\nPritesh$deposit(20)\n#> Current balance is: 120\n\nPritesh$withdraw(150)\n#> Error: Can't withdraw more than your current balance: 120\nbankAccountFee <- R6::R6Class(\n  \"bankAccountFee\",\n  inherit = bankAccount,\n  public = list(\n    withdraw = function(amount) {\n      super$withdraw(amount)\n\n      if (self$balance) {\n        self$balance <- self$balance - 10\n        message(\"You're withdrawing more than your current balance. You will be charged a fee of 10 euros.\")\n      }\n    }\n  )\n)\nMangesh <- bankAccountFee$new(balance = 100)\n\nMangesh$deposit(20)\n#> Current balance is: 120\n\nMangesh$withdraw(150)\n#> Current balance is: -30\n#> You're withdrawing more than your current balance. You will be charged a fee of 10 euros.\nsuit <- c(\"♠\", \"♥\", \"♦\", \"♣\")\nvalue <- c(\"A\", 2:10, \"J\", \"Q\", \"K\")\ncards <- paste0(rep(value, 4), suit)\nsuit <- c(\"♠\", \"♥\", \"♦\", \"♣\")\nvalue <- c(\"A\", 2:10, \"J\", \"Q\", \"K\")\ncards <- paste(rep(value, 4), suit)\n\nDeck <- R6::R6Class(\n  \"Deck\",\n  public = list(\n    initialize = function(deck) {\n      private$cards <- sample(deck)\n    },\n    draw = function(n) {\n      if (n > length(private$cards)) {\n        stop(\n          paste0(\"Can't draw more than remaining number of cards: \", length(private$cards)),\n          call. = FALSE\n        )\n      }\n\n      drawn_cards <- sample(private$cards, n)\n      private$cards <- private$cards[-which(private$cards %in% drawn_cards)]\n      message(paste0(\"Remaining number of cards: \", length(private$cards)))\n\n      return(drawn_cards)\n    },\n    reshuffle = function() {\n      private$cards <- sample(private$cards)\n      invisible(self)\n    }\n  ),\n  private = list(\n    cards = NULL\n  )\n)\nmyDeck <- Deck$new(cards)\n\nmyDeck$draw(4)\n#> Remaining number of cards: 48\n#> [1] \"2 ♠\"  \"10 ♦\" \"9 ♦\"  \"3 ♦\"\n\nmyDeck$reshuffle()$draw(5)\n#> Remaining number of cards: 43\n#> [1] \"6 ♦\"  \"10 ♥\" \"2 ♥\"  \"A ♥\"  \"8 ♥\"\n\nmyDeck$draw(50)\n#> Error: Can't draw more than remaining number of cards: 43\nCurrentTimeZone <- R6::R6Class(\"CurrentTimeZone\",\n  public = list(\n    setTimeZone = function(tz) {\n      stopifnot(tz %in% OlsonNames())\n      Sys.setenv(TZ = tz)\n    },\n    getTimeZone = function() {\n      Sys.timezone()\n    }\n  )\n)\nmyCurrentTimeZone <- CurrentTimeZone$new()\n\nmyCurrentTimeZone$getTimeZone()\n#> [1] \"UTC\"\n\nmyCurrentTimeZone$setTimeZone(\"Asia/Kolkata\")\nmyCurrentTimeZone$getTimeZone()\n#> [1] \"Asia/Kolkata\"\n\nmyCurrentTimeZone$setTimeZone(\"Europe/Berlin\")\nManageDirectory <- R6::R6Class(\"ManageDirectory\",\n  public = list(\n    setWorkingDirectory = function(dir) {\n      setwd(dir)\n    },\n    getWorkingDirectory = function() {\n      getwd()\n    }\n  )\n)\nmyDirManager <- ManageDirectory$new()\n\n# current working directory\nmyDirManager$getWorkingDirectory()\n\n# change and check if that worked\nmyDirManager$setWorkingDirectory(\"..\")\nmyDirManager$getWorkingDirectory()\n\n# revert this change\nmyDirManager$setWorkingDirectory(\"/Advanced-R-exercises\")\nExample <- R6::R6Class(\"Example\")\nmyExample <- Example$new()\ntypeof(myExample)\n#> [1] \"environment\"\n\nrlang::env_print(myExample)\n#> <environment: 0x55ffdaaac290> [L]\n#> Parent: <environment: empty>\n#> Class: Example, R6\n#> Bindings:\n#> • .__enclos_env__: <env>\n#> • clone: <fn> [L]\nattributes(myExample)\n#> $class\n#> [1] \"Example\" \"R6\""},{"path":"r6.html","id":"controlling-access-exercises-14.3.3","chapter":"14 R6","heading":"14.2 Controlling access (Exercises 14.3.3)","text":"Q1. Create bank account class prevents directly setting account balance, can still withdraw deposit . Throw error attempt go overdraft.A1. bank account class satisfies specified requirements:Let’s check works expected:Q2. Create class write-$password field. $check_password(password) method returns TRUE FALSE, way view complete password.A2. implementation class needed properties:, course, everything possible:Q3. Extend Rando class another active binding allows access previous random value. Ensure active binding way access value.A3. modified version Rando class meet specified requirements:Let’s try :Q4. Can subclasses access private fields/methods parent? Perform experiment find .A4. Unlike common OOP languages (e.g. C++), R6 subclasses (derived classes) also access private methods superclass (base class).instance, following example, Duck class private method $quack(), subclass Mallard can access using super$quack().","code":"\nSafeBankAccount <- R6::R6Class(\n  classname = \"SafeBankAccount\",\n  public = list(\n    deposit = function(deposit_amount) {\n      private$.balance <- private$.balance + deposit_amount\n      print(paste(\"Current balance:\", private$.balance))\n\n      invisible(self)\n    },\n    withdraw = function(withdrawal_amount) {\n      if (withdrawal_amount > private$.balance) {\n        stop(\"You can't withdraw more than your current balance.\", call. = FALSE)\n      }\n\n      private$.balance <- private$.balance - withdrawal_amount\n      print(paste(\"Current balance:\", private$.balance))\n\n      invisible(self)\n    }\n  ),\n  private = list(\n    .balance = 0\n  )\n)\nmySafeBankAccount <- SafeBankAccount$new()\n\nmySafeBankAccount$deposit(100)\n#> [1] \"Current balance: 100\"\n\nmySafeBankAccount$withdraw(50)\n#> [1] \"Current balance: 50\"\n\nmySafeBankAccount$withdraw(100)\n#> Error: You can't withdraw more than your current balance.\nlibrary(R6)\n\ncheckCredentials <- R6Class(\n  \"checkCredentials\",\n  public = list(\n    # setter\n    set_password = function(password) {\n      private$.password <- password\n    },\n\n    # checker\n    check_password = function(password) {\n      if (is.null(private$.password)) {\n        stop(\"No password set to check against.\")\n      }\n\n      identical(password, private$.password)\n    },\n\n    # the default print method prints the private fields as well\n    print = function() {\n      cat(\"Password: XXXX\")\n\n      # for method chaining\n      invisible(self)\n    }\n  ),\n  private = list(\n    .password = NULL\n  )\n)\n\nmyCheck <- checkCredentials$new()\n\nmyCheck$set_password(\"1234\")\nprint(myCheck)\n#> Password: XXXX\n\nmyCheck$check_password(\"abcd\")\n#> [1] FALSE\nmyCheck$check_password(\"1234\")\n#> [1] TRUE\nmyCheck$.__enclos_env__$private$.password\n#> [1] \"1234\"\nRando <- R6::R6Class(\"Rando\",\n  active = list(\n    random = function(value) {\n      if (missing(value)) {\n        newValue <- runif(1)\n        private$.previousRandom <- private$.currentRandom\n        private$.currentRandom <- newValue\n        return(private$.currentRandom)\n      } else {\n        stop(\"Can't set `$random`\", call. = FALSE)\n      }\n    },\n    previousRandom = function(value) {\n      if (missing(value)) {\n        if (is.null(private$.previousRandom)) {\n          message(\"No random value has been generated yet.\")\n        } else {\n          return(private$.previousRandom)\n        }\n      } else {\n        stop(\"Can't set `$previousRandom`\", call. = FALSE)\n      }\n    }\n  ),\n  private = list(\n    .currentRandom = NULL,\n    .previousRandom = NULL\n  )\n)\nmyRando <- Rando$new()\n\n# first time\nmyRando$random\n#> [1] 0.5549124\nmyRando$previousRandom\n#> No random value has been generated yet.\n#> NULL\n\n# second time\nmyRando$random\n#> [1] 0.3482785\nmyRando$previousRandom\n#> [1] 0.5549124\n\n# third time\nmyRando$random\n#> [1] 0.2187275\nmyRando$previousRandom\n#> [1] 0.3482785\nDuck <- R6Class(\"Duck\",\n  private = list(quack = function() print(\"Quack Quack\"))\n)\n\nMallard <- R6Class(\"Mallard\",\n  inherit = Duck,\n  public = list(quack = function() super$quack())\n)\n\nmyMallard <- Mallard$new()\nmyMallard$quack()\n#> [1] \"Quack Quack\""},{"path":"r6.html","id":"reference-semantics-exercises-14.4.4","chapter":"14 R6","heading":"14.3 Reference semantics (Exercises 14.4.4)","text":"Q1. Create class allows write line specified file. open connection file $initialize(), append line using cat() $append_line(), close connection $finalize().A1. class allows write line specified file:Let’s check works expected:","code":"\nfileEditor <- R6Class(\n  \"fileEditor\",\n  public = list(\n    initialize = function(filePath) {\n      private$.connection <- file(filePath, open = \"wt\")\n    },\n    append_line = function(text) {\n      cat(\n        text,\n        file = private$.connection,\n        sep = \"\\n\",\n        append = TRUE\n      )\n    }\n  ),\n  private = list(\n    .connection = NULL,\n    # according to R6 docs, the destructor method should be private\n    finalize = function() {\n      print(\"Closing the file connection!\")\n      close(private$.connection)\n    }\n  )\n)\ngreetMom <- function() {\n  f <- tempfile()\n  myfileEditor <- fileEditor$new(f)\n\n  readLines(f)\n\n  myfileEditor$append_line(\"Hi mom!\")\n  myfileEditor$append_line(\"It's a beautiful day!\")\n\n  readLines(f)\n}\n\ngreetMom()\n#> [1] \"Hi mom!\"               \"It's a beautiful day!\"\n\n# force garbage collection\ngc()\n#> [1] \"Closing the file connection!\"\n#>           used (Mb) gc trigger (Mb) max used (Mb)\n#> Ncells  746818 39.9    1365334   73  1365334 73.0\n#> Vcells 1372826 10.5    8388608   64  2594866 19.8"},{"path":"r6.html","id":"session-information-11","chapter":"14 R6","heading":"14.4 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       Europe/Berlin\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  R6          * 2.5.1   2021-08-19 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"s4.html","id":"s4","chapter":"15 S4","heading":"15 S4","text":"","code":""},{"path":"s4.html","id":"basics-exercises-15.2.1","chapter":"15 S4","heading":"15.1 Basics (Exercises 15.2.1)","text":"Q1. lubridate::period() returns S4 class. slots ? class slot? accessors provide?A1. Let’s first create instance Period class:following slots:Additionally, base type slot (numeric) can seen str() output:lubridate package provides accessors slots:Q2. ways can find help method? Read ?\"?\" summarise details.A2. \"?\" operator allows access documentation three ways. demonstrate different ways access documentation, let’s define new S4 class.Ways access documentation:general documentation generic can found ?topic:expression type?topic look overall documentation methods function f.","code":"\nlibrary(lubridate)\nx <- lubridate::period(c(2, 43, 6), c(\"hour\", \"second\", \"minute\"))\nx\n#> [1] \"2H 6M 43S\"\nslotNames(x)\n#> [1] \".Data\"  \"year\"   \"month\"  \"day\"    \"hour\"   \"minute\"\nstr(x)\n#> Formal class 'Period' [package \"lubridate\"] with 6 slots\n#>   ..@ .Data : num 43\n#>   ..@ year  : num 0\n#>   ..@ month : num 0\n#>   ..@ day   : num 0\n#>   ..@ hour  : num 2\n#>   ..@ minute: num 6\nyear(x)\n#> [1] 0\nmonth(x)\n#> [1] 0\nday(x)\n#> [1] 0\nhour(x)\n#> [1] 2\nminute(x)\n#> [1] 6\nsecond(x)\n#> [1] 43\npow <- function(x, exp) c(x, exp)\nsetGeneric(\"pow\")\n#> [1] \"pow\"\nsetMethod(\"pow\", c(\"numeric\", \"numeric\"), function(x, exp) x^exp)\n?pow\n?pow # produces the function documentation\n\nmethods?pow # looks for the overall methods documentation"},{"path":"s4.html","id":"classes-exercises-15.3.6","chapter":"15 S4","heading":"15.2 Classes (Exercises 15.3.6)","text":"Q1. Extend Person class fields match utils::person(). Think slots need, class slot , ’ll need check validity method.A1. code extends Person class described book match closely utils::person().Let’s make sure validation works expected:Q2. happens define new S4 class doesn’t slots? (Hint: read virtual classes ?setClass.)A2. define new S4 class doesn’t slots, create virtual classes:can’t create instance class:useful? mentioned ?setClass docs:Classes exist actual objects can created, virtual classes.common useful form virtual class class union, virtual class defined call setClassUnion() rather call setClass().virtual classes can still inherited:addition specifying slots, another way create virtual classes:Calls setClass() also create virtual class, either Class argument supplied (slots superclasses) contains= argument includes special class name \"VIRTUAL\".Q3. Imagine going reimplement factors, dates, data frames S4. Sketch setClass() calls use define classes. Think appropriate slots prototype.A3. reimplementation following classes S4 might definitions like following.factorFor simplicity, won’t provide options factor() provides. Note x pseudo-class accept objects type.DateJust like base-R version, integer values.data.frameThe tricky part supporting ... argument data.frame(). , can let users pass (named) list.","code":"\nsetClass(\"Person\",\n  slots = c(\n    age     = \"numeric\",\n    given   = \"character\",\n    family  = \"character\",\n    middle  = \"character\",\n    email   = \"character\",\n    role    = \"character\",\n    comment = \"character\"\n  ),\n  prototype = list(\n    age     = NA_real_,\n    given   = NA_character_,\n    family  = NA_character_,\n    middle  = NA_character_,\n    email   = NA_character_,\n    role    = NA_character_,\n    comment = NA_character_\n  )\n)\n\n# Helper function to create an instance of the `Person` class\nPerson <- function(given,\n                   family,\n                   middle = NA_character_,\n                   age = NA_real_,\n                   email = NA_character_,\n                   role = NA_character_,\n                   comment = NA_character_) {\n  age <- as.double(age)\n\n  new(\"Person\",\n    age     = age,\n    given   = given,\n    family  = family,\n    middle  = middle,\n    email   = email,\n    role    = role,\n    comment = comment\n  )\n}\n\n# Validator to ensure that each slot is of length one and that the specified\n# role is one of the possible roles\nsetValidity(\"Person\", function(object) {\n  invalid_length <- NULL\n  slot_lengths <- c(\n    length(object@age),\n    length(object@given),\n    length(object@middle),\n    length(object@family),\n    length(object@email),\n    length(object@comment)\n  )\n\n  if (any(slot_lengths > 1L)) {\n    invalid_length <- \"\\nFollowing slots must be of length 1:\\n @age, @given, @family, @middle, @email, @comment\"\n  }\n\n  possible_roles <- c(\n    NA_character_, \"aut\", \"com\", \"cph\", \"cre\", \"ctb\", \"ctr\", \"dtc\", \"fnd\", \"rev\", \"ths\", \"trl\"\n  )\n\n  if (any(!object@role %in% possible_roles)) {\n    invalid_length <- paste(\n      invalid_length,\n      \"\\nSlot @role(s) must be one of the following:\\n\",\n      paste(possible_roles, collapse = \", \")\n    )\n  }\n\n  if (!is.null(invalid_length)) {\n    return(invalid_length)\n  } else {\n    return(TRUE)\n  }\n})\n#> Class \"Person\" [in \".GlobalEnv\"]\n#> \n#> Slots:\n#>                                                         \n#> Name:        age     given    family    middle     email\n#> Class:   numeric character character character character\n#>                           \n#> Name:       role   comment\n#> Class: character character\n# length of first argument not 1\nPerson(c(\"Indrajeet\", \"Surendra\"), \"Patil\")\n#> Error in validObject(.Object): invalid class \"Person\" object: \n#> Following slots must be of length 1:\n#>  @age, @given, @family, @middle, @email, @comment\n\n# role not recognized\nPerson(\"Indrajeet\", \"Patil\", role = \"xyz\")\n#> Error in validObject(.Object): invalid class \"Person\" object:  \n#> Slot @role(s) must be one of the following:\n#>  NA, aut, com, cph, cre, ctb, ctr, dtc, fnd, rev, ths, trl\n\n# all okay\nPerson(\"Indrajeet\", \"Patil\", role = c(\"aut\", \"cph\"))\n#> An object of class \"Person\"\n#> Slot \"age\":\n#> [1] NA\n#> \n#> Slot \"given\":\n#> [1] \"Indrajeet\"\n#> \n#> Slot \"family\":\n#> [1] \"Patil\"\n#> \n#> Slot \"middle\":\n#> [1] NA\n#> \n#> Slot \"email\":\n#> [1] NA\n#> \n#> Slot \"role\":\n#> [1] \"aut\" \"cph\"\n#> \n#> Slot \"comment\":\n#> [1] NA\nsetClass(\"Empty\")\n\nisVirtualClass(\"Empty\")\n#> [1] TRUE\nnew(\"Empty\")\n#> Error in new(\"Empty\"): trying to generate an object from a virtual class (\"Empty\")\nsetClass(\"Nothing\", contains = \"Empty\")\nsetClass(\"Factor\",\n  slots = c(\n    x       = \"ANY\",\n    levels  = \"character\",\n    ordered = \"logical\"\n  ),\n  prototype = list(\n    x       = character(),\n    levels  = character(),\n    ordered = FALSE\n  )\n)\n\nnew(\"Factor\", x = letters[1:3], levels = LETTERS[1:3])\n#> An object of class \"Factor\"\n#> Slot \"x\":\n#> [1] \"a\" \"b\" \"c\"\n#> \n#> Slot \"levels\":\n#> [1] \"A\" \"B\" \"C\"\n#> \n#> Slot \"ordered\":\n#> [1] FALSE\n\nnew(\"Factor\", x = 1:3, levels = letters[1:3])\n#> An object of class \"Factor\"\n#> Slot \"x\":\n#> [1] 1 2 3\n#> \n#> Slot \"levels\":\n#> [1] \"a\" \"b\" \"c\"\n#> \n#> Slot \"ordered\":\n#> [1] FALSE\n\nnew(\"Factor\", x = c(TRUE, FALSE, TRUE), levels = c(\"x\", \"y\", \"x\"))\n#> An object of class \"Factor\"\n#> Slot \"x\":\n#> [1]  TRUE FALSE  TRUE\n#> \n#> Slot \"levels\":\n#> [1] \"x\" \"y\" \"x\"\n#> \n#> Slot \"ordered\":\n#> [1] FALSE\nsetClass(\"Date2\",\n  slots = list(\n    data = \"integer\"\n  ),\n  prototype = list(\n    data = integer()\n  )\n)\n\nnew(\"Date2\", data = 1342L)\n#> An object of class \"Date2\"\n#> Slot \"data\":\n#> [1] 1342\nsetClass(\"DataFrame\",\n  slots = c(\n    data      = \"list\",\n    row.names = \"character\"\n  ),\n  prototype = list(\n    data      = list(),\n    row.names = character(0L)\n  )\n)\n\nnew(\"DataFrame\", data = list(x = c(\"a\", \"b\"), y = c(1L, 2L)))\n#> An object of class \"DataFrame\"\n#> Slot \"data\":\n#> $x\n#> [1] \"a\" \"b\"\n#> \n#> $y\n#> [1] 1 2\n#> \n#> \n#> Slot \"row.names\":\n#> character(0)"},{"path":"s4.html","id":"generics-and-methods-exercises-15.4.5","chapter":"15 S4","heading":"15.3 Generics and methods (Exercises 15.4.5)","text":"Q1. Add age() accessors Person class.A1. first define generic method class:Q2. definition generic, necessary repeat name generic twice?A2. Let’s look generic just defined; generic name \"age\" repeated twice.:\"age\" passed argument name provides name genericthe \"age\" passed argument def supplies method dispatchThis reminiscent defined S3 generic, also repeat name twice:Q3. show() method defined Section Show method use (object)[[1]]? (Hint: try printing employee subclass.)A3. wish define show() method specific class, need disregard super-/sub-classes.Always using first element ensures method defined class question:Q4. happens define method different argument names generic?A4. Let’s experiment method defined Q1. study behavior.original method worked expected since argument name generic method matched:case, either get warning get error depending many arguments specified:","code":"\nIndra <- Person(\"Indrajeet\", \"Patil\", role = c(\"aut\", \"cph\"), age = 34)\n\nsetGeneric(\"age\", function(x) standardGeneric(\"age\"))\n#> [1] \"age\"\nsetMethod(\"age\", \"Person\", function(x) x@age)\n\nage(Indra)\n#> [1] 34\nsetGeneric(name = \"age\", def = function(x) standardGeneric(\"age\"))\nage <- function(x) {\n  UseMethod(\"age\")\n}\nAlice <- new(\"Employee\")\n\nis(Alice)\n#> [1] \"Employee\" \"Person\"\n\nis(Alice)[[1]]\n#> [1] \"Employee\"\nsetMethod(\"age\", \"Person\", function(x) x@age)\nsetMethod(\"age\", \"Person\", function(object) object@age)\n#> Warning: For function 'age', signature 'Person': argument\n#> in method definition changed from (object) to (x)\n\nsetMethod(\"age\", \"Person\", function(object, x) object@age)\n#> Error in rematchDefinition(definition, fdef, mnames, fnames, signature): methods can add arguments to the generic 'age' only if '...' is an argument to the generic\n\nsetMethod(\"age\", \"Person\", function(...) ...elt(1)@age)\n#> Warning: For function 'age', signature 'Person': argument\n#> in method definition changed from (...) to (x)\n\nsetMethod(\"age\", \"Person\", function(x, ...) x@age)\n#> Error in rematchDefinition(definition, fdef, mnames, fnames, signature): methods can add arguments to the generic 'age' only if '...' is an argument to the generic"},{"path":"s4.html","id":"method-dispatch-exercises-15.5.5","chapter":"15 S4","heading":"15.4 Method dispatch (Exercises 15.5.5)","text":"Q1. Draw method graph f(😅, 😽).A1. don’t prepare visual illustrations used book, linking illustration official solution manual:Q2. Draw method graph f(😃, 😉, 😙).A2. don’t prepare visual illustrations used book, linking illustration official solution manual:Q3. Take last example shows multiple dispatch two classes use multiple inheritance. happens define method terminal classes? method dispatch save us much work ?A3. one class distance 2 terminal nodes four distance 1 two terminal nodes , introduce ambiguity.Method dispatch save us much work resolve ambiguity define five methods (one per class combination).","code":""},{"path":"s4.html","id":"s4-and-s3-exercises-15.6.3","chapter":"15 S4","heading":"15.5 S4 and S3 (Exercises 15.6.3)","text":"Q1. full setOldClass() definition look like ordered factor (.e. add slots prototype definition )?A1. can register old-style/S3 ordered class formally defined class using setOldClass().Let’s use see works expected.Q2. Define length method Person class.A2. Person class can used create objects represent multiple people, let’s say length() method returns many persons object.can define S3 method class:Alternatively, can also write S4 method:","code":"\nsetClass(\"factor\",\n  contains = \"integer\",\n  slots = c(\n    levels = \"character\"\n  ),\n  prototype = structure(\n    integer(),\n    levels = character()\n  )\n)\nsetOldClass(\"factor\", S4Class = \"factor\")\n#> Warning in rm(list = what, pos = classWhere): object\n#> '.__C__factor' not found\n\nsetClass(\"Ordered\",\n  contains = \"factor\",\n  slots = c(\n    levels  = \"character\",\n    ordered = \"logical\"\n  ),\n  prototype = structure(\n    integer(),\n    levels  = character(),\n    ordered = logical()\n  )\n)\n\nsetOldClass(\"ordered\", S4Class = \"Ordered\")\nx <- new(\"Ordered\", 1L:4L, levels = letters[1:4], ordered = TRUE)\n\nx\n#> Object of class \"Ordered\"\n#> [1] a b c d\n#> Levels: a b c d\n#> Slot \"ordered\":\n#> [1] TRUE\n\nstr(x)\n#> Formal class 'Ordered' [package \".GlobalEnv\"] with 4 slots\n#>   ..@ .Data   : int [1:4] 1 2 3 4\n#>   ..@ levels  : chr [1:4] \"a\" \"b\" \"c\" \"d\"\n#>   ..@ ordered : logi TRUE\n#>   ..@ .S3Class: chr \"factor\"\n\nclass(x)\n#> [1] \"Ordered\"\n#> attr(,\"package\")\n#> [1] \".GlobalEnv\"\nFriends <- new(\"Person\", name = c(\"Vishu\", \"Aditi\"))\nlength.Person <- function(x) length(x@name)\n\nlength(Friends)\n#> [1] 2\nsetMethod(\"length\", \"Person\", function(x) length(x@name))\n\nlength(Friends)\n#> [1] 2"},{"path":"s4.html","id":"session-information-12","chapter":"15 S4","heading":"15.6 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version    date (UTC) lib source\n#>  assertthat    0.2.1      2019-03-21 [1] RSPM\n#>  base        * 4.3.2      2023-11-01 [3] local\n#>  bookdown      0.37       2023-12-01 [1] RSPM\n#>  bslib         0.6.1      2023-11-28 [1] RSPM\n#>  cachem        1.0.8      2023-05-01 [1] RSPM\n#>  cli           3.6.2      2023-12-11 [1] RSPM\n#>  compiler      4.3.2      2023-11-01 [3] local\n#>  crayon        1.5.2      2022-09-29 [1] RSPM\n#>  datasets    * 4.3.2      2023-11-01 [3] local\n#>  digest        0.6.33     2023-07-07 [1] RSPM\n#>  downlit       0.4.3      2023-06-29 [1] RSPM\n#>  emo           0.0.0.9000 2023-11-05 [1] Github (hadley/emo@3f03b11)\n#>  evaluate      0.23       2023-11-01 [1] RSPM\n#>  fastmap       1.1.1      2023-02-24 [1] RSPM\n#>  fs            1.6.3      2023-07-20 [1] RSPM\n#>  generics      0.1.3      2022-07-05 [1] RSPM\n#>  glue          1.6.2      2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2      2023-11-01 [3] local\n#>  grDevices   * 4.3.2      2023-11-01 [3] local\n#>  highr         0.10       2022-12-22 [1] RSPM\n#>  htmltools     0.5.7      2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4      2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8      2023-12-04 [1] RSPM\n#>  knitr         1.45       2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4      2023-11-07 [1] RSPM\n#>  lubridate   * 1.9.3      2023-09-27 [1] RSPM\n#>  magrittr    * 2.0.3      2022-03-30 [1] RSPM\n#>  memoise       2.0.1      2021-11-26 [1] RSPM\n#>  methods     * 4.3.2      2023-11-01 [3] local\n#>  purrr         1.0.2      2023-08-10 [1] RSPM\n#>  R6            2.5.1      2021-08-19 [1] RSPM\n#>  rlang         1.1.2      2023-11-04 [1] RSPM\n#>  rmarkdown     2.25       2023-09-18 [1] RSPM\n#>  sass          0.4.8      2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2      2021-12-06 [1] RSPM\n#>  stats       * 4.3.2      2023-11-01 [3] local\n#>  stringi       1.8.3      2023-12-11 [1] RSPM\n#>  stringr       1.5.1      2023-11-14 [1] RSPM\n#>  timechange    0.2.0      2023-01-11 [1] RSPM\n#>  tools         4.3.2      2023-11-01 [3] local\n#>  utils       * 4.3.2      2023-11-01 [3] local\n#>  vctrs         0.6.5      2023-12-01 [1] RSPM\n#>  withr         2.5.2      2023-10-30 [1] RSPM\n#>  xfun          0.41       2023-11-01 [1] RSPM\n#>  xml2          1.3.6      2023-12-04 [1] RSPM\n#>  yaml          2.3.8      2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"trade-offs.html","id":"trade-offs","chapter":"16 Trade-offs","heading":"16 Trade-offs","text":"exercises.","code":""},{"path":"big-picture.html","id":"big-picture","chapter":"17 Big Picture","heading":"17 Big Picture","text":"exercises.","code":""},{"path":"expressions.html","id":"expressions","chapter":"18 Expressions","heading":"18 Expressions","text":"Attaching needed libraries:","code":"\nlibrary(rlang, warn.conflicts = FALSE)\nlibrary(lobstr, warn.conflicts = FALSE)"},{"path":"expressions.html","id":"abstract-syntax-trees-exercises-18.2.4","chapter":"18 Expressions","heading":"18.1 Abstract syntax trees (Exercises 18.2.4)","text":"Q1. Reconstruct code represented trees :A1. reconstructed code.can confirm drawing ASTs :Q2. Draw following trees hand check answers ast().A2. Successfully drawn hand. Checking using ast():Q3. ’s happening ASTs ? (Hint: carefully read ?\"^\".)A3. str2expression() helps make sense ASTs.non-syntactic names parsed names. Thus, backticks removed AST.mentioned docs ^:** translated parser ^rightward assignment parsed leftward assignment:Q4. special AST ?A4. mentioned section:Like objects R, functions can also possess number additional attributes(). One attribute used base R srcref, short source reference. points source code used create function. srcref used printing , unlike body(), contains code comments formatting.Therefore, last leaf AST, although specified function call, represents source reference attribute.Q5. call tree statement multiple else conditions look like? ?A5. nothing special tree. just shows nested loop structure inherent code multiple else statements.","code":"#> █─f \n#> └─█─g \n#>   └─█─h\n#> █─`+` \n#> ├─█─`+` \n#> │ ├─1 \n#> │ └─2 \n#> └─3\n#> █─`*` \n#> ├─█─`(` \n#> │ └─█─`+` \n#> │   ├─x \n#> │   └─y \n#> └─z\nf(g(h()))\n1 + 2 + 3\n(x + y) * z\nast(f(g(h())))\n#> █─f \n#> └─█─g \n#>   └─█─h\n\nast(1 + 2 + 3)\n#> █─`+` \n#> ├─█─`+` \n#> │ ├─1 \n#> │ └─2 \n#> └─3\n\nast((x + y) * z)\n#> █─`*` \n#> ├─█─`(` \n#> │ └─█─`+` \n#> │   ├─x \n#> │   └─y \n#> └─z\nf(g(h(i(1, 2, 3))))\nf(1, g(2, h(3, i())))\nf(g(1, 2), h(3, i(4, 5)))\nast(f(g(h(i(1, 2, 3)))))\n#> █─f \n#> └─█─g \n#>   └─█─h \n#>     └─█─i \n#>       ├─1 \n#>       ├─2 \n#>       └─3\n\nast(f(1, g(2, h(3, i()))))\n#> █─f \n#> ├─1 \n#> └─█─g \n#>   ├─2 \n#>   └─█─h \n#>     ├─3 \n#>     └─█─i\n\nast(f(g(1, 2), h(3, i(4, 5))))\n#> █─f \n#> ├─█─g \n#> │ ├─1 \n#> │ └─2 \n#> └─█─h \n#>   ├─3 \n#>   └─█─i \n#>     ├─4 \n#>     └─5\nast(`x` + `y`)\n#> █─`+` \n#> ├─x \n#> └─y\nast(x**y)\n#> █─`^` \n#> ├─x \n#> └─y\nast(1 -> x)\n#> █─`<-` \n#> ├─x \n#> └─1\nstr2expression(\"`x` + `y`\")\n#> expression(x + y)\nstr2expression(\"x**y\")\n#> expression(x^y)\nstr2expression(\"1 -> x\")\n#> expression(x <- 1)\nast(function(x = 1, y = 2) {})\n#> █─`function` \n#> ├─█─x = 1 \n#> │ └─y = 2 \n#> ├─█─`{` \n#> └─<inline srcref>\nast(if (FALSE) 1 else if (FALSE) 2 else if (FALSE) 3 else 4)\n#> █─`if` \n#> ├─FALSE \n#> ├─1 \n#> └─█─`if` \n#>   ├─FALSE \n#>   ├─2 \n#>   └─█─`if` \n#>     ├─FALSE \n#>     ├─3 \n#>     └─4"},{"path":"expressions.html","id":"expressions-exercises-18.3.5","chapter":"18 Expressions","heading":"18.2 Expressions (Exercises 18.3.5)","text":"Q1. two six types atomic vector can’t appear expression? ? Similarly, can’t create expression contains atomic vector length greater one?A1. six types atomic vectors, two can’t appear expression : complex raw.Complex numbers created via function call (using +), can seen AST:Similarly, raw vectors (using raw()):Contrast atomic vectors:reason, can’t create expression contains atomic vector length greater one since function call uses c() function:Q2. happens subset call object remove first element? e.g. expr(read.csv(\"foo.csv\", header = TRUE))[-1]. ?A2. captured function call like following creates call object:mentioned respective section:first element call object function position.Therefore, first element call object removed, next one moves function position, get observed output:Q3. Describe differences following call objects.A4. differences constructed call objects due different type arguments supplied first two parameters call2() function.Types arguments supplied .fn:Types arguments supplied dynamic dots:following outputs can understood using following properties:.fn argument closure, function inlined constructed function callwhen x symbol, value passed function callImportantly, constructed call objects evaluate give result:Q4. call_standardise() doesn’t work well following calls. ? makes mean() special?A4. ellipsis mean() function signature:mentioned respective section:function uses ... ’s possible standardise arguments.mean() S3 generic dots passed underlying S3 methods., output can improved using specific method. example:Q5. code make sense?A5. doesn’t make sense first position call object reserved function (function position), assigning names element just ignored R:Q6. Construct expression (x > 1) \"\" else \"b\" using multiple calls call2(). code structure reflect structure AST?A6. Using multiple calls construct required expression:construction follows prefix form expression, revealed AST:","code":"\nx_complex <- expr(1 + 1i)\ntypeof(x_complex)\n#> [1] \"language\"\n\nast(1 + 1i)\n#> █─`+` \n#> ├─1 \n#> └─1i\nx_raw <- expr(raw(2))\ntypeof(x_raw)\n#> [1] \"language\"\n\nast(raw(2))\n#> █─raw \n#> └─2\nx_int <- expr(2L)\ntypeof(x_int)\n#> [1] \"integer\"\n\nast(2L)\n#> 2L\nx_vec <- expr(c(1, 2))\ntypeof(x_vec)\n#> [1] \"language\"\n\nast(c(1, 2))\n#> █─c \n#> ├─1 \n#> └─2\nexpr(read.csv(\"foo.csv\", header = TRUE))\n#> read.csv(\"foo.csv\", header = TRUE)\n\ntypeof(expr(read.csv(\"foo.csv\", header = TRUE)))\n#> [1] \"language\"\nexpr(read.csv(\"foo.csv\", header = TRUE))[-1]\n#> \"foo.csv\"(header = TRUE)\nx <- 1:10\ncall2(median, x, na.rm = TRUE)\ncall2(expr(median), x, na.rm = TRUE)\ncall2(median, expr(x), na.rm = TRUE)\ncall2(expr(median), expr(x), na.rm = TRUE)\ntypeof(median)\n#> [1] \"closure\"\ntypeof(expr(median))\n#> [1] \"symbol\"\nx <- 1:10\ntypeof(x)\n#> [1] \"integer\"\ntypeof(expr(x))\n#> [1] \"symbol\"\nx <- 1:10\n\ncall2(median, x, na.rm = TRUE)\n#> (function (x, na.rm = FALSE, ...) \n#> UseMethod(\"median\"))(1:10, na.rm = TRUE)\n\ncall2(expr(median), x, na.rm = TRUE)\n#> median(1:10, na.rm = TRUE)\n\ncall2(median, expr(x), na.rm = TRUE)\n#> (function (x, na.rm = FALSE, ...) \n#> UseMethod(\"median\"))(x, na.rm = TRUE)\n\ncall2(expr(median), expr(x), na.rm = TRUE)\n#> median(x, na.rm = TRUE)\nx <- 1:10\n\neval(call2(median, x, na.rm = TRUE))\n#> [1] 5.5\n\neval(call2(expr(median), x, na.rm = TRUE))\n#> [1] 5.5\n\neval(call2(median, expr(x), na.rm = TRUE))\n#> [1] 5.5\n\neval(call2(expr(median), expr(x), na.rm = TRUE))\n#> [1] 5.5\ncall_standardise(quote(mean(1:10, na.rm = TRUE)))\n#> Warning: `call_standardise()` is deprecated as of rlang 0.4.11\n#> This warning is displayed once every 8 hours.\n#> mean(x = 1:10, na.rm = TRUE)\ncall_standardise(quote(mean(n = T, 1:10)))\n#> mean(x = 1:10, n = T)\ncall_standardise(quote(mean(x = 1:10, , TRUE)))\n#> mean(x = 1:10, , TRUE)\nmean\n#> function (x, ...) \n#> UseMethod(\"mean\")\n#> <bytecode: 0x55d34a3275a8>\n#> <environment: namespace:base>\ncall_standardise(quote(mean.default(n = T, 1:10)))\n#> mean.default(x = 1:10, na.rm = T)\nx <- expr(foo(x = 1))\nnames(x) <- c(\"x\", \"y\")\nx <- expr(foo(x = 1))\nx\n#> foo(x = 1)\n\nnames(x) <- c(\"x\", \"y\")\nx\n#> foo(y = 1)\nx <- 5\ncall_obj1 <- call2(\">\", expr(x), 1)\ncall_obj1\n#> x > 1\n\ncall_obj2 <- call2(\"if\", cond = call_obj1, cons.expr = \"a\", alt.expr = \"b\")\ncall_obj2\n#> if (x > 1) \"a\" else \"b\"\nast(if (x > 1) \"a\" else \"b\")\n#> █─`if` \n#> ├─█─`>` \n#> │ ├─x \n#> │ └─1 \n#> ├─\"a\" \n#> └─\"b\""},{"path":"expressions.html","id":"parsing-and-grammar-exercises-18.4.4","chapter":"18 Expressions","heading":"18.3 Parsing and grammar (Exercises 18.4.4)","text":"Q1. R uses parentheses two slightly different ways illustrated two calls:Compare contrast two uses referencing AST.A1. Let’s first look AST:, can see ( used two separate ways:function right \"`(`\"part prefix syntax (f()), AST f((1)), see one \"`(`\" (first use case), f(), part function syntax (second use case).Q2. = can also used two ways. Construct simple example shows uses.A2. simple example illustrating = can also used two ways:assignmentfor named arguments function callsWe can also look AST:Q3. -2^2 yield 4 -4? ?A3. expression -2^2 evaluates -4 operator ^ higher precedence unary - operator:can also seen AST:less confusing way write :Q4. !1 + !1 return? ?A3. expression !1 + !1 evaluates FALSE.! operator higher precedence unary + operator. Thus, !1 evaluates FALSE, added 1 + FALSE, evaluates 1, logically negated !1, FALSE.can easily seen AST:Q5. x1 <- x2 <- x3 <- 0 work? Describe two reasons.A5. two reasons following works expected:<- operator right associative.Therefore, order assignment :<- operator invisibly returns assigned value.easy surmise AST:Q6. Compare ASTs x + y %+% z x ^ y %+% z. learned precedence custom infix functions?A6. Looking ASTs expressions,can say custom infix operator %+% :higher precedence + operatorlower precedence ^ operatorQ7. happens call parse_expr() string generates multiple expressions? e.g. parse_expr(\"x + 1; y + 1\")A7. produced error:expected based docs:parse_expr() returns one expression. text contains one expression (separated semicolons new lines), error issued.instead need use parse_exprs():Q8. happens attempt parse invalid expression? e.g. \"+\" \"f())\".A8. invalid expression produces error:Since underlying parse() function produces error:Q9. deparse() produces vectors input long. example, following call produces vector length two:expr_text() instead?A9. difference deparse() expr_text() latter turns (possibly multi-line) expression single string.Q10. pairwise.t.test() assumes deparse() always returns length one character vector. Can construct input violates expectation? happens?A10 Since R 4.0, possible violate expectation since new implementation produces single string matter input:New function deparse1() produces one string, wrapping deparse(), used typically deparse1(substitute(*))","code":"\nf((1))\n`(`(1 + 1)\nast(f((1)))\n#> █─f \n#> └─█─`(` \n#>   └─1\nast(`(`(1 + 1))\n#> █─`(` \n#> └─█─`+` \n#>   ├─1 \n#>   └─1\nm <- mean(x = 1)\nast({\n  m <- mean(x = 1)\n})\n#> █─`{` \n#> └─█─`<-` \n#>   ├─m \n#>   └─█─mean \n#>     └─x = 1\n-2^2\n#> [1] -4\nast(-2^2)\n#> █─`-` \n#> └─█─`^` \n#>   ├─2 \n#>   └─2\n-(2^2)\n#> [1] -4\nast(!1 + !1)\n#> █─`!` \n#> └─█─`+` \n#>   ├─1 \n#>   └─█─`!` \n#>     └─1\nx1 <- x2 <- x3 <- 0\n(x3 <- 0)\n(x2 <- x3)\n(x1 <- x2)\n(x <- 1)\n#> [1] 1\nast(x1 <- x2 <- x3 <- 0)\n#> █─`<-` \n#> ├─x1 \n#> └─█─`<-` \n#>   ├─x2 \n#>   └─█─`<-` \n#>     ├─x3 \n#>     └─0\nast(x + y %+% z)\n#> █─`+` \n#> ├─x \n#> └─█─`%+%` \n#>   ├─y \n#>   └─z\n\nast(x^y %+% z)\n#> █─`%+%` \n#> ├─█─`^` \n#> │ ├─x \n#> │ └─y \n#> └─z\nparse_expr(\"x + 1; y + 1\")\n#> Error in `parse_expr()`:\n#> ! `x` must contain exactly 1 expression, not 2.\nparse_exprs(\"x + 1; y + 1\")\n#> [[1]]\n#> x + 1\n#> \n#> [[2]]\n#> y + 1\nparse_expr(\"a +\")\n#> Error in parse(text = x, keep.source = FALSE): <text>:2:0: unexpected end of input\n#> 1: a +\n#>    ^\n\nparse_expr(\"f())\")\n#> Error in parse(text = x, keep.source = FALSE): <text>:1:4: unexpected ')'\n#> 1: f())\n#>        ^\nparse(text = \"a +\")\n#> Error in parse(text = \"a +\"): <text>:2:0: unexpected end of input\n#> 1: a +\n#>    ^\n\nparse(text = \"f())\")\n#> Error in parse(text = \"f())\"): <text>:1:4: unexpected ')'\n#> 1: f())\n#>        ^\nexpr <- expr(g(a + b + c + d + e + f + g + h + i + j + k + l +\n  m + n + o + p + q + r + s + t + u + v + w + x + y + z))\ndeparse(expr)\nexpr <- expr(g(a + b + c + d + e + f + g + h + i + j + k + l +\n  m + n + o + p + q + r + s + t + u + v + w + x + y + z))\n\ndeparse(expr)\n#> [1] \"g(a + b + c + d + e + f + g + h + i + j + k + l + m + n + o + \"\n#> [2] \"    p + q + r + s + t + u + v + w + x + y + z)\"\n\nexpr_text(expr)\n#> [1] \"g(a + b + c + d + e + f + g + h + i + j + k + l + m + n + o + \\n    p + q + r + s + t + u + v + w + x + y + z)\""},{"path":"expressions.html","id":"walking-ast-with-recursive-functions-exercises-18.5.3","chapter":"18 Expressions","heading":"18.4 Walking AST with recursive functions (Exercises 18.5.3)","text":"Q1. logical_abbr() returns TRUE T(1, 2, 3). modify logical_abbr_rec() ignores function calls use T F?A1. avoid function calls use T F, just need ignore function position call objects:Let’s try :Q2. logical_abbr() works expressions. currently fails give function. ? modify logical_abbr() make work? components function need recurse ?A2. Surprisingly, logical_abbr() currently doesn’t fail closures:see , let’s see type object produced capture user provided closure:Given closures converted call objects, surprise function works:function fails can’t find negative case. example, instead returning FALSE, produces error reasons remain (yet) elusive :Q3. Modify find_assign also detect assignment using replacement functions, .e. names(x) <- y.A3. Although simple assignment (x <- y) assignment using replacement functions (names(x) <- y) <- operator call, latter case, names(x) call object symbol:’s can detect kind assignment checking second element expression symbol language type object.Let’s try :Q4. Write function extracts calls specified function.A4. function extracts calls specified function:","code":"\nlogical_abbr_rec(expr(T(1, 2, 3)))\n#> [1] FALSE\n\nlogical_abbr_rec(expr(F(1, 2, 3)))\n#> [1] FALSE\n\nlogical_abbr_rec(expr(T))\n#> [1] TRUE\n\nlogical_abbr_rec(expr(F))\n#> [1] TRUE\nlogical_abbr(function(x = TRUE) {\n  g(x + T)\n})\nprint_enexpr <- function(.f) {\n  print(typeof(enexpr(.f)))\n  print(is.call(enexpr(.f)))\n}\n\nprint_enexpr(function(x = TRUE) {\n  g(x + T)\n})\n#> [1] \"language\"\n#> [1] TRUE\nlogical_abbr(function(x = TRUE) {\n  g(x + T)\n})\n#> [1] TRUE\nlogical_abbr(function(x = TRUE) {\n  g(x + TRUE)\n})\n#> Error: Don't know how to handle type integer\nexpr1 <- expr(names(x) <- y)\nas.list(expr1)\n#> [[1]]\n#> `<-`\n#> \n#> [[2]]\n#> names(x)\n#> \n#> [[3]]\n#> y\ntypeof(expr1[[2]])\n#> [1] \"language\"\n\nexpr2 <- expr(x <- y)\nas.list(expr2)\n#> [[1]]\n#> `<-`\n#> \n#> [[2]]\n#> x\n#> \n#> [[3]]\n#> y\ntypeof(expr2[[2]])\n#> [1] \"symbol\"\nexpr_type <- function(x) {\n  if (is_syntactic_literal(x)) {\n    \"constant\"\n  } else if (is.symbol(x)) {\n    \"symbol\"\n  } else if (is.call(x)) {\n    \"call\"\n  } else if (is.pairlist(x)) {\n    \"pairlist\"\n  } else {\n    typeof(x)\n  }\n}\n\nswitch_expr <- function(x, ...) {\n  switch(expr_type(x),\n    ...,\n    stop(\"Don't know how to handle type \", typeof(x), call. = FALSE)\n  )\n}\n\nflat_map_chr <- function(.x, .f, ...) {\n  purrr::flatten_chr(purrr::map(.x, .f, ...))\n}\n\nextract_symbol <- function(x) {\n  if (is_symbol(x[[2]])) {\n    as_string(x[[2]])\n  } else {\n    extract_symbol(as.list(x[[2]]))\n  }\n}\n\nfind_assign_call <- function(x) {\n  if (is_call(x, \"<-\") && is_symbol(x[[2]])) {\n    lhs <- as_string(x[[2]])\n    children <- as.list(x)[-1]\n  } else if (is_call(x, \"<-\") && is_call(x[[2]])) {\n    lhs <- extract_symbol(as.list(x[[2]]))\n    children <- as.list(x)[-1]\n  } else {\n    lhs <- character()\n    children <- as.list(x)\n  }\n\n  c(lhs, flat_map_chr(children, find_assign_rec))\n}\n\nfind_assign_rec <- function(x) {\n  switch_expr(x,\n    # Base cases\n    constant = ,\n    symbol = character(),\n\n    # Recursive cases\n    pairlist = flat_map_chr(x, find_assign_rec),\n    call = find_assign_call(x)\n  )\n}\n\nfind_assign <- function(x) find_assign_rec(enexpr(x))\nfind_assign(names(x))\n#> character(0)\n\nfind_assign(names(x) <- y)\n#> [1] \"x\"\n\nfind_assign(names(f(x)) <- y)\n#> [1] \"x\"\n\nfind_assign(names(x) <- y <- z <- NULL)\n#> [1] \"x\" \"y\" \"z\"\n\nfind_assign(a <- b <- c <- 1)\n#> [1] \"a\" \"b\" \"c\"\n\nfind_assign(system.time(x <- print(y <- 5)))\n#> [1] \"x\" \"y\"\nfind_function_call <- function(x, .f) {\n  if (is_call(x)) {\n    if (is_call(x, .f)) {\n      list(x)\n    } else {\n      purrr::map(as.list(x), ~ find_function_call(.x, .f)) %>%\n        purrr::compact() %>%\n        unlist(use.names = FALSE)\n    }\n  }\n}\n\n# example-1: with infix operator `:`\nfind_function_call(expr(mean(1:2)), \":\")\n#> [[1]]\n#> 1:2\n\nfind_function_call(expr(sum(mean(1:2))), \":\")\n#> [[1]]\n#> 1:2\n\nfind_function_call(expr(list(1:5, 4:6, 3:9)), \":\")\n#> [[1]]\n#> 1:5\n#> \n#> [[2]]\n#> 4:6\n#> \n#> [[3]]\n#> 3:9\n\nfind_function_call(expr(list(1:5, sum(4:6), mean(3:9))), \":\")\n#> [[1]]\n#> 1:5\n#> \n#> [[2]]\n#> 4:6\n#> \n#> [[3]]\n#> 3:9\n\n# example-2: with assignment operator `<-`\nfind_function_call(expr(names(x)), \"<-\")\n#> NULL\n\nfind_function_call(expr(names(x) <- y), \"<-\")\n#> [[1]]\n#> names(x) <- y\n\nfind_function_call(expr(names(f(x)) <- y), \"<-\")\n#> [[1]]\n#> names(f(x)) <- y\n\nfind_function_call(expr(names(x) <- y <- z <- NULL), \"<-\")\n#> [[1]]\n#> names(x) <- y <- z <- NULL\n\nfind_function_call(expr(a <- b <- c <- 1), \"<-\")\n#> [[1]]\n#> a <- b <- c <- 1\n\nfind_function_call(expr(system.time(x <- print(y <- 5))), \"<-\")\n#> [[1]]\n#> x <- print(y <- 5)"},{"path":"expressions.html","id":"session-information-13","chapter":"18 Expressions","heading":"18.5 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  crayon        1.5.2   2022-09-29 [1] RSPM\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  lobstr      * 1.1.2   2022-06-22 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  purrr         1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"quasiquotation.html","id":"quasiquotation","chapter":"19 Quasiquotation","heading":"19 Quasiquotation","text":"Attaching needed libraries:","code":"\nlibrary(rlang)\nlibrary(purrr)\nlibrary(lobstr)\nlibrary(dplyr)\nlibrary(ggplot2)"},{"path":"quasiquotation.html","id":"motivation-exercises-19.2.2","chapter":"19 Quasiquotation","heading":"19.1 Motivation (Exercises 19.2.2)","text":"Q1. function following base R code, identify arguments quoted evaluated.A1. identify arguments quoted evaluated, can use trick mentioned book:’re ever unsure whether argument quoted evaluated, try executing code outside function. doesn’t work something different, argument quoted.library(MASS)package argument library() quoted:subset(mtcars, cyl == 4)argument x evaluated, argument subset quoted.(mtcars2, sum(vs))argument data evaluated, expr argument quoted.sum(mtcars2$)argument ... evaluated.rm(mtcars2)trick using far won’t work since trying print mtcars2 always fail rm() made pass .can instead look docs ...:… objects removed, names (unquoted) character strings (quoted).Thus, argument evaluated, rather quoted.Q2. function following tidyverse code, identify arguments quoted evaluated.A2. seen answer Q1., library() quotes first argument:following code:%>% (lazily) evaluates argumentgroup_by() summarise() quote argumentsIn following code:ggplot() evaluates data argumentaes() quotes arguments","code":"\nlibrary(MASS)\n\nmtcars2 <- subset(mtcars, cyl == 4)\n\nwith(mtcars2, sum(vs))\nsum(mtcars2$am)\n\nrm(mtcars2)\nlibrary(MASS)\n\nMASS\n#> Error in eval(expr, envir, enclos): object 'MASS' not found\nmtcars2 <- subset(mtcars, cyl == 4)\n\ninvisible(mtcars)\n\ncyl == 4\n#> Error in eval(expr, envir, enclos): object 'cyl' not found\nwith(mtcars2, sum(vs))\n#> [1] 10\n\ninvisible(mtcars2)\n\nsum(vs)\n#> Error in eval(expr, envir, enclos): object 'vs' not found\nsum(mtcars2$am)\n#> [1] 8\n\nmtcars2$am\n#>  [1] 1 0 0 1 1 1 0 1 1 1 1\nrm(mtcars2)\nlibrary(dplyr)\nlibrary(ggplot2)\n\nby_cyl <- mtcars %>%\n  group_by(cyl) %>%\n  summarise(mean = mean(mpg))\n\nggplot(by_cyl, aes(cyl, mean)) +\n  geom_point()\nlibrary(dplyr)\nlibrary(ggplot2)\nby_cyl <- mtcars %>%\n  group_by(cyl) %>%\n  summarise(mean = mean(mpg))\nggplot(by_cyl, aes(cyl, mean)) +\n  geom_point()"},{"path":"quasiquotation.html","id":"quoting-exercises-19.3.6","chapter":"19 Quasiquotation","heading":"19.2 Quoting (Exercises 19.3.6)","text":"Q1. expr() implemented? Look source code.A1. Looking source code, can see expr() simple wrapper around enexpr(), captures returns user-entered expressions:example:turn, enexpr() calls native code:Q2. Compare contrast following two functions. Can predict output running ?A2. exprs() captures returns expressions specified developer instead values:hand, enexprs() captures user-entered expressions returns values:Q3. happens try use enexpr() expression (.e. enexpr(x + y)? happens enexpr() passed missing argument?A3. try use enexpr() expression, fails works symbol.enexpr() passed missing argument, returns missing argument:Q4. exprs() exprs(= ) different? Think input output.A4. key difference exprs() exprs(= ) former return unnamed list, latter return named list. former interpreted unnamed argument, latter named argument.cases, treated symbol:, argument missing latter case, since name corresponding value provided:Q5. differences exprs() alist()? Read documentation named arguments exprs() find .A5. additional differences exprs() alist().Names: inputs named, exprs() provides way name automatically using .named argument.Ignoring empty arguments: .ignore_empty argument exprs() gives much finer control empty arguments, alist() doesn’t provide way ignore arguments.Names injection: Using .unquote_names argument exprs(), can inject name argument.Q6. documentation substitute() says:Substitution takes place examining component parse tree\nfollows:bound symbol env, unchanged.promise object (.e., formal argument function)\nexpression slot promise replaces symbol.ordinary variable, value substituted, unless\nenv .GlobalEnv case symbol left unchanged.Create examples illustrate cases.A6. See examples illustrate -mentioned cases.bound symbol env, unchanged.Symbol x bound env, remains unchanged.promise object (.e., formal argument function)\nexpression slot promise replaces symbol.ordinary variable, value substituted, unless\nenv .GlobalEnv case symbol left unchanged.","code":"\nrlang::expr\n#> function (expr) \n#> {\n#>     enexpr(expr)\n#> }\n#> <bytecode: 0x55b385e4f448>\n#> <environment: namespace:rlang>\nx <- expr(x <- 1)\nx\n#> x <- 1\nrlang::enexpr\n#> function (arg) \n#> {\n#>     .Call(ffi_enexpr, substitute(arg), parent.frame())\n#> }\n#> <bytecode: 0x55b3829aaae0>\n#> <environment: namespace:rlang>\nf1 <- function(x, y) {\n  exprs(x = x, y = y)\n}\nf2 <- function(x, y) {\n  enexprs(x = x, y = y)\n}\nf1(a + b, c + d)\nf2(a + b, c + d)\nf1 <- function(x, y) {\n  exprs(x = x, y = y)\n}\n\nf1(a + b, c + d)\n#> $x\n#> x\n#> \n#> $y\n#> y\nf2 <- function(x, y) {\n  enexprs(x = x, y = y)\n}\n\nf2(a + b, c + d)\n#> $x\n#> a + b\n#> \n#> $y\n#> c + d\nenexpr(x + y)\n#> Error in `enexpr()`:\n#> ! `arg` must be a symbol\narg <- missing_arg()\n\nenexpr(arg)\n\nis_missing(enexpr(arg))\n#> [1] TRUE\nexprs(a)\n#> [[1]]\n#> a\n\nexprs(a = )\n#> $a\nmap_lgl(exprs(a), is_symbol)\n#>      \n#> TRUE\n\nmap_lgl(exprs(a = ), is_symbol)\n#>    a \n#> TRUE\nmap_lgl(exprs(a), is_missing)\n#>       \n#> FALSE\n\nmap_lgl(exprs(a = ), is_missing)\n#>    a \n#> TRUE\nalist(\"x\" = 1, TRUE, \"z\" = expr(x + y))\n#> $x\n#> [1] 1\n#> \n#> [[2]]\n#> [1] TRUE\n#> \n#> $z\n#> expr(x + y)\n\nexprs(\"x\" = 1, TRUE, \"z\" = expr(x + y), .named = TRUE)\n#> $x\n#> [1] 1\n#> \n#> $`TRUE`\n#> [1] TRUE\n#> \n#> $z\n#> expr(x + y)\nalist(\"x\" = 1, , TRUE, )\n#> $x\n#> [1] 1\n#> \n#> [[2]]\n#> \n#> \n#> [[3]]\n#> [1] TRUE\n#> \n#> [[4]]\n\nexprs(\"x\" = 1, , TRUE, , .ignore_empty = \"trailing\")\n#> $x\n#> [1] 1\n#> \n#> [[2]]\n#> \n#> \n#> [[3]]\n#> [1] TRUE\n\nexprs(\"x\" = 1, , TRUE, , .ignore_empty = \"none\")\n#> $x\n#> [1] 1\n#> \n#> [[2]]\n#> \n#> \n#> [[3]]\n#> [1] TRUE\n#> \n#> [[4]]\n\nexprs(\"x\" = 1, , TRUE, , .ignore_empty = \"all\")\n#> $x\n#> [1] 1\n#> \n#> [[2]]\n#> [1] TRUE\nalist(foo := bar)\n#> [[1]]\n#> `:=`(foo, bar)\n\nexprs(foo := bar, .unquote_names = FALSE)\n#> [[1]]\n#> `:=`(foo, bar)\n\nexprs(foo := bar, .unquote_names = TRUE)\n#> $foo\n#> bar\nsubstitute(x + y, env = list(y = 2))\n#> x + 2\nmsg <- \"old\"\ndelayedAssign(\"myVar\", msg) # creates a promise\nsubstitute(myVar)\n#> myVar\nmsg <- \"new!\"\nmyVar\n#> [1] \"new!\"\nsubstitute(x + y, env = env(x = 2, y = 1))\n#> 2 + 1\n\nx <- 2\ny <- 1\nsubstitute(x + y, env = .GlobalEnv)\n#> x + y"},{"path":"quasiquotation.html","id":"unquoting-exercises-19.4.8","chapter":"19 Quasiquotation","heading":"19.3 Unquoting (Exercises 19.4.8)","text":"Q1. Given following components:Use quasiquotation construct following calls:A1. Using quasiquotation construct specified calls:Q2. following two calls print , actually different:’s difference? one natural?A2. can see difference two expression convert lists:can seen, second element call object, b integer vector:can also noticed ASTs expressions:first call natural, since second one inlines vector directly call, something rarely done.","code":"\nxy <- expr(x + y)\nxz <- expr(x + z)\nyz <- expr(y + z)\nabc <- exprs(a, b, c)\n(x + y) / (y + z)\n-(x + z)^(y + z)\n(x + y) + (y + z) - (x + y)\natan2(x + y, y + z)\nsum(x + y, x + y, y + z)\nsum(a, b, c)\nmean(c(a, b, c), na.rm = TRUE)\nfoo(a = x + y, b = y + z)\nxy <- expr(x + y)\nxz <- expr(x + z)\nyz <- expr(y + z)\nabc <- exprs(a, b, c)\n\nexpr((!!xy) / (!!yz))\n#> (x + y)/(y + z)\n\nexpr(-(!!xz)^(!!yz))\n#> -(x + z)^(y + z)\n\nexpr(((!!xy)) + (!!yz) - (!!xy))\n#> (x + y) + (y + z) - (x + y)\n\ncall2(\"atan2\", expr(!!xy), expr(!!yz))\n#> atan2(x + y, y + z)\n\ncall2(\"sum\", expr(!!xy), expr(!!xy), expr(!!yz))\n#> sum(x + y, x + y, y + z)\n\ncall2(\"sum\", !!!abc)\n#> sum(a, b, c)\n\nexpr(mean(c(!!!abc), na.rm = TRUE))\n#> mean(c(a, b, c), na.rm = TRUE)\n\ncall2(\"foo\", a = expr(!!xy), b = expr(!!yz))\n#> foo(a = x + y, b = y + z)\n(a <- expr(mean(1:10)))\n#> mean(1:10)\n(b <- expr(mean(!!(1:10))))\n#> mean(1:10)\nidentical(a, b)\n#> [1] FALSE\nas.list(expr(mean(1:10)))\n#> [[1]]\n#> mean\n#> \n#> [[2]]\n#> 1:10\n\nas.list(expr(mean(!!(1:10))))\n#> [[1]]\n#> mean\n#> \n#> [[2]]\n#>  [1]  1  2  3  4  5  6  7  8  9 10\nwaldo::compare(a, b)\n#> `old[[2]]` is a call\n#> `new[[2]]` is an integer vector (1, 2, 3, 4, 5, ...)\nast(expr(mean(1:10)))\n#> █─expr \n#> └─█─mean \n#>   └─█─`:` \n#>     ├─1 \n#>     └─10\n\nast(expr(mean(!!(1:10))))\n#> █─expr \n#> └─█─mean \n#>   └─<inline integer>"},{"path":"quasiquotation.html","id":"dot-dot-dot-exercises-19.6.5","chapter":"19 Quasiquotation","heading":"19.4 ... (dot-dot-dot) (Exercises 19.6.5)","text":"Q1. One way implement exec() shown . Describe works. key ideas?A1. keys ideas underlie implementation exec() function following:constructs call using function f argument ..., evaluates call environment .env.constructs call using function f argument ..., evaluates call environment .env.uses dynamic dots via list2(), means can splice arguments using !!!, can inject names using :=, trailing commas problem.uses dynamic dots via list2(), means can splice arguments using !!!, can inject names using :=, trailing commas problem.example:Q2. Carefully read source code interaction(), expand.grid(), par(). Compare contrast techniques use switching dots list behaviour.A2. Source code reveals following comparison table:functions capture dots list.Using dots, functions check:list entered argument checking number argumentsif count 1, checking argument listQ3. Explain problem definition set_attr()A3. set_attr() function signature parameter called x, additionally uses dynamic dots pass multiple arguments specify additional attributes x., shown example, creates problem attribute named x. Naming arguments won’t help either:can avoid issues renaming parameter:","code":"\nexec <- function(f, ..., .env = caller_env()) {\n  args <- list2(...)\n  do.call(f, args, envir = .env)\n}\nvec <- c(1:5, NA)\nargs_list <- list(trim = 0, na.rm = TRUE)\n\nexec(mean, vec, !!!args_list, , .env = caller_env())\n#> [1] 3\n\nrm(\"exec\")\nset_attr <- function(x, ...) {\n  attr <- rlang::list2(...)\n  attributes(x) <- attr\n  x\n}\nset_attr(1:10, x = 10)\n#> Error in attributes(x) <- attr: attributes must be named\nset_attr <- function(x, ...) {\n  attr <- rlang::list2(...)\n  attributes(x) <- attr\n  x\n}\nset_attr(x = 1:10, x = 10)\n#> Error in set_attr(x = 1:10, x = 10): formal argument \"x\" matched by multiple actual arguments\nset_attr <- function(.x, ...) {\n  attr <- rlang::list2(...)\n  attributes(.x) <- attr\n  .x\n}\n\nset_attr(.x = 1:10, x = 10)\n#>  [1]  1  2  3  4  5  6  7  8  9 10\n#> attr(,\"x\")\n#> [1] 10"},{"path":"quasiquotation.html","id":"case-studies-exercises-19.7.5","chapter":"19 Quasiquotation","heading":"19.5 Case studies (Exercises 19.7.5)","text":"Q1. linear-model example, replace expr() reduce(summands, ~ expr(!!.x + !!.y)) call2(): reduce(summands, call2, \"+\"). Compare contrast two approaches. think easier read?A1. can rewrite linear() function chapter using call2() follows:personally find version call2() much readable since !! syntax bit esoteric.Q2. Re-implement Box-Cox transform defined using unquoting new_function():A2. Re-implementation Box-Cox transform using unquoting new_function():Let’s try see produces output :Q3. Re-implement simple compose() defined using quasiquotation new_function():A3. Following re-implementation compose() using quasiquotation new_function():Checking new version behaves way original version:","code":"\nlinear <- function(var, val) {\n  var <- ensym(var)\n  coef_name <- map(seq_along(val[-1]), ~ expr((!!var)[[!!.x]]))\n\n  summands <- map2(val[-1], coef_name, ~ expr((!!.x * !!.y)))\n  summands <- c(val[[1]], summands)\n\n  reduce(summands, ~ call2(\"+\", .x, .y))\n}\n\nlinear(x, c(10, 5, -4))\n#> 10 + (5 * x[[1L]]) + (-4 * x[[2L]])\nbc <- function(lambda) {\n  if (lambda == 0) {\n    function(x) log(x)\n  } else {\n    function(x) (x^lambda - 1) / lambda\n  }\n}\nbc_new <- function(lambda) {\n  lambda <- enexpr(lambda)\n\n  if (!!lambda == 0) {\n    new_function(\n      exprs(x = ),\n      expr(log(x))\n    )\n  } else {\n    new_function(\n      exprs(x = ),\n      expr((x^(!!lambda) - 1) / (!!lambda))\n    )\n  }\n}\nbc(0)(1)\n#> [1] 0\nbc_new(0)(1)\n#> [1] 0\n\nbc(2)(2)\n#> [1] 1.5\nbc_new(2)(2)\n#> [1] 1.5\ncompose <- function(f, g) {\n  function(...) f(g(...))\n}\ncompose_new <- function(f, g) {\n  f <- enexpr(f) # or ensym(f)\n  g <- enexpr(g) # or ensym(g)\n\n  new_function(\n    exprs(... = ),\n    expr((!!f)((!!g)(...)))\n  )\n}\nnot_null <- compose(`!`, is.null)\nnot_null(4)\n#> [1] TRUE\n\nnot_null2 <- compose_new(`!`, is.null)\nnot_null2(4)\n#> [1] TRUE"},{"path":"quasiquotation.html","id":"session-information-14","chapter":"19 Quasiquotation","heading":"19.6 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  colorspace    2.1-0   2023-01-23 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  crayon        1.5.2   2022-09-29 [1] RSPM\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  diffobj       0.3.5   2021-10-05 [1] RSPM\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr       * 1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  farver        2.1.1   2022-07-06 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  ggplot2     * 3.4.4   2023-10-12 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  grid          4.3.2   2023-11-01 [3] local\n#>  gtable        0.3.4   2023-08-21 [1] RSPM\n#>  highr         0.10    2022-12-22 [1] RSPM\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  labeling      0.4.3   2023-08-29 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  lobstr      * 1.1.2   2022-06-22 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  MASS        * 7.3-60  2023-05-04 [3] CRAN (R 4.3.2)\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  munsell       0.5.0   2018-06-12 [1] RSPM\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  purrr       * 1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rematch2      2.1.2   2020-05-01 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  scales        1.3.0   2023-11-28 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  waldo         0.5.2   2023-11-02 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"evaluation.html","id":"evaluation","chapter":"20 Evaluation","heading":"20 Evaluation","text":"Attaching needed libraries:","code":"\nlibrary(rlang)"},{"path":"evaluation.html","id":"evaluation-basics-exercises-20.2.4","chapter":"20 Evaluation","heading":"20.1 Evaluation basics (Exercises 20.2.4)","text":"Q1. Carefully read documentation source(). environment use default? supply local = TRUE? provide custom environment?A1. parameter local source() decides environment parsed expressions evaluated.default local = FALSE, corresponds user’s workspace (global environment, .e.).local = TRUE, environment source() called used.specify custom environment, sys.source() function can used, provides envir parameter.Q2. Predict results following lines code:A2. Correctly predicted 😉Q3. Fill function bodies re-implement get() using sym() eval(), assign() using sym(), expr(), eval(). Don’t worry multiple ways choosing environment get() assign() support; assume user supplies explicitly.A3. required re-implementations:get()assign()Q4. Modify source2() returns result every expression, just last one. Can eliminate loop?A4. can use purrr::map() iterate every expression return result every expression:Q5. can make base::local() slightly easier understand spreading multiple lines:Explain local() works words. (Hint: might want print(call) help understand substitute() , read documentation remind environment new.env() inherit .)A5. order figure function works, let’s add suggested print(call):docs substitute() mention:Substituting quoting often cause confusion argument expression(…). result call expression constructor function needs evaluated eval give actual expression object.Thus, get actual expression object, quoted expression needs evaluated using eval():Finally, generated call evaluated caller environment. final function call looks like following:Note bindings x y found inner environment, bindings functions eval(), quote(), etc. found outer environment.","code":"\nwithr::with_tempdir(\n  code = {\n    f <- tempfile()\n    writeLines(\"rlang::env_print()\", f)\n    foo <- function() source(f, local = FALSE)\n    foo()\n  }\n)\n#> <environment: global>\n#> Parent: <environment: package:rlang>\n#> Bindings:\n#> • .Random.seed: <int>\n#> • foo: <fn>\n#> • f: <chr>\nwithr::with_tempdir(\n  code = {\n    f <- tempfile()\n    writeLines(\"rlang::env_print()\", f)\n    foo <- function() source(f, local = TRUE)\n    foo()\n  }\n)\n#> <environment: 0x56269fbe7e18>\n#> Parent: <environment: global>\neval(expr(eval(expr(eval(expr(2 + 2))))))\neval(eval(expr(eval(expr(eval(expr(2 + 2)))))))\nexpr(eval(expr(eval(expr(eval(expr(2 + 2)))))))\neval(expr(eval(expr(eval(expr(2 + 2))))))\n#> [1] 4\n\neval(eval(expr(eval(expr(eval(expr(2 + 2)))))))\n#> [1] 4\n\nexpr(eval(expr(eval(expr(eval(expr(2 + 2)))))))\n#> eval(expr(eval(expr(eval(expr(2 + 2))))))\n# name is a string\nget2 <- function(name, env) {}\nassign2 <- function(name, value, env) {}\nget2 <- function(name, env = caller_env()) {\n  name <- sym(name)\n  eval(name, env)\n}\n\nx <- 2\n\nget2(\"x\")\n#> [1] 2\nget(\"x\")\n#> [1] 2\n\ny <- 1:4\nassign(\"y[1]\", 2)\n\nget2(\"y[1]\")\n#> [1] 2\nget(\"y[1]\")\n#> [1] 2\nassign2 <- function(name, value, env = caller_env()) {\n  name <- sym(name)\n  eval(expr(!!name <- !!value), env)\n}\n\nassign(\"y1\", 4)\ny1\n#> [1] 4\n\nassign2(\"y2\", 4)\ny2\n#> [1] 4\nsource2 <- function(path, env = caller_env()) {\n  file <- paste(readLines(path, warn = FALSE), collapse = \"\\n\")\n  exprs <- parse_exprs(file)\n  purrr::map(exprs, ~ eval(.x, env))\n}\n\nwithr::with_tempdir(\n  code = {\n    f <- tempfile(fileext = \".R\")\n    writeLines(\"1 + 1; 2 + 4\", f)\n    source2(f)\n  }\n)\n#> [[1]]\n#> [1] 2\n#> \n#> [[2]]\n#> [1] 6\nlocal3 <- function(expr, envir = new.env()) {\n  call <- substitute(eval(quote(expr), envir))\n  eval(call, envir = parent.frame())\n}\nlocal3 <- function(expr, envir = new.env()) {\n  call <- substitute(eval(quote(expr), envir))\n  print(call)\n\n  eval(call, envir = parent.frame())\n}\n\nlocal3({\n  x <- 10\n  y <- 200\n  x + y\n})\n#> eval(quote({\n#>     x <- 10\n#>     y <- 200\n#>     x + y\n#> }), new.env())\n#> [1] 210\nis_expression(eval(quote({\n  x <- 10\n  y <- 200\n  x + y\n}), new.env()))\n#> [1] TRUE\n# outer environment\neval(\n  # inner environment\n  eval(quote({\n    x <- 10\n    y <- 200\n    x + y\n  }), new.env()),\n  envir = parent.frame()\n)"},{"path":"evaluation.html","id":"quosures-exercises-20.3.6","chapter":"20 Evaluation","heading":"20.2 Quosures (Exercises 20.3.6)","text":"Q1. Predict following quosures return evaluated.A1. Correctly predicted 😉Q2. Write enenv() function captures environment associated argument. (Hint: require two function calls.)A2. can make use get_env() helper get environment associated argument:","code":"\nq1 <- new_quosure(expr(x), env(x = 1))\nq1\n#> <quosure>\n#> expr: ^x\n#> env:  0x56269fa39110\nq2 <- new_quosure(expr(x + !!q1), env(x = 10))\nq2\n#> <quosure>\n#> expr: ^x + (^x)\n#> env:  0x5626a008bd48\nq3 <- new_quosure(expr(x + !!q2), env(x = 100))\nq3\n#> <quosure>\n#> expr: ^x + (^x + (^x))\n#> env:  0x5626a034a820\nq1 <- new_quosure(expr(x), env(x = 1))\neval_tidy(q1)\n#> [1] 1\n\nq2 <- new_quosure(expr(x + !!q1), env(x = 10))\neval_tidy(q2)\n#> [1] 11\n\nq3 <- new_quosure(expr(x + !!q2), env(x = 100))\neval_tidy(q3)\n#> [1] 111\nenenv <- function(x) {\n  x <- enquo(x)\n  get_env(x)\n}\n\nenenv(x)\n#> <environment: R_GlobalEnv>\n\nfoo <- function(x) enenv(x)\nfoo()\n#> <environment: 0x5626a12b70c8>"},{"path":"evaluation.html","id":"data-masks-exercises-20.4.6","chapter":"20 Evaluation","heading":"20.3 Data masks (Exercises 20.4.6)","text":"Q1. use loop transform2() instead map()? Consider transform2(df, x = x * 2, x = x * 2).A1. see map() appropriate function, let’s create version function map() see happens.use () loop, iteration, updating x column current expression evaluation. , repeatedly modifying column works.use map() instead, trying evaluate expressions time; .e., column attempted modify using multiple expressions.Q2. ’s alternative implementation subset2():Compare contrast subset3() subset2(). advantages disadvantages?A2. Let’s first juxtapose functions outputs can compare better.Disadvantages subset3() subset2()filtering conditions specified rows don’t evaluate logical, function doesn’t fail informatively. Indeed, silently returns incorrect result.Advantages subset3() subset2()might argue function shorter advantage, much subjective preference.Q3. following function implements basics dplyr::arrange(). Annotate line comment explaining . Can explain !!.na.last strictly correct, omitting !! unlikely cause problems?A3. Annotated version function:see doesn’t matter whether whether unquote .na.last argument , let’s look smaller example:can seen:without unquoting, .na.last found function environmentwith unquoting, .na.last included order call object ","code":"\ntransform2 <- function(.data, ...) {\n  dots <- enquos(...)\n\n  for (i in seq_along(dots)) {\n    name <- names(dots)[[i]]\n    dot <- dots[[i]]\n\n    .data[[name]] <- eval_tidy(dot, .data)\n  }\n\n  .data\n}\n\ntransform3 <- function(.data, ...) {\n  dots <- enquos(...)\n\n  purrr::map(dots, function(x, .data = .data) {\n    name <- names(x)\n    dot <- x\n\n    .data[[name]] <- eval_tidy(dot, .data)\n\n    .data\n  })\n}\ndf <- data.frame(x = 1:3)\ntransform2(df, x = x * 2, x = x * 2)\n#>    x\n#> 1  4\n#> 2  8\n#> 3 12\ndf <- data.frame(x = 1:3)\ntransform3(df, x = x * 2, x = x * 2)\n#> Error in `purrr::map()`:\n#> ℹ In index: 1.\n#> ℹ With name: x.\n#> Caused by error:\n#> ! promise already under evaluation: recursive default argument reference or earlier problems?\nsubset3 <- function(data, rows) {\n  rows <- enquo(rows)\n  eval_tidy(expr(data[!!rows, , drop = FALSE]), data = data)\n}\ndf <- data.frame(x = 1:3)\nsubset3(df, x == 1)\nsubset2 <- function(data, rows) {\n  rows <- enquo(rows)\n  rows_val <- eval_tidy(rows, data)\n  stopifnot(is.logical(rows_val))\n\n  data[rows_val, , drop = FALSE]\n}\n\ndf <- data.frame(x = 1:3)\nsubset2(df, x == 1)\n#>   x\n#> 1 1\nsubset3 <- function(data, rows) {\n  rows <- enquo(rows)\n  eval_tidy(expr(data[!!rows, , drop = FALSE]), data = data)\n}\n\nsubset3(df, x == 1)\n#>   x\n#> 1 1\nrm(\"x\")\nexists(\"x\")\n#> [1] FALSE\n\nsubset2(df, x + 1)\n#> Error in subset2(df, x + 1): is.logical(rows_val) is not TRUE\n\nsubset3(df, x + 1)\n#>     x\n#> 2   2\n#> 3   3\n#> NA NA\narrange2 <- function(.df, ..., .na.last = TRUE) {\n  args <- enquos(...)\n  order_call <- expr(order(!!!args, na.last = !!.na.last))\n  ord <- eval_tidy(order_call, .df)\n  stopifnot(length(ord) == nrow(.df))\n  .df[ord, , drop = FALSE]\n}\narrange2 <- function(.df, ..., .na.last = TRUE) {\n  # capture user-supplied expressions (and corresponding environments) as quosures\n  args <- enquos(...)\n\n  # create a call object by splicing a list of quosures\n  order_call <- expr(order(!!!args, na.last = !!.na.last))\n\n  # and evaluate the constructed call in the data frame\n  ord <- eval_tidy(order_call, .df)\n\n  # sanity check\n  stopifnot(length(ord) == nrow(.df))\n\n  .df[ord, , drop = FALSE]\n}\nx <- TRUE\neval(expr(c(x = !!x)))\n#>    x \n#> TRUE\neval(expr(c(x = x)))\n#>    x \n#> TRUE"},{"path":"evaluation.html","id":"using-tidy-evaluation-exercises-20.5.4","chapter":"20 Evaluation","heading":"20.4 Using tidy evaluation (Exercises 20.5.4)","text":"Q1. ’ve included alternative implementation threshold_var() . makes different approach used ? makes harder?A1. First, let’s compare two definitions function make sure produce output:key difference subsetting operator used:old version uses non-quoting [[ operator. Thus, var argument first needs converted string.new version uses quoting $ operator. Thus, var argument first quoted unquoted (using !!).","code":"\nthreshold_var <- function(df, var, val) {\n  var <- ensym(var)\n  subset2(df, `$`(.data, !!var) >= !!val)\n}\nthreshold_var_old <- function(df, var, val) {\n  var <- as_string(ensym(var))\n  subset2(df, .data[[var]] >= !!val)\n}\n\nthreshold_var_new <- threshold_var\n\ndf <- data.frame(x = 1:10)\n\nidentical(\n  threshold_var(df, x, 8),\n  threshold_var(df, x, 8)\n)\n#> [1] TRUE"},{"path":"evaluation.html","id":"base-evaluation-exercises-20.6.3","chapter":"20 Evaluation","heading":"20.5 Base evaluation (Exercises 20.6.3)","text":"Q1. function fail?A1. doesn’t work lm_call call evaluated caller_env(), finds binding base::data() function, data execution environment.make work, need unquote data expression:Q2. model building, typically response data relatively constant rapidly experiment different predictors. Write small wrapper allows reduce duplication code .A2. small wrapper allows enter predictors:function flexible enough also allow changing data dependent variable:Q3. Another way write resample_lm() include resample expression (data[sample(nrow(data), replace = TRUE), , drop = FALSE]) data argument. Implement approach. advantages? disadvantages?A3. variant resample_lm(), providing resampled data argument.makes use R’s lazy evaluation function arguments. , resample_data argument evaluated needed function.","code":"\nlm3a <- function(formula, data) {\n  formula <- enexpr(formula)\n  lm_call <- expr(lm(!!formula, data = data))\n  eval(lm_call, caller_env())\n}\n\nlm3a(mpg ~ disp, mtcars)$call\n#> Error in as.data.frame.default(data, optional = TRUE):\n#> cannot coerce class ‘\"function\"’ to a data.frame\nlm3a <- function(formula, data) {\n  formula <- enexpr(formula)\n  lm_call <- expr(lm(!!formula, data = !!data))\n  eval(lm_call, caller_env())\n}\n\nis_call(lm3a(mpg ~ disp, mtcars)$call)\n#> [1] TRUE\nlm(mpg ~ disp, data = mtcars)\nlm(mpg ~ I(1 / disp), data = mtcars)\nlm(mpg ~ disp * cyl, data = mtcars)\nlm_custom <- function(data = mtcars, x, y = mpg) {\n  x <- enexpr(x)\n  y <- enexpr(y)\n  data <- enexpr(data)\n\n  lm_call <- expr(lm(formula = !!y ~ !!x, data = !!data))\n\n  eval(lm_call, caller_env())\n}\n\nidentical(\n  lm_custom(x = disp),\n  lm(mpg ~ disp, data = mtcars)\n)\n#> [1] TRUE\n\nidentical(\n  lm_custom(x = I(1 / disp)),\n  lm(mpg ~ I(1 / disp), data = mtcars)\n)\n#> [1] TRUE\n\nidentical(\n  lm_custom(x = disp * cyl),\n  lm(mpg ~ disp * cyl, data = mtcars)\n)\n#> [1] TRUE\nlm_custom(data = iris, x = Sepal.Length, y = Petal.Width)\n#> \n#> Call:\n#> lm(formula = Petal.Width ~ Sepal.Length, data = iris)\n#> \n#> Coefficients:\n#>  (Intercept)  Sepal.Length  \n#>      -3.2002        0.7529\nresample_lm3 <- function(formula,\n                         data,\n                         resample_data = data[sample(nrow(data), replace = TRUE), , drop = FALSE],\n                         env = current_env()) {\n  formula <- enexpr(formula)\n  lm_call <- expr(lm(!!formula, data = resample_data))\n  expr_print(lm_call)\n  eval(lm_call, env)\n}\n\ndf <- data.frame(x = 1:10, y = 5 + 3 * (1:10) + round(rnorm(10), 2))\nresample_lm3(y ~ x, data = df)\n#> lm(y ~ x, data = resample_data)\n#> \n#> Call:\n#> lm(formula = y ~ x, data = resample_data)\n#> \n#> Coefficients:\n#> (Intercept)            x  \n#>       2.654        3.420"},{"path":"evaluation.html","id":"session-information-15","chapter":"20 Evaluation","heading":"20.6 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  purrr         1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"translation.html","id":"translation","chapter":"21 Translation","heading":"21 Translation","text":"Needed libraries:","code":"\nlibrary(rlang)\nlibrary(purrr)"},{"path":"translation.html","id":"html-exercises-21.2.6","chapter":"21 Translation","heading":"21.1 HTML (Exercises 21.2.6)","text":"Q1. escaping rules <script> tags different contain JavaScript, HTML. Instead escaping angle brackets ampersands, need escape <\/script> tag isn’t closed early. example, script(\"'<\/script>'\"), shouldn’t generate :ButAdapt escape() follow rules new argument script set TRUE.A1. Let’s first start boilerplate code included book:also need tweak boilerplate pass additional parameter escape():Q2. use ... functions big downsides. ’s input validation little information documentation autocomplete used function. Create new function , given named list tags attribute names (like ), creates tag functions named arguments.tags get class id attributes.Q3. Reason following code calls with_html() referencing objects environment. work fail? ? Run code verify predictions.A3. work , first need copy-paste relevant code book:Note with_html() uses eval_tidy(), therefore code argument evaluated first html_tags named list, acts data mask, object found data mask, searches caller environment.reason, first example code work:following code, however, going work already address element data mask, p() take function address() input, escape() doesn’t know deal objects function type:Q4. Currently HTML doesn’t look terribly pretty, ’s hard see structure. adapt tag() indenting formatting? (may need research block inline tags.)A4. Let’s first look currently looks like:can improve follow Google HTML/CSS Style Guide., need create new function indent code conditionally:can update body() function use new helper:new formatting looks much better:","code":"  <script>'<\/script>'<\/script>  <script>'<\\/script>'<\/script>\nescape <- function(x, ...) UseMethod(\"escape\")\n\nescape.character <- function(x, script = FALSE) {\n  if (script) {\n    x <- gsub(\"<\/script>\", \"<\\\\/script>\", x, fixed = TRUE)\n  } else {\n    x <- gsub(\"&\", \"&amp;\", x)\n    x <- gsub(\"<\", \"&lt;\", x)\n    x <- gsub(\">\", \"&gt;\", x)\n  }\n\n  html(x)\n}\n\nescape.advr_html <- function(x, ...) x\nhtml <- function(x) structure(x, class = \"advr_html\")\n\nprint.advr_html <- function(x, ...) {\n  out <- paste0(\"<HTML> \", x)\n  cat(paste(strwrap(out), collapse = \"\\n\"), \"\\n\", sep = \"\")\n}\n\ndots_partition <- function(...) {\n  dots <- list2(...)\n\n  if (is.null(names(dots))) {\n    is_named <- rep(FALSE, length(dots))\n  } else {\n    is_named <- names(dots) != \"\"\n  }\n\n  list(\n    named = dots[is_named],\n    unnamed = dots[!is_named]\n  )\n}\n\ntag <- function(tag, script = FALSE) {\n  force(script)\n  new_function(\n    exprs(... = ),\n    expr({\n      dots <- dots_partition(...)\n      attribs <- html_attributes(dots$named)\n      children <- map_chr(.x = dots$unnamed, .f = ~ escape(.x, !!script))\n\n      html(paste0(\n        !!paste0(\"<\", tag), attribs, \">\",\n        paste(children, collapse = \"\"),\n        !!paste0(\"<\/\", tag, \">\")\n      ))\n    }),\n    caller_env()\n  )\n}\n\nvoid_tag <- function(tag) {\n  new_function(\n    exprs(... = ),\n    expr({\n      dots <- dots_partition(...)\n      if (length(dots$unnamed) > 0) {\n        abort(!!paste0(\"<\", tag, \"> must not have unnamed arguments\"))\n      }\n      attribs <- html_attributes(dots$named)\n\n      html(paste0(!!paste0(\"<\", tag), attribs, \" />\"))\n    }),\n    caller_env()\n  )\n}\n\np <- tag(\"p\")\nscript <- tag(\"script\", script = TRUE)\nscript(\"'<\/script>'\")\n#> <HTML> <script>'<\\/script>'<\/script>\nlist(\n  a = c(\"href\"),\n  img = c(\"src\", \"width\", \"height\")\n)\ngreeting <- \"Hello!\"\nwith_html(p(greeting))\np <- function() \"p\"\naddress <- \"123 anywhere street\"\nwith_html(p(address))\ntags <- c(\n  \"a\", \"abbr\", \"address\", \"article\", \"aside\", \"audio\",\n  \"b\", \"bdi\", \"bdo\", \"blockquote\", \"body\", \"button\", \"canvas\",\n  \"caption\", \"cite\", \"code\", \"colgroup\", \"data\", \"datalist\",\n  \"dd\", \"del\", \"details\", \"dfn\", \"div\", \"dl\", \"dt\", \"em\",\n  \"eventsource\", \"fieldset\", \"figcaption\", \"figure\", \"footer\",\n  \"form\", \"h1\", \"h2\", \"h3\", \"h4\", \"h5\", \"h6\", \"head\", \"header\",\n  \"hgroup\", \"html\", \"i\", \"iframe\", \"ins\", \"kbd\", \"label\",\n  \"legend\", \"li\", \"mark\", \"map\", \"menu\", \"meter\", \"nav\",\n  \"noscript\", \"object\", \"ol\", \"optgroup\", \"option\", \"output\",\n  \"p\", \"pre\", \"progress\", \"q\", \"ruby\", \"rp\", \"rt\", \"s\", \"samp\",\n  \"script\", \"section\", \"select\", \"small\", \"span\", \"strong\",\n  \"style\", \"sub\", \"summary\", \"sup\", \"table\", \"tbody\", \"td\",\n  \"textarea\", \"tfoot\", \"th\", \"thead\", \"time\", \"title\", \"tr\",\n  \"u\", \"ul\", \"var\", \"video\"\n)\n\nvoid_tags <- c(\n  \"area\", \"base\", \"br\", \"col\", \"command\", \"embed\",\n  \"hr\", \"img\", \"input\", \"keygen\", \"link\", \"meta\", \"param\",\n  \"source\", \"track\", \"wbr\"\n)\n\nhtml_tags <- c(\n  tags %>% set_names() %>% map(tag),\n  void_tags %>% set_names() %>% map(void_tag)\n)\n\nwith_html <- function(code) {\n  code <- enquo(code)\n  eval_tidy(code, html_tags)\n}\ngreeting <- \"Hello!\"\nwith_html(p(greeting))\n#> <HTML> <p>Hello!<\/p>\n\"address\" %in% names(html_tags)\n#> [1] TRUE\n\np <- function() \"p\"\naddress <- \"123 anywhere street\"\nwith_html(p(address))\n#> Error in `map_chr()`:\n#> ℹ In index: 1.\n#> Caused by error in `UseMethod()`:\n#> ! no applicable method for 'escape' applied to an object of class \"function\"\nwith_html(\n  body(\n    h1(\"A heading\", id = \"first\"),\n    p(\"Some text &\", b(\"some bold text.\")),\n    img(src = \"myimg.png\", width = 100, height = 100)\n  )\n)\n#> <HTML> <body><h1 id='first'>A heading<\/h1><p>Some\n#> text &amp;<b>some bold text.<\/b><\/p><img\n#> src='myimg.png' width='100' height='100' /><\/body>\nprint.advr_html <- function(x, ...) {\n  cat(paste(\"<HTML>\", x, sep = \"\\n\"))\n}\n\nindent <- function(x) {\n  paste0(\"  \", gsub(\"\\n\", \"\\n  \", x))\n}\n\nformat_code <- function(children, indent = FALSE) {\n  if (indent) {\n    paste0(\"\\n\", paste0(indent(children), collapse = \"\\n\"), \"\\n\")\n  } else {\n    paste(children, collapse = \"\")\n  }\n}\nhtml_tags$body <- function(...) {\n  dots <- dots_partition(...)\n  attribs <- html_attributes(dots$named)\n  children <- map_chr(dots$unnamed, escape)\n\n  html(paste0(\n    \"<body\", attribs, \">\",\n    format_code(children, indent = TRUE),\n    \"<\/body>\"\n  ))\n}\nwith_html(\n  body(\n    h1(\"A heading\", id = \"first\"),\n    p(\"Some text &\", b(\"some bold text.\")),\n    img(src = \"myimg.png\", width = 100, height = 100)\n  )\n)\n#> <HTML>\n#> <body>\n#>   <h1 id='first'>A heading<\/h1>\n#>   <p>Some text &amp;<b>some bold text.<\/b><\/p>\n#>   <img src='myimg.png' width='100' height='100' />\n#> <\/body>"},{"path":"translation.html","id":"latex-exercises-21.3.8","chapter":"21 Translation","heading":"21.2 LaTeX (Exercises 21.3.8)","text":"didn’t manage solve exercises, ’d recommend checking solutions official solutions manual.Q1. Add escaping. special symbols escaped adding backslash front \\, $, %. Just HTML, ’ll need make sure don’t end double-escaping. ’ll need create small S3 class use function operators. also allow embed arbitrary LaTeX needed.Q2. Complete DSL support functions plotmath supports.","code":""},{"path":"translation.html","id":"session-information-16","chapter":"21 Translation","heading":"21.3 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  purrr       * 1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"debugging.html","id":"debugging","chapter":"22 Debugging","heading":"22 Debugging","text":"exercises.","code":""},{"path":"measuring-performance.html","id":"measuring-performance","chapter":"23 Measuring performance","heading":"23 Measuring performance","text":"Attaching needed libraries:","code":"\nlibrary(profvis, warn.conflicts = FALSE)\nlibrary(dplyr, warn.conflicts = FALSE)"},{"path":"measuring-performance.html","id":"profiling-exercises-23.2.4","chapter":"23 Measuring performance","heading":"23.1 Profiling (Exercises 23.2.4)","text":"Q1. Profile following function torture = TRUE. surprising? Read source code rm() figure ’s going .A1. Let’s source functions mentioned exercises.First, try without torture = TRUE: returns meaningful results.mentioned docs, setting torture = TRUETriggers garbage collection every torture memory allocation call.process somehow never seems finish crashes RStudio session stops!question says documentation rm() may provide clues:still couldn’t figure . recommend checking official answer.","code":"\nf <- function(n = 1e5) {\n  x <- rep(1, n)\n  rm(x)\n}\nsource(\"profiling-exercises.R\")\nprofvis(f())\n#> Error in parse_rprof_lines(lines, expr_source): No parsing data available. Maybe your function was too fast?\nprofvis(f(), torture = TRUE)\nrm\n#> function (..., list = character(), pos = -1, envir = as.environment(pos), \n#>     inherits = FALSE) \n#> {\n#>     if (...length()) {\n#>         dots <- match.call(expand.dots = FALSE)$...\n#>         if (!all(vapply(dots, function(x) is.symbol(x) || is.character(x), \n#>             NA, USE.NAMES = FALSE))) \n#>             stop(\"... must contain names or character strings\")\n#>         list <- .Primitive(\"c\")(list, vapply(dots, as.character, \n#>             \"\"))\n#>     }\n#>     .Internal(remove(list, envir, inherits))\n#> }\n#> <bytecode: 0x56294c35adf0>\n#> <environment: namespace:base>"},{"path":"measuring-performance.html","id":"microbenchmarking-exercises-23.3.3","chapter":"23 Measuring performance","heading":"23.2 Microbenchmarking (Exercises 23.3.3)","text":"Q1. Instead using bench::mark(), use built-function system.time(). system.time() much less precise, ’ll need repeat operation many times loop, divide find average time operation, code .estimates system.time() compare bench::mark()? different?A1. Let’s benchmark first using two approaches:Now can compare results alternatives:comparison reveals two approaches yield quite similar results. Slight differences exact values possibly due differences precision timers used internally functions.Q2. two ways compute square root vector. think fastest? slowest? Use microbenchmarking test answers.A2. Microbenchmarking ways compute square root vector mentioned chapter.specialized primitive function sqrt() (written C) fastest way compute square root.","code":"\nn <- 1e6\nsystem.time(for (i in 1:n) sqrt(x)) / n\nsystem.time(for (i in 1:n) x^0.5) / n\nn <- 1e6\nx <- runif(100)\n\n# bench -------------------\n\nbench_df <- bench::mark(\n  sqrt(x),\n  x^0.5,\n  iterations = n,\n  time_unit = \"us\"\n)\n\nt_bench_df <- bench_df %>%\n  select(expression, time) %>%\n  rowwise() %>%\n  mutate(bench_mean = mean(unlist(time))) %>%\n  ungroup() %>%\n  select(-time)\n\n# system.time -------------------\n\n# garbage collection performed immediately before the timing\nt1_systime_gc <- system.time(for (i in 1:n) sqrt(x), gcFirst = TRUE) / n\nt2_systime_gc <- system.time(for (i in 1:n) x^0.5, gcFirst = TRUE) / n\n\n# garbage collection not performed immediately before the timing\nt1_systime_nogc <- system.time(for (i in 1:n) sqrt(x), gcFirst = FALSE) / n\nt2_systime_nogc <- system.time(for (i in 1:n) x^0.5, gcFirst = FALSE) / n\n\nt_systime_df <- tibble(\n  \"expression\" = bench_df$expression,\n  \"systime_with_gc\" = c(t1_systime_gc[\"elapsed\"], t2_systime_gc[\"elapsed\"]),\n  \"systime_with_nogc\" = c(t1_systime_nogc[\"elapsed\"], t2_systime_nogc[\"elapsed\"])\n) %>%\n  mutate(\n    systime_with_gc = systime_with_gc * 1e6, # in microseconds\n    systime_with_nogc = systime_with_nogc * 1e6 # in microseconds\n  )\n# note that system time columns report time in microseconds\nfull_join(t_bench_df, t_systime_df, by = \"expression\")\n#> # A tibble: 2 × 4\n#>   expression bench_mean systime_with_gc systime_with_nogc\n#>   <bch:expr>   <bch:tm>           <dbl>             <dbl>\n#> 1 sqrt(x)      670.07ns           0.629             0.443\n#> 2 x^0.5          2.19µs           2.01              2.02\nx^(1 / 2)\nexp(log(x) / 2)\nx <- runif(1000)\n\nbench::mark(\n  sqrt(x),\n  x^0.5,\n  x^(1 / 2),\n  exp(log(x) / 2),\n  iterations = 1000\n) %>%\n  select(expression, median) %>%\n  arrange(median)\n#> # A tibble: 4 × 2\n#>   expression      median\n#>   <bch:expr>    <bch:tm>\n#> 1 sqrt(x)         3.03µs\n#> 2 exp(log(x)/2)  12.62µs\n#> 3 x^0.5          18.86µs\n#> 4 x^(1/2)           19µs"},{"path":"measuring-performance.html","id":"session-information-17","chapter":"23 Measuring performance","heading":"23.3 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bench         1.1.3   2023-05-04 [1] RSPM\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr       * 1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  htmlwidgets   1.6.4   2023-12-06 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  profmem       0.6.0   2020-12-13 [1] RSPM\n#>  profvis     * 0.3.8   2023-05-02 [1] RSPM\n#>  purrr         1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  stringi       1.8.3   2023-12-11 [1] RSPM\n#>  stringr       1.5.1   2023-11-14 [1] RSPM\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"improving-performance.html","id":"improving-performance","chapter":"24 Improving performance","heading":"24 Improving performance","text":"Attaching needed libraries:","code":"\nlibrary(ggplot2)\nlibrary(dplyr)\nlibrary(purrr)"},{"path":"improving-performance.html","id":"exercises-24.3.1","chapter":"24 Improving performance","heading":"24.1 Exercises 24.3.1","text":"Q1. faster alternatives lm()? specifically designed work larger datasets?A1. Faster alternatives lm() can found visiting CRAN Task View: High-Performance Parallel Computing R page.available options:speedglm::speedlm() (large datasets)speedglm::speedlm() (large datasets)biglm::biglm() (specifically designed data large fit memory)biglm::biglm() (specifically designed data large fit memory)RcppEigen::fastLm() (using Eigen linear algebra library)RcppEigen::fastLm() (using Eigen linear algebra library)High performances can obtained packages especially R linked optimized BLAS, ATLAS. can check information using sessionInfo():Comparing performance different alternatives:results might change depending size dataset, performance benefits accruing bigger dataset.experiment different algorithms find one fits needs dataset best.Q2. package implements version match() ’s faster repeated look ups? much faster ?A2. package (respective function) fastmatch::fmatch()7.documentation function notes:slightly faster built-version uses specialized code, addition retains hash table within table object can re-used, dramatically reducing look-time especially large table.small vector, fmatch() slightly faster, order magnitude., larger vector, fmatch() orders magnitude faster! ⚡can also look hash table:Additionally, fastmatch provides equivalent familiar infix operator:Q3. List four functions (just base R) convert string date time object. strengths weaknesses?A3. four functions convert string date time object:base::.POSIXct()base::.POSIXlt()lubridate::ymd_hms()fasttime::fastPOSIXct()can also compare performance:many packages implement way convert string date time object. , see CRAN Task View: Time Series AnalysisQ4. packages provide ability compute rolling mean?A4. packages respective functions provide way compute rolling mean:RcppRoll::roll_mean()data.table::frollmean()roll::roll_mean()zoo::rollmean()slider::slide_dbl()Q5. alternatives optim()?A5. optim() function provides general-purpose optimization. noted docs:General-purpose optimization based Nelder–Mead, quasi-Newton conjugate-gradient algorithms. includes option box-constrained optimization simulated annealing.many alternatives exact one want choose depend type optimization like .available options can seen CRAN Task View: Optimization Mathematical Programming.","code":"\nsessInfo <- sessionInfo()\nsessInfo$matprod\n#> [1] \"default\"\nsessInfo$LAPACK\n#> [1] \"/usr/lib/x86_64-linux-gnu/openblas-pthread/libopenblasp-r0.3.20.so\"\nlibrary(gapminder)\n\n# having a look at the data\nglimpse(gapminder)\n#> Rows: 1,704\n#> Columns: 6\n#> $ country   <fct> \"Afghanistan\", \"Afghanistan\", \"Afghanist…\n#> $ continent <fct> Asia, Asia, Asia, Asia, Asia, Asia, Asia…\n#> $ year      <int> 1952, 1957, 1962, 1967, 1972, 1977, 1982…\n#> $ lifeExp   <dbl> 28.801, 30.332, 31.997, 34.020, 36.088, …\n#> $ pop       <int> 8425333, 9240934, 10267083, 11537966, 13…\n#> $ gdpPercap <dbl> 779.4453, 820.8530, 853.1007, 836.1971, …\n\nbench::mark(\n  \"lm\"       = stats::lm(lifeExp ~ continent * gdpPercap, gapminder),\n  \"speedglm\" = speedglm::speedlm(lifeExp ~ continent * gdpPercap, gapminder),\n  \"biglm\"    = biglm::biglm(lifeExp ~ continent * gdpPercap, gapminder),\n  \"fastLm\"   = RcppEigen::fastLm(lifeExp ~ continent * gdpPercap, gapminder),\n  check      = FALSE,\n  iterations = 1000\n)[1:5]\n#> # A tibble: 4 × 5\n#>   expression      min   median `itr/sec` mem_alloc\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 lm         860.49µs 897.15µs     1094.    1.27MB\n#> 2 speedglm     1.56ms   1.63ms      611.   75.01MB\n#> 3 biglm      767.92µs 797.64µs     1189.   589.1KB\n#> 4 fastLm     987.36µs   1.01ms      982.    4.48MB\nlibrary(fastmatch, warn.conflicts = FALSE)\n\nsmall_vec <- c(\"a\", \"b\", \"x\", \"m\", \"n\", \"y\")\n\nlength(small_vec)\n#> [1] 6\n\nbench::mark(\n  \"base\" = match(c(\"x\", \"y\"), small_vec),\n  \"fastmatch\" = fmatch(c(\"x\", \"y\"), small_vec)\n)[1:5]\n#> # A tibble: 2 × 5\n#>   expression      min   median `itr/sec` mem_alloc\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 base         1.14µs   1.18µs   796519.     2.8KB\n#> 2 fastmatch  990.93ns   1.04µs   929883.    2.66KB\nlarge_vec <- c(rep(c(\"a\", \"b\"), 1e4), \"x\", rep(c(\"m\", \"n\"), 1e6), \"y\")\n\nlength(large_vec)\n#> [1] 2020002\n\nbench::mark(\n  \"base\" = match(c(\"x\", \"y\"), large_vec),\n  \"fastmatch\" = fmatch(c(\"x\", \"y\"), large_vec)\n)[1:5]\n#> # A tibble: 2 × 5\n#>   expression      min   median `itr/sec` mem_alloc\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 base         23.3ms   23.4ms      42.7    31.4MB\n#> 2 fastmatch   990.9ns   1.04µs  917170.         0B\nfmatch.hash(c(\"x\", \"y\"), small_vec)\n#> [1] \"a\" \"b\" \"x\" \"m\" \"n\" \"y\"\n#> attr(,\".match.hash\")\n#> <hash table>\nlibrary(fastmatch)\n\nsmall_vec <- c(\"a\", \"b\", \"x\", \"m\", \"n\", \"y\")\n\nc(\"x\", \"y\") %in% small_vec\n#> [1] TRUE TRUE\n\nc(\"x\", \"y\") %fin% small_vec\n#> [1] TRUE TRUE\nbase::as.POSIXct(\"2022-05-05 09:23:22\")\n#> [1] \"2022-05-05 09:23:22 UTC\"\nbase::as.POSIXlt(\"2022-05-05 09:23:22\")\n#> [1] \"2022-05-05 09:23:22 UTC\"\nlubridate::ymd_hms(\"2022-05-05-09-23-22\")\n#> [1] \"2022-05-05 09:23:22 UTC\"\nfasttime::fastPOSIXct(\"2022-05-05 09:23:22\")\n#> [1] \"2022-05-05 09:23:22 UTC\"\nbench::mark(\n  \"as.POSIXct\" = base::as.POSIXct(\"2022-05-05 09:23:22\"),\n  \"as.POSIXlt\" = base::as.POSIXlt(\"2022-05-05 09:23:22\"),\n  \"ymd_hms\" = lubridate::ymd_hms(\"2022-05-05-09-23-22\"),\n  \"fastPOSIXct\" = fasttime::fastPOSIXct(\"2022-05-05 09:23:22\"),\n  check = FALSE,\n  iterations = 1000\n)\n#> # A tibble: 4 × 6\n#>   expression       min   median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1 as.POSIXct   29.97µs   31.6µs    31256.        0B     0   \n#> 2 as.POSIXlt   20.65µs  21.83µs    45281.        0B     0   \n#> 3 ymd_hms       2.22ms   2.29ms      432.    21.5KB     3.48\n#> 4 fastPOSIXct   1.17µs   1.22µs   779859.        0B     0"},{"path":"improving-performance.html","id":"exercises-24.4.3","chapter":"24 Improving performance","heading":"24.2 Exercises 24.4.3","text":"Q1. ’s difference rowSums() .rowSums()?A1. documentation functions state:versions initial dot name (.colSums() etc) ‘bare-bones’ versions use programming: apply numeric (like) matrices name result.Looking source code,rowSums() function number checks validate arguments acceptable.rowSums() directly proceeds computation using internal code built R interpreterBut comparable performance:Q2. Make faster version chisq.test() computes chi-square test statistic input two numeric vectors missing values. can try simplifying chisq.test() coding mathematical definition.A2. function supposed accept two numeric vectors without missing values, can make chisq.test() less work removing code corresponding following :checks data frame matrix inputsgoodness--fit testsimulating p-valueschecking missing valuesThis leaves us much simpler, bare bones implementation:, indeed, custom function performs slightly better8 base equivalent:Q3. Can make faster version table() case input two integer vectors missing values? Can use speed chi-square test?A3. order make leaner version table(), can take similar approach trim unnecessary input checks light new API accepting just two vectors without missing values. can remove following components code:extracting data objects entered ... argumentdealing missing valuesother input validation checksIn addition removal, can also use fastmatch::fmatch() instead match():custom function indeed performs slightly better:can also use function custom chi-squared test function see performance improves :, indeed, new version custom function performs even better previously :","code":"\nrowSums\n#> function (x, na.rm = FALSE, dims = 1L) \n#> {\n#>     if (is.data.frame(x)) \n#>         x <- as.matrix(x)\n#>     if (!is.array(x) || length(dn <- dim(x)) < 2L) \n#>         stop(\"'x' must be an array of at least two dimensions\")\n#>     if (dims < 1L || dims > length(dn) - 1L) \n#>         stop(\"invalid 'dims'\")\n#>     p <- prod(dn[-(id <- seq_len(dims))])\n#>     dn <- dn[id]\n#>     z <- if (is.complex(x)) \n#>         .Internal(rowSums(Re(x), prod(dn), p, na.rm)) + (0+1i) * \n#>             .Internal(rowSums(Im(x), prod(dn), p, na.rm))\n#>     else .Internal(rowSums(x, prod(dn), p, na.rm))\n#>     if (length(dn) > 1L) {\n#>         dim(z) <- dn\n#>         dimnames(z) <- dimnames(x)[id]\n#>     }\n#>     else names(z) <- dimnames(x)[[1L]]\n#>     z\n#> }\n#> <bytecode: 0x55a2a63ce4b0>\n#> <environment: namespace:base>\n.rowSums\n#> function (x, m, n, na.rm = FALSE) \n#> .Internal(rowSums(x, m, n, na.rm))\n#> <bytecode: 0x55a2ac309f60>\n#> <environment: namespace:base>\nx <- cbind(x1 = 3, x2 = c(4:1e4, 2:1e5))\n\nbench::mark(\n  \"rowSums\" = rowSums(x),\n  \".rowSums\" = .rowSums(x, dim(x)[[1]], dim(x)[[2]])\n)[1:5]\n#> # A tibble: 2 × 5\n#>   expression      min   median `itr/sec` mem_alloc\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 rowSums       821µs   1.29ms      761.     859KB\n#> 2 .rowSums      818µs   1.28ms      837.     859KB\nmy_chisq_test <- function(x, y) {\n  x <- table(x, y)\n  n <- sum(x)\n\n  nr <- as.integer(nrow(x))\n  nc <- as.integer(ncol(x))\n\n  sr <- rowSums(x)\n  sc <- colSums(x)\n  E <- outer(sr, sc, \"*\") / n\n  v <- function(r, c, n) c * r * (n - r) * (n - c) / n^3\n  V <- outer(sr, sc, v, n)\n  dimnames(E) <- dimnames(x)\n\n  STATISTIC <- sum((abs(x - E))^2 / E)\n  PARAMETER <- (nr - 1L) * (nc - 1L)\n  PVAL <- pchisq(STATISTIC, PARAMETER, lower.tail = FALSE)\n\n  names(STATISTIC) <- \"X-squared\"\n  names(PARAMETER) <- \"df\"\n\n  structure(\n    list(\n      statistic = STATISTIC,\n      parameter = PARAMETER,\n      p.value = PVAL,\n      method = \"Pearson's Chi-squared test\",\n      observed = x,\n      expected = E,\n      residuals = (x - E) / sqrt(E),\n      stdres = (x - E) / sqrt(V)\n    ),\n    class = \"htest\"\n  )\n}\nm <- c(rep(\"a\", 1000), rep(\"b\", 9000))\nn <- c(rep(c(\"x\", \"y\"), 5000))\n\nbench::mark(\n  \"base\" = chisq.test(m, n)$statistic[[1]],\n  \"custom\" = my_chisq_test(m, n)$statistic[[1]]\n)[1:5]\n#> # A tibble: 2 × 5\n#>   expression      min   median `itr/sec` mem_alloc\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 base          904µs    927µs     1008.    1.57MB\n#> 2 custom        708µs    721µs     1376.    1.13MB\nmy_table <- function(x, y) {\n  x_sorted <- sort(unique(x))\n  y_sorted <- sort(unique(y))\n\n  x_length <- length(x_sorted)\n  y_length <- length(y_sorted)\n\n  bin <-\n    fastmatch::fmatch(x, x_sorted) +\n    x_length * fastmatch::fmatch(y, y_sorted) -\n    x_length\n\n  y <- tabulate(bin, x_length * y_length)\n\n  y <- array(\n    y,\n    dim = c(x_length, y_length),\n    dimnames = list(x = x_sorted, y = y_sorted)\n  )\n\n  class(y) <- \"table\"\n  y\n}\nx <- c(rep(\"a\", 1000), rep(\"b\", 9000))\ny <- c(rep(c(\"x\", \"y\"), 5000))\n\n# `check = FALSE` because the custom function has an additional attribute:\n# \".match.hash\"\nbench::mark(\n  \"base\" = table(x, y),\n  \"custom\" = my_table(x, y),\n  check = FALSE\n)[1:5]\n#> # A tibble: 2 × 5\n#>   expression      min   median `itr/sec` mem_alloc\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 base          638µs    652µs     1519.     960KB\n#> 2 custom        355µs    362µs     2736.     488KB\nmy_chisq_test2 <- function(x, y) {\n  x <- my_table(x, y)\n  n <- sum(x)\n\n  nr <- as.integer(nrow(x))\n  nc <- as.integer(ncol(x))\n\n  sr <- rowSums(x)\n  sc <- colSums(x)\n  E <- outer(sr, sc, \"*\") / n\n  v <- function(r, c, n) c * r * (n - r) * (n - c) / n^3\n  V <- outer(sr, sc, v, n)\n  dimnames(E) <- dimnames(x)\n\n  STATISTIC <- sum((abs(x - E))^2 / E)\n  PARAMETER <- (nr - 1L) * (nc - 1L)\n  PVAL <- pchisq(STATISTIC, PARAMETER, lower.tail = FALSE)\n\n  names(STATISTIC) <- \"X-squared\"\n  names(PARAMETER) <- \"df\"\n\n  structure(\n    list(\n      statistic = STATISTIC,\n      parameter = PARAMETER,\n      p.value = PVAL,\n      method = \"Pearson's Chi-squared test\",\n      observed = x,\n      expected = E,\n      residuals = (x - E) / sqrt(E),\n      stdres = (x - E) / sqrt(V)\n    ),\n    class = \"htest\"\n  )\n}\nm <- c(rep(\"a\", 1000), rep(\"b\", 9000))\nn <- c(rep(c(\"x\", \"y\"), 5000))\n\nbench::mark(\n  \"base\" = chisq.test(m, n)$statistic[[1]],\n  \"custom\" = my_chisq_test2(m, n)$statistic[[1]]\n)[1:5]\n#> # A tibble: 2 × 5\n#>   expression      min   median `itr/sec` mem_alloc\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 base          909µs    940µs     1059.    1.28MB\n#> 2 custom        418µs    427µs     2314.  594.45KB"},{"path":"improving-performance.html","id":"exercises-24.5.1","chapter":"24 Improving performance","heading":"24.3 Exercises 24.5.1","text":"Q1. density functions, e.g., dnorm(), common interface. arguments vectorised ? rnorm(10, mean = 10:1) ?A1. density function family following interface:Reading documentation reveals following parameters vectorized:\nx, q, p, mean, sd.means something like following work:, functions don’t multiple vectorized parameters, won’t. example,following function call generates 10 random numbers (since n = 10) 10 different distributions means supplied vector 10:1.Q2. Compare speed apply(x, 1, sum) rowSums(x) varying sizes x.A2. can write custom function vary number rows matrix extract data frame comparing performance two functions.Plotting data reveals rowSums(x) O(1) behavior, O(n) behavior.Q3. can use crossprod() compute weighted sum? much faster naive sum(x * w)?A3. functions provide way compute weighted sum:benchmarking performance reveals latter significantly faster former!","code":"\ndnorm(x, mean = 0, sd = 1, log = FALSE)\npnorm(q, mean = 0, sd = 1, lower.tail = TRUE, log.p = FALSE)\nqnorm(p, mean = 0, sd = 1, lower.tail = TRUE, log.p = FALSE)\nrnorm(n, mean = 0, sd = 1)\nrnorm(c(1, 2, 3), mean = c(0, -1, 5))\n#> [1] 1.124335 0.930398 3.844935\npnorm(c(1, 2, 3), mean = c(0, -1, 5), log.p = c(FALSE, TRUE, TRUE))\n#> [1] 0.84134475 0.99865010 0.02275013\nrnorm(n = 10, mean = 10:1)\n#>  [1]  8.2421770  9.3920474  7.1362118  7.5789906  5.2551688\n#>  [6]  6.0143714  4.6147891  1.1096247  2.8759129 -0.6756857\nbenc_perform <- function(nRow, nCol = 100) {\n  x <- matrix(data = rnorm(nRow * nCol), nrow = nRow, ncol = nCol)\n\n  bench::mark(\n    rowSums(x),\n    apply(x, 1, sum)\n  )[1:5]\n}\n\nnRowList <- list(10, 100, 500, 1000, 5000, 10000, 50000, 100000)\n\nnames(nRowList) <- as.character(nRowList)\n\nbenchDF <- map_dfr(\n  .x = nRowList,\n  .f = ~ benc_perform(.x),\n  .id = \"nRows\"\n) %>%\n  mutate(nRows = as.numeric(nRows))\nggplot(\n  benchDF,\n  aes(\n    x = as.numeric(nRows),\n    y = median,\n    group = as.character(expression),\n    color = as.character(expression)\n  )\n) +\n  geom_point() +\n  geom_line() +\n  labs(\n    x = \"Number of Rows\",\n    y = \"Median Execution Time\",\n    colour = \"Function used\"\n  )\nx <- c(1:6, 2, 3)\nw <- rnorm(length(x))\n\ncrossprod(x, w)[[1]]\n#> [1] 15.94691\nsum(x * w)[[1]]\n#> [1] 15.94691\nbench::mark(\n  crossprod(x, w)[[1]],\n  sum(x * w)[[1]],\n  iterations = 1e6\n)[1:5]\n#> # A tibble: 2 × 5\n#>   expression                min   median `itr/sec` mem_alloc\n#>   <bch:expr>           <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 crossprod(x, w)[[1]]    871ns    922ns  1001700.        0B\n#> 2 sum(x * w)[[1]]         460ns    511ns  1796479.        0B"},{"path":"rewriting-r-code-in-c.html","id":"rewriting-r-code-in-c","chapter":"25 Rewriting R code in C++","heading":"25 Rewriting R code in C++","text":"","code":"\nlibrary(Rcpp, warn.conflicts = FALSE)"},{"path":"rewriting-r-code-in-c.html","id":"getting-started-with-c-exercises-25.2.6","chapter":"25 Rewriting R code in C++","heading":"25.1 Getting started with C++ (Exercises 25.2.6)","text":"Q1. basics C++ hand, ’s now great time practice reading writing simple C++ functions. following functions, read code figure corresponding base R function . might understand every part code yet, able figure basics function .A1.f1() mean():f2() cumsum():f3() ():f4() Position():f5() pmin():Q2. practice function writing skills, convert following functions C++. now, assume inputs missing values.().().cumprod(), cummin(), cummax().cumprod(), cummin(), cummax().diff(). Start assuming lag 1, generalise lag n.diff(). Start assuming lag 1, generalise lag n.range().range().var(). Read approaches can take Wikipedia. Whenever implementing numerical algorithm, ’s always good check already known problem.var(). Read approaches can take Wikipedia. Whenever implementing numerical algorithm, ’s always good check already known problem.A2. performance benefits going observed function primitive since already tuned max R performance. , expect performance gain diff() var().()cumprod()cumminC()cummaxC()diff()range()var()","code":"#include <Rcpp.h>\nusing namespace Rcpp;\n\n// [[Rcpp::export]]\ndouble f1(NumericVector x) {\n  int n = x.size();\n  double y = 0;\n\n  for(int i = 0; i < n; ++i) {\n    y += x[i] / n;\n  }\n  return y;\n}\n\n// [[Rcpp::export]]\nNumericVector f2(NumericVector x) {\n  int n = x.size();\n  NumericVector out(n);\n\n  out[0] = x[0];\n  for(int i = 1; i < n; ++i) {\n    out[i] = out[i - 1] + x[i];\n  }\n  return out;\n}\n\n// [[Rcpp::export]]\nbool f3(LogicalVector x) {\n  int n = x.size();\n\n  for(int i = 0; i < n; ++i) {\n    if (x[i]) return true;\n  }\n  return false;\n}\n\n// [[Rcpp::export]]\nint f4(Function pred, List x) {\n  int n = x.size();\n\n  for(int i = 0; i < n; ++i) {\n    LogicalVector res = pred(x[i]);\n    if (res[0]) return i + 1;\n  }\n  return 0;\n}\n\n// [[Rcpp::export]]\nNumericVector f5(NumericVector x, NumericVector y) {\n  int n = std::max(x.size(), y.size());\n  NumericVector x1 = rep_len(x, n);\n  NumericVector y1 = rep_len(y, n);\n\n  NumericVector out(n);\n\n  for (int i = 0; i < n; ++i) {\n    out[i] = std::min(x1[i], y1[i]);\n  }\n\n  return out;\n}\nx <- c(1, 2, 3, 4, 5, 6)\n\nf1(x)\n#> [1] 3.5\nmean(x)\n#> [1] 3.5\nx <- c(1, 3, 5, 6)\n\nf2(x)\n#> [1]  1  4  9 15\ncumsum(x)\n#> [1]  1  4  9 15\nx1 <- c(TRUE, FALSE, FALSE, TRUE)\nx2 <- c(FALSE, FALSE)\n\nf3(x1)\n#> [1] TRUE\nany(x1)\n#> [1] TRUE\n\nf3(x2)\n#> [1] FALSE\nany(x2)\n#> [1] FALSE\nx <- list(\"a\", TRUE, \"m\", 2)\n\nf4(is.numeric, x)\n#> [1] 4\nPosition(is.numeric, x)\n#> [1] 4\nv1 <- c(1, 3, 4, 5, 6, 7)\nv2 <- c(1, 2, 7, 2, 8, 1)\n\nf5(v1, v2)\n#> [1] 1 2 4 2 6 1\npmin(v1, v2)\n#> [1] 1 2 4 2 6 1\nis.primitive(all)\n#> [1] TRUE\nis.primitive(cumprod)\n#> [1] TRUE\nis.primitive(diff)\n#> [1] FALSE\nis.primitive(range)\n#> [1] TRUE\nis.primitive(var)\n#> [1] FALSE#include <vector>\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nbool allC(std::vector<bool> x)\n{\n    for (const auto& xElement : x)\n    {\n        if (!xElement) return false;\n    }\n\n    return true;\n}\nv1 <- rep(TRUE, 10)\nv2 <- c(rep(TRUE, 5), rep(FALSE, 5))\n\nall(v1)\n#> [1] TRUE\nallC(v1)\n#> [1] TRUE\n\nall(v2)\n#> [1] FALSE\nallC(v2)\n#> [1] FALSE\n\n# performance benefits?\nbench::mark(\n  all(c(rep(TRUE, 1000), rep(FALSE, 1000))),\n  allC(c(rep(TRUE, 1000), rep(FALSE, 1000))),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression                                      min\n#>   <bch:expr>                                 <bch:tm>\n#> 1 all(c(rep(TRUE, 1000), rep(FALSE, 1000)))    6.16µs\n#> 2 allC(c(rep(TRUE, 1000), rep(FALSE, 1000)))   8.59µs\n#>     median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1   6.46µs   152676.    15.8KB        0\n#> 2   8.95µs   109377.    18.3KB        0#include <vector>\n\n// [[Rcpp::export]]\nstd::vector<double> cumprodC(const std::vector<double> &x)\n{\n    std::vector<double> out{x};\n\n    for (std::size_t i = 1; i < x.size(); i++)\n    {\n        out[i] = out[i - 1] * x[i];\n    }\n\n    return out;\n}\nv1 <- c(10, 4, 6, 8)\n\ncumprod(v1)\n#> [1]   10   40  240 1920\ncumprodC(v1)\n#> [1]   10   40  240 1920\n\n# performance benefits?\nbench::mark(\n  cumprod(v1),\n  cumprodC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression        min   median `itr/sec` mem_alloc\n#>   <bch:expr>   <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 cumprod(v1)  110.01ns 130.04ns  6991133.        0B\n#> 2 cumprodC(v1)   1.44µs   1.53µs   619348.    6.62KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0#include <vector>\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::vector<double> cumminC(const std::vector<double> &x)\n{\n    std::vector<double> out{x};\n\n    for (std::size_t i = 1; i < x.size(); i++)\n    {\n        out[i] = (out[i] < out[i - 1]) ? out[i] : out[i - 1];\n    }\n\n    return out;\n}\nv1 <- c(3:1, 2:0, 4:2)\n\ncummin(v1)\n#> [1] 3 2 1 1 1 0 0 0 0\ncumminC(v1)\n#> [1] 3 2 1 1 1 0 0 0 0\n\n# performance benefits?\nbench::mark(\n  cummin(v1),\n  cumminC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression       min   median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1 cummin(v1)  130.97ns 151.57ns  5389797.        0B        0\n#> 2 cumminC(v1)   1.52µs   1.83µs   514890.    6.62KB        0#include <vector>\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::vector<double> cummaxC(const std::vector<double> &x)\n{\n    std::vector<double> out{x};\n\n    for (std::size_t i = 1; i < x.size(); i++)\n    {\n        out[i] = (out[i] > out[i - 1]) ? out[i] : out[i - 1];\n    }\n    \n    return out;\n}\nv1 <- c(3:1, 2:0, 4:2)\n\ncummax(v1)\n#> [1] 3 3 3 3 3 3 4 4 4\ncummaxC(v1)\n#> [1] 3 3 3 3 3 3 4 4 4\n\n# performance benefits?\nbench::mark(\n  cummax(v1),\n  cummaxC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression       min   median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1 cummax(v1)  110.01ns 120.14ns  6760800.        0B        0\n#> 2 cummaxC(v1)   1.55µs   1.72µs   549982.    6.62KB        0#include <vector>\n#include <functional>\n#include <algorithm>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::vector<double> diffC(const std::vector<double> &x, int lag)\n{\n    std::vector<double> vec_start;\n    std::vector<double> vec_lagged;\n    std::vector<double> vec_diff;\n\n    for (std::size_t i = lag; i < x.size(); i++)\n    {\n        vec_lagged.push_back(x[i]);\n    }\n\n    for (std::size_t i = 0; i < (x.size() - lag); i++)\n    {\n        vec_start.push_back(x[i]);\n    }\n\n    std::transform(\n        vec_lagged.begin(), vec_lagged.end(),\n        vec_start.begin(), std::back_inserter(vec_diff),\n        std::minus<double>());\n\n    return vec_diff;\n}\nv1 <- c(1, 2, 4, 8, 13)\nv2 <- c(1, 2, NA, 8, 13)\n\ndiff(v1, 2)\n#> [1] 3 6 9\ndiffC(v1, 2)\n#> [1] 3 6 9\n\ndiff(v2, 2)\n#> [1] NA  6 NA\ndiffC(v2, 2)\n#> [1] NA  6 NA\n\n# performance benefits?\nbench::mark(\n  diff(v1, 2),\n  diffC(v1, 2),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression        min   median `itr/sec` mem_alloc\n#>   <bch:expr>   <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 diff(v1, 2)    3.78µs   4.44µs   218299.        0B\n#> 2 diffC(v1, 2)   1.91µs   2.12µs   450551.    2.49KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0#include <iostream>\n#include <vector>\n#include <algorithm>\nusing namespace std;\n\n// [[Rcpp::export]]\nstd::vector<double> rangeC(std::vector<double> x)\n{\n    std::vector<double> rangeVec{0.0, 0.0};\n\n    rangeVec.at(0) = *std::min_element(x.begin(), x.end());\n    rangeVec.at(1) = *std::max_element(x.begin(), x.end());\n\n    return rangeVec;\n}\nv1 <- c(10, 4, 6, 8)\n\nrange(v1)\n#> [1]  4 10\nrangeC(v1)\n#> [1]  4 10\n\n# performance benefits?\nbench::mark(\n  range(v1),\n  rangeC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression      min   median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1 range(v1)    1.64µs   1.81µs   526003.        0B        0\n#> 2 rangeC(v1)   1.51µs   1.66µs   482154.    6.62KB        0#include <vector>\n#include <cmath>\n#include <numeric>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\ndouble variance(std::vector<double> x)\n{\n    double sumSquared{0};\n\n    double mean = std::accumulate(x.begin(), x.end(), 0.0) / x.size();\n\n    for (const auto& xElement : x)\n    {\n        sumSquared += pow(xElement - mean, 2.0);\n    }\n\n    return sumSquared / (x.size() - 1);\n}\nv1 <- c(1, 4, 7, 8)\n\nvar(v1)\n#> [1] 10\nvariance(v1)\n#> [1] 10\n\n# performance benefits?\nbench::mark(\n  var(v1),\n  variance(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression        min   median `itr/sec` mem_alloc\n#>   <bch:expr>   <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 var(v1)         7.4µs   8.31µs   117887.        0B\n#> 2 variance(v1)   1.46µs   1.61µs   587188.    6.62KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0"},{"path":"rewriting-r-code-in-c.html","id":"missing-values-exercises-25.4.5","chapter":"25 Rewriting R code in C++","heading":"25.2 Missing values (Exercises 25.4.5)","text":"Q1. Rewrite functions Exercise 25.2.6 deal missing values. na.rm true, ignore missing values. na.rm false, return missing value input contains missing values. good functions practice min(), max(), range(), mean(), var().A1. create version range() deals missing values. principle applies others:Q2. Rewrite cumsum() diff() can handle missing values. Note functions slightly complicated behaviour.A2. cumsum() docs say:NA value x causes corresponding following elements return value NA, integer overflow cumsum (warning).Similarly, diff() docs say:NA’s propagate.Therefore, functions don’t allow removing missing values NAs propagate.seen examples , diffC() already behaves way.Similarly, cumsumC() propagates NAs well.","code":"#include <iostream>\n#include <vector>\n#include <algorithm>\n#include <math.h>\n#include <Rcpp.h>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::vector<double> rangeC_NA(std::vector<double> x, bool removeNA = true)\n{\n    std::vector<double> rangeVec{0.0, 0.0};\n\n    bool naPresent = std::any_of(\n        x.begin(),\n        x.end(),\n        [](double d)\n        { return isnan(d); });\n\n    if (naPresent)\n    {\n        if (removeNA)\n        {\n            std::remove(x.begin(), x.end(), NAN);\n        }\n        else\n        {\n            rangeVec.at(0) = NA_REAL; // NAN;\n            rangeVec.at(1) = NA_REAL; // NAN;\n\n            return rangeVec;\n        }\n    }\n\n    rangeVec.at(0) = *std::min_element(x.begin(), x.end());\n    rangeVec.at(1) = *std::max_element(x.begin(), x.end());\n\n    return rangeVec;\n}\nv1 <- c(10, 4, NA, 6, 8)\n\nrange(v1, na.rm = FALSE)\n#> [1] NA NA\nrangeC_NA(v1, FALSE)\n#> [1] NA NA\n\nrange(v1, na.rm = TRUE)\n#> [1]  4 10\nrangeC_NA(v1, TRUE)\n#> [1]  4 10#include <Rcpp.h>\nusing namespace Rcpp;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nNumericVector cumsumC(NumericVector x) {\n  int n = x.size();\n  NumericVector out(n);\n  \n  out[0] = x[0];\n  for(int i = 1; i < n; ++i) {\n    out[i] = out[i - 1] + x[i];\n  }\n  \n  return out;\n}\nv1 <- c(1, 2, 3, 4)\nv2 <- c(1, 2, NA, 4)\n\ncumsum(v1)\n#> [1]  1  3  6 10\ncumsumC(v1)\n#> [1]  1  3  6 10\n\ncumsum(v2)\n#> [1]  1  3 NA NA\ncumsumC(v2)\n#> [1]  1  3 NA NA"},{"path":"rewriting-r-code-in-c.html","id":"standard-template-library-exercises-25.5.7","chapter":"25 Rewriting R code in C++","heading":"25.3 Standard Template Library (Exercises 25.5.7)","text":"Q1. practice using STL algorithms data structures, implement following using R functions C++, using hints provided:A1.median.default() using partial_sort.%% using unordered_set find() count() methods.unique() using unordered_set (challenge: one line!).Note functions comparable. far can see, way get output R version function using unordered_set data structure.can make comparable version using set data structure:min() using std::min(), max() using std::max()..min() using min_element, .max() using max_element.setdiff(), union(), intersect() integers using sorted ranges set_union, set_intersection set_difference.Note following C++ implementations given functions strictly equivalent R versions. far can see, way identical satisfying specifications mentioned question.union()intersect()setdiff()","code":"#include <iostream>\n#include <vector>\n#include <algorithm>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\ndouble medianC(std::vector<double> &x)\n{\n    int middleIndex = static_cast<int>(x.size() / 2);\n\n    std::partial_sort(x.begin(), x.begin() + middleIndex, x.end());\n\n    // for even number of observations\n    if (x.size() % 2 == 0)\n    {\n        return (x[middleIndex - 1] + x[middleIndex]) / 2;\n    }\n\n    return x[middleIndex];\n}\nv1 <- c(1, 3, 3, 6, 7, 8, 9)\nv2 <- c(1, 2, 3, 4, 5, 6, 8, 9)\n\nmedian.default(v1)\n#> [1] 6\nmedianC(v1)\n#> [1] 6\n\nmedian.default(v2)\n#> [1] 4.5\nmedianC(v2)\n#> [1] 4.5\n\n# performance benefits?\nbench::mark(\n  median.default(v2),\n  medianC(v2),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression              min   median `itr/sec` mem_alloc\n#>   <bch:expr>         <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 median.default(v2)  21.91µs  23.52µs    40942.        0B\n#> 2 medianC(v2)          1.45µs   1.67µs   565586.    2.49KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0#include <vector>\n#include <unordered_set>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::vector<bool> matchC(const std::vector<double> &x, const std::vector<double> &table)\n{\n    std::unordered_set<double> tableUnique(table.begin(), table.end());\n    std::vector<bool> out;\n\n    for (const auto &xElem : x)\n    {\n        out.push_back(tableUnique.find(xElem) != tableUnique.end() ? true : false);\n    }\n\n    return out;\n}\nx1 <- c(3, 4, 8)\nx2 <- c(1, 2, 3, 3, 4, 4, 5, 6)\n\nx1 %in% x2\n#> [1]  TRUE  TRUE FALSE\nmatchC(x1, x2)\n#> [1]  TRUE  TRUE FALSE\n\n# performance benefits?\nbench::mark(\n  x1 %in% x2,\n  matchC(x1, x2),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression          min   median `itr/sec` mem_alloc\n#>   <bch:expr>     <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 x1 %in% x2     881.03ns   1.11µs   703925.        0B\n#> 2 matchC(x1, x2)   2.07µs   2.35µs   409169.    6.62KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0#include <unordered_set>\n#include <vector>\n#include <iostream>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::unordered_set<double> uniqueC(const std::vector<double> &x)\n{\n    std::unordered_set<double> xSet(x.begin(), x.end());\n\n    return xSet;\n}\nv1 <- c(1, 3, 3, 6, 7, 8, 9)\n\nunique(v1)\n#> [1] 1 3 6 7 8 9\nuniqueC(v1)\n#> [1] 9 8 7 6 3 1#include <set>\n#include <vector>\n#include <iostream>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::set<double> uniqueC2(const std::vector<double> &x)\n{\n    std::set<double> xSet(x.begin(), x.end());\n\n    return xSet;\n}\nv1 <- c(1, 3, 3, 6, 7, 8, 9)\n\nunique(v1)\n#> [1] 1 3 6 7 8 9\nuniqueC2(v1)\n#> [1] 1 3 6 7 8 9\n\n# performance benefits?\nbench::mark(\n  unique(v1),\n  uniqueC2(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression        min   median `itr/sec` mem_alloc\n#>   <bch:expr>   <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 unique(v1)     2.13µs   2.43µs   343230.        0B\n#> 2 uniqueC2(v1)   1.69µs   1.85µs   512921.    6.62KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0#include <iostream>\n#include <vector>\n#include <algorithm>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nconst double minC(const std::vector<double> &x)\n{\n     return *std::min_element(x.begin(), x.end());\n}\n\n// [[Rcpp::export]]\nconst double maxC(std::vector<double> x)\n{\n     return *std::max_element(x.begin(), x.end());\n}\nv1 <- c(3, 3, 6, 1, 9, 7, 8)\n\nmin(v1)\n#> [1] 1\nminC(v1)\n#> [1] 1\n\n# performance benefits?\nbench::mark(\n  min(v1),\n  minC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression      min   median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1 min(v1)    240.05ns 251.11ns  3549516.        0B        0\n#> 2 minC(v1)     1.39µs   1.51µs   612452.    6.62KB        0\n\nmax(v1)\n#> [1] 9\nmaxC(v1)\n#> [1] 9\n\n# performance benefits?\nbench::mark(\n  max(v1),\n  maxC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression      min   median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1 max(v1)    220.03ns  250.1ns  3345068.        0B        0\n#> 2 maxC(v1)     1.39µs    1.6µs   595400.    6.62KB        0#include <vector>\n#include <algorithm>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nint which_maxC(std::vector<double> &x)\n{\n    int maxIndex = std::distance(x.begin(), std::max_element(x.begin(), x.end()));\n  \n    // R is 1-index based, while C++ is 0-index based\n    return maxIndex + 1;\n}\n\n// [[Rcpp::export]]\nint which_minC(std::vector<double> &x)\n{\n    int minIndex = std::distance(x.begin(), std::min_element(x.begin(), x.end()));\n  \n    // R is 1-index based, while C++ is 0-index based\n    return minIndex + 1;\n}\nv1 <- c(3, 3, 6, 1, 9, 7, 8)\n\nwhich.min(v1)\n#> [1] 4\nwhich_minC(v1)\n#> [1] 4\n\n# performance benefits?\nbench::mark(\n  which.min(v1),\n  which_minC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression          min   median `itr/sec` mem_alloc\n#>   <bch:expr>     <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 which.min(v1)   401.1ns 431.09ns  2009835.        0B\n#> 2 which_minC(v1)    1.4µs   1.62µs   589790.    6.62KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0\n\nwhich.max(v1)\n#> [1] 5\nwhich_maxC(v1)\n#> [1] 5\n\n# performance benefits?\nbench::mark(\n  which.max(v1),\n  which_maxC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression          min   median `itr/sec` mem_alloc\n#>   <bch:expr>     <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 which.max(v1)  390.11ns 461.12ns  1940961.        0B\n#> 2 which_maxC(v1)   1.39µs   1.55µs   608448.    6.62KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0#include <algorithm>\n#include <iostream>\n#include <vector>\n#include <set>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::set<int> unionC(std::vector<int> &v1, std::vector<int> &v2)\n{\n    std::sort(v1.begin(), v1.end());\n    std::sort(v2.begin(), v2.end());\n\n    std::vector<int> union_vec(v1.size() + v2.size());\n    auto it = std::set_union(v1.begin(), v1.end(), v2.begin(), v2.end(), union_vec.begin());\n\n    union_vec.resize(it - union_vec.begin());\n    std::set<int> union_set(union_vec.begin(), union_vec.end());\n\n    return union_set;\n}\nv1 <- c(1, 4, 5, 5, 5, 6, 2)\nv2 <- c(4, 1, 6, 8)\n\nunion(v1, v2)\n#> [1] 1 4 5 6 2 8\nunionC(v1, v2)\n#> [1] 1 2 4 5 6 8#include <algorithm>\n#include <iostream>\n#include <vector>\n#include <set>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::set<int> intersectC(std::vector<int> &v1, std::vector<int> &v2)\n{\n    std::sort(v1.begin(), v1.end());\n    std::sort(v2.begin(), v2.end());\n\n    std::vector<int> union_vec(v1.size() + v2.size());\n    auto it = std::set_intersection(v1.begin(), v1.end(), v2.begin(), v2.end(), union_vec.begin());\n\n    union_vec.resize(it - union_vec.begin());\n    std::set<int> union_set(union_vec.begin(), union_vec.end());\n\n    return union_set;\n}\nv1 <- c(1, 4, 5, 5, 5, 6, 2)\nv2 <- c(4, 1, 6, 8)\n\nintersect(v1, v2)\n#> [1] 1 4 6\nintersectC(v1, v2)\n#> [1] 1 4 6#include <algorithm>\n#include <iostream>\n#include <vector>\n#include <set>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::set<int> setdiffC(std::vector<int> &v1, std::vector<int> &v2)\n{\n    std::sort(v1.begin(), v1.end());\n    std::sort(v2.begin(), v2.end());\n\n    std::vector<int> union_vec(v1.size() + v2.size());\n    auto it = std::set_difference(v1.begin(), v1.end(), v2.begin(), v2.end(), union_vec.begin());\n\n    union_vec.resize(it - union_vec.begin());\n    std::set<int> union_set(union_vec.begin(), union_vec.end());\n\n    return union_set;\n}\nv1 <- c(1, 4, 5, 5, 5, 6, 2)\nv2 <- c(4, 1, 6, 8)\n\nsetdiff(v1, v2)\n#> [1] 5 2\nsetdiffC(v1, v2)\n#> [1] 2 5"},{"path":"rewriting-r-code-in-c.html","id":"session-information-18","chapter":"25 Rewriting R code in C++","heading":"25.4 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2023-12-31\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bench         1.1.3   2023-05-04 [1] RSPM\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  profmem       0.6.0   2020-12-13 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  Rcpp        * 1.0.11  2023-07-06 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"}]
+[{"path":"index.html","id":"about","chapter":"About","heading":"About","text":"book provides solutions exercises Hadley Wickham’s Advanced R (2nd edition) book.started working book part process learn solving book’s exercises. comparing solutions official solutions manual, realized solutions took different approaches least explained differently. ’m sharing solutions case others might find another perspective explanation official solution manual helpful building understanding.Although tried make sure solutions correct, blame inaccuracies lies solely . ’d much appreciate suggestions corrections.","code":""},{"path":"introduction.html","id":"introduction","chapter":"1 Introduction","heading":"1 Introduction","text":"exercises.","code":""},{"path":"names-and-values.html","id":"names-and-values","chapter":"2 Names and values","heading":"2 Names and values","text":"Loading needed libraries:","code":"\nlibrary(lobstr)"},{"path":"names-and-values.html","id":"binding-basics-exercise-2.2.2","chapter":"2 Names and values","heading":"2.1 Binding basics (Exercise 2.2.2)","text":"Q1. Explain relationship , b, c d following code:A1. names (, b, c) values point object memory, can seen identical memory addresses:Except d, different object, even value , b, c:Q2. following code accesses mean function multiple ways. point underlying function object? Verify lobstr::obj_addr().A2. listed function calls point underlying function object memory, shown object’s memory address:Q3. default, base R data import functions, like read.csv(), automatically convert non-syntactic names syntactic ones. might problematic? option allows suppress behaviour?A3. conversion non-syntactic names syntactic ones can sometimes corrupt data. datasets may require non-syntactic names.suppress behavior, one can set check.names = FALSE.Q4. rules make.names() use convert non-syntactic names syntactic ones?A4. make.names() uses following rules convert non-syntactic names syntactic ones:prepends non-syntactic names Xit converts invalid characters (like @) .adds . suffix name reserved keywordQ5. slightly simplified rules govern syntactic names. .123e1 syntactic name? Read ?make.names full details.A5. .123e1 syntacti name parsed number, string:docs mention (emphasis mine):syntactically valid name consists letters, numbers dot underline characters starts letter dot followed number.","code":"\na <- 1:10\nb <- a\nc <- b\nd <- 1:10\nobj_addrs <- obj_addrs(list(a, b, c))\nunique(obj_addrs)\n#> [1] \"0x56426e6d12c8\"\nobj_addr(d)\n#> [1] \"0x56426e38afc8\"\nmean\nbase::mean\nget(\"mean\")\nevalq(mean)\nmatch.fun(\"mean\")\nobj_addrs <- obj_addrs(list(\n  mean,\n  base::mean,\n  get(\"mean\"),\n  evalq(mean),\n  match.fun(\"mean\")\n))\n\nunique(obj_addrs)\n#> [1] \"0x56426c0b3500\"\nmake.names(c(\"123abc\", \"@me\", \"_yu\", \"  gh\", \"else\"))\n#> [1] \"X123abc\" \"X.me\"    \"X_yu\"    \"X..gh\"   \"else.\"\ntypeof(.123e1)\n#> [1] \"double\""},{"path":"names-and-values.html","id":"copy-on-modify-exercise-2.3.6","chapter":"2 Names and values","heading":"2.2 Copy-on-modify (Exercise 2.3.6)","text":"Q1. tracemem(1:10) useful?A1. tracemem() traces copying objects R. example:since object created memory 1:10 assigned name, can’t addressed modified R, nothing trace.Q2. Explain tracemem() shows two copies run code. Hint: carefully look difference code code shown earlier section.A2. initial atomic vector type integer, 4 (4L) type double. new copy created.Trying integer create another copy:understand still produces copy, explanation official solutions manual:Please aware running code RStudio result additional copies reference environment pane.Q3. Sketch relationship following objects:A3. can understand relationship objects looking memory addresses:learn:name references object 1:10 memory.name b bound list two references memory address .name c also bound list references b, 1:10 object (bound name).Q4. happens run code?Draw picture.A4.don’t access OmniGraffle software, including figure official solution manual:","code":"\nx <- 1:10\n\ntracemem(x)\n#> [1] \"<0x5642701b6628>\"\n\nx <- x + 1\n\nuntracemem(x)\nobj_addr(1:10)\n#> [1] \"0x56426fc8b0f8\"\n\ntracemem(1:10)\n#> [1] \"<0x56426fcf77a0>\"\nx <- c(1L, 2L, 3L)\ntracemem(x)\n\nx[[3]] <- 4\nuntracemem(x)\nx <- c(1L, 2L, 3L)\ntypeof(x)\n#> [1] \"integer\"\ntracemem(x)\n#> [1] \"<0x56427053bad8>\"\n\nx[[3]] <- 4\n#> tracemem[0x56427053bad8 -> 0x564270667558]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local \n#> tracemem[0x564270667558 -> 0x564270675188]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local\nuntracemem(x)\n\ntypeof(x)\n#> [1] \"double\"\nx <- c(1L, 2L, 3L)\ntypeof(x)\n#> [1] \"integer\"\ntracemem(x)\n#> [1] \"<0x564270b34198>\"\n\nx[[3]] <- 4L\n#> tracemem[0x564270b34198 -> 0x564270c5fe18]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local\nuntracemem(x)\n\ntypeof(x)\n#> [1] \"integer\"\na <- 1:10\nb <- list(a, a)\nc <- list(b, a, 1:10)\na <- 1:10\nb <- list(a, a)\nc <- list(b, a, 1:10)\n\nref(a)\n#> [1:0x56427131bcc8] <int>\n\nref(b)\n#> █ [1:0x5642718493e8] <list> \n#> ├─[2:0x56427131bcc8] <int> \n#> └─[2:0x56427131bcc8]\n\nref(c)\n#> █ [1:0x5642718cfda8] <list> \n#> ├─█ [2:0x5642718493e8] <list> \n#> │ ├─[3:0x56427131bcc8] <int> \n#> │ └─[3:0x56427131bcc8] \n#> ├─[3:0x56427131bcc8] \n#> └─[4:0x5642718d2178] <int>\nx <- list(1:10)\nx[[2]] <- x\nx <- list(1:10)\nx\n#> [[1]]\n#>  [1]  1  2  3  4  5  6  7  8  9 10\nobj_addr(x)\n#> [1] \"0x56426f689e78\"\n\nx[[2]] <- x\nx\n#> [[1]]\n#>  [1]  1  2  3  4  5  6  7  8  9 10\n#> \n#> [[2]]\n#> [[2]][[1]]\n#>  [1]  1  2  3  4  5  6  7  8  9 10\nobj_addr(x)\n#> [1] \"0x56426fe84148\"\n\nref(x)\n#> █ [1:0x56426fe84148] <list> \n#> ├─[2:0x56426eee5b38] <int> \n#> └─█ [3:0x56426f689e78] <list> \n#>   └─[2:0x56426eee5b38]"},{"path":"names-and-values.html","id":"object-size-exercise-2.4.1","chapter":"2 Names and values","heading":"2.3 Object size (Exercise 2.4.1)","text":"Q1. following example, object.size(y) obj_size(y) radically different? Consult documentation object.size().A1. mentioned docs object.size():function…detect elements list shared.sizes different:Q2. Take following list. size somewhat misleading?A2. functions externally created objects R, always available part base packages, doesn’t make much sense measure size never going available.Q3. Predict output following code:A3. Correctly predicted 😉Key pieces information keep mind make correct predictions:Size empty vectorSize single double: 8 bytesCopy--modify semantics","code":"\ny <- rep(list(runif(1e4)), 100)\n\nobject.size(y)\nobj_size(y)\ny <- rep(list(runif(1e4)), 100)\n\nobject.size(y)\n#> 8005648 bytes\n\nobj_size(y)\n#> 80.90 kB\nfuns <- list(mean, sd, var)\nobj_size(funs)\nfuns <- list(mean, sd, var)\nobj_size(funs)\n#> 17.55 kB\na <- runif(1e6)\nobj_size(a)\n\nb <- list(a, a)\nobj_size(b)\nobj_size(a, b)\n\nb[[1]][[1]] <- 10\nobj_size(b)\nobj_size(a, b)\n\nb[[2]][[1]] <- 10\nobj_size(b)\nobj_size(a, b)\na <- runif(1e6)\nobj_size(a)\n#> 8.00 MB\n\nb <- list(a, a)\nobj_size(b)\n#> 8.00 MB\nobj_size(a, b)\n#> 8.00 MB\n\nb[[1]][[1]] <- 10\nobj_size(b)\n#> 16.00 MB\nobj_size(a, b)\n#> 16.00 MB\n\nb[[2]][[1]] <- 10\nobj_size(b)\n#> 16.00 MB\nobj_size(a, b)\n#> 24.00 MB\nobj_size(double())\n#> 48 B\nobj_size(double(1))\n#> 56 B"},{"path":"names-and-values.html","id":"modify-in-place-exercise-2.5.3","chapter":"2 Names and values","heading":"2.4 Modify-in-place (Exercise 2.5.3)","text":"Q1. Explain following code doesn’t create circular list.A1. Copy--modify prevents creation circular list.Q2. Wrap two methods subtracting medians two functions, use ‘bench’ package carefully compare speeds. performance change number columns increase?A2. Let’s first microbenchmark functions create copies varying lengths number columns.Plotting benchmarks reveals performance gets increasingly worse number data frames increases:Q3. happens attempt use tracemem() environment?A3. doesn’t work documentation tracemem() makes clear :useful trace NULL, environments, promises, weak references, external pointer objects, duplicated","code":"\nx <- list()\nx[[1]] <- x\nx <- list()\n\nobj_addr(x)\n#> [1] \"0x5642702ece68\"\n\ntracemem(x)\n#> [1] \"<0x5642702ece68>\"\n\nx[[1]] <- x\n#> tracemem[0x5642702ece68 -> 0x56426fbe3420]: eval eval eval_with_user_handlers withVisible withCallingHandlers handle timing_fn evaluate_call <Anonymous> evaluate in_dir in_input_dir eng_r block_exec call_block process_group.block process_group withCallingHandlers withCallingHandlers handle_error process_file <Anonymous> <Anonymous> do.call eval eval eval eval eval.parent local\n\nobj_addr(x[[1]])\n#> [1] \"0x5642702ece68\"\nlibrary(bench)\nlibrary(tidyverse)\n\ngenerateDataFrame <- function(ncol) {\n  as.data.frame(matrix(runif(100 * ncol), nrow = 100))\n}\n\nwithCopy <- function(ncol) {\n  x <- generateDataFrame(ncol)\n  medians <- vapply(x, median, numeric(1))\n\n  for (i in seq_along(medians)) {\n    x[[i]] <- x[[i]] - medians[[i]]\n  }\n\n  return(x)\n}\n\nwithoutCopy <- function(ncol) {\n  x <- generateDataFrame(ncol)\n  medians <- vapply(x, median, numeric(1))\n\n  y <- as.list(x)\n\n  for (i in seq_along(medians)) {\n    y[[i]] <- y[[i]] - medians[[i]]\n  }\n\n  return(y)\n}\n\nbenchComparison <- function(ncol) {\n  bench::mark(\n    withCopy(ncol),\n    withoutCopy(ncol),\n    iterations = 100,\n    check = FALSE\n  ) %>%\n    dplyr::select(expression:total_time)\n}\n\nnColList <- list(1, 10, 50, 100, 250, 500, 1000)\n\nnames(nColList) <- as.character(nColList)\n\nbenchDf <- purrr::map_dfr(\n  .x = nColList,\n  .f = benchComparison,\n  .id = \"nColumns\"\n)\nggplot(\n  benchDf,\n  aes(\n    x = as.numeric(nColumns),\n    y = median,\n    group = as.character(expression),\n    color = as.character(expression)\n  )\n) +\n  geom_line() +\n  labs(\n    x = \"Number of Columns\",\n    y = \"Median Execution Time (ms)\",\n    colour = \"Type of function\"\n  )\ne <- rlang::env(a = 1, b = \"3\")\ntracemem(e)\n#> Error in tracemem(e): 'tracemem' is not useful for promise and environment objects"},{"path":"names-and-values.html","id":"session-information","chapter":"2 Names and values","heading":"2.5 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bench       * 1.1.3   2023-05-04 [1] RSPM\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  colorspace    2.1-0   2023-01-23 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  crayon        1.5.2   2022-09-29 [1] RSPM\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr       * 1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  farver        2.1.1   2022-07-06 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  forcats     * 1.0.0   2023-01-29 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  ggplot2     * 3.4.4   2023-10-12 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  grid          4.3.2   2023-11-01 [3] local\n#>  gtable        0.3.4   2023-08-21 [1] RSPM\n#>  highr         0.10    2022-12-22 [1] RSPM\n#>  hms           1.1.3   2023-03-21 [1] RSPM\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  labeling      0.4.3   2023-08-29 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  lobstr      * 1.1.2   2022-06-22 [1] RSPM\n#>  lubridate   * 1.9.3   2023-09-27 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  munsell       0.5.0   2018-06-12 [1] RSPM\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  prettyunits   1.2.0   2023-09-24 [1] RSPM\n#>  profmem       0.6.0   2020-12-13 [1] RSPM\n#>  purrr       * 1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  readr       * 2.1.4   2023-02-10 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  scales        1.3.0   2023-11-28 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  stringi       1.8.3   2023-12-11 [1] RSPM\n#>  stringr     * 1.5.1   2023-11-14 [1] RSPM\n#>  tibble      * 3.2.1   2023-03-20 [1] RSPM\n#>  tidyr       * 1.3.0   2023-01-24 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tidyverse   * 2.0.0   2023-02-22 [1] RSPM\n#>  timechange    0.2.0   2023-01-11 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  tzdb          0.4.0   2023-05-12 [1] RSPM\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"vectors.html","id":"vectors","chapter":"3 Vectors","heading":"3 Vectors","text":"","code":""},{"path":"vectors.html","id":"atomic-vectors-exercises-3.2.5","chapter":"3 Vectors","heading":"3.1 Atomic vectors (Exercises 3.2.5)","text":"Q1. create raw complex scalars? (See ?raw ?complex.)A1. R, scalars nothing vectors length 1, can created using constructor.Raw vectorsThe raw type holds raw bytes, can created using charToRaw(). example,alternative use .raw():Complex vectorsComplex vectors used represent (surprise!) complex numbers.Example complex scalar:Q2. Test knowledge vector coercion rules predicting output following uses c():A2. vector coercion rules dictate data type smaller size converted data type bigger size.Q3. 1 == \"1\" true? -1 < FALSE true? \"one\" < 2 false?A3. coercion rules vectors reveal comparisons return results .Q4. default missing value, NA, logical vector? ’s special logical vectors? (Hint: think c(FALSE, NA_character_).)A4. \"logical\" type lowest coercion hierarchy.NA defaulting type (e.g. \"numeric\") mean time missing element vector, rest elements converted type higher hierarchy, problematic types lower hierarchy.Q5. Precisely .atomic(), .numeric(), .vector() test ?A5. Let’s discuss one--one..atomic()function checks object vector atomic type (NULL).Quoting docs:.atomic true atomic types (“logical”, “integer”, “numeric”, “complex”, “character” “raw”) NULL..numeric()documentation says:.numeric return true base type class double integer values can reasonably regarded numericTherefore, function checks double integer base types types based top types (factor, Date, POSIXt, difftime)..vector()per documentation:.vector returns TRUE x vector specified mode attributes names. returns FALSE otherwise.Thus, function can incorrectif object attributes names.better way check vector:","code":"\nx <- \"A string\"\n\n(y <- charToRaw(x))\n#> [1] 41 20 73 74 72 69 6e 67\n\ntypeof(y)\n#> [1] \"raw\"\nas.raw(\"–\") # en-dash\n#> Warning: NAs introduced by coercion\n#> Warning: out-of-range values treated as 0 in coercion to\n#> raw\n#> [1] 00\nas.raw(\"—\") # em-dash\n#> Warning: NAs introduced by coercion\n\n#> Warning: out-of-range values treated as 0 in coercion to\n#> raw\n#> [1] 00\n(x <- complex(length.out = 1, real = 1, imaginary = 8))\n#> [1] 1+8i\n\ntypeof(x)\n#> [1] \"complex\"\nc(1, FALSE)\nc(\"a\", 1)\nc(TRUE, 1L)\nc(1, FALSE)\n#> [1] 1 0\n\nc(\"a\", 1)\n#> [1] \"a\" \"1\"\n\nc(TRUE, 1L)\n#> [1] 1 1\n1 == \"1\"\n#> [1] TRUE\n\nc(1, \"1\")\n#> [1] \"1\" \"1\"\n-1 < FALSE\n#> [1] TRUE\n\nc(-1, FALSE)\n#> [1] -1  0\n\"one\" < 2\n#> [1] FALSE\n\nc(\"one\", 2)\n#> [1] \"one\" \"2\"\n\nsort(c(\"one\", 2))\n#> [1] \"2\"   \"one\"\ntypeof(NA)\n#> [1] \"logical\"\n\nc(FALSE, NA_character_)\n#> [1] \"FALSE\" NA\nis.atomic(NULL)\n#> [1] TRUE\n\nis.atomic(list(NULL))\n#> [1] FALSE\nis.numeric(1L)\n#> [1] TRUE\n\nis.numeric(factor(1L))\n#> [1] FALSE\nx <- c(\"x\" = 1, \"y\" = 2)\n\nis.vector(x)\n#> [1] TRUE\n\nattr(x, \"m\") <- \"abcdef\"\n\nis.vector(x)\n#> [1] FALSE\nis.null(dim(x))\n#> [1] TRUE"},{"path":"vectors.html","id":"attributes-exercises-3.3.4","chapter":"3 Vectors","heading":"3.2 Attributes (Exercises 3.3.4)","text":"Q1. setNames() implemented? unname() implemented? Read source code.A1. Let’s look implementations functions.setNames()Given function signature, can see , first argument given, result still named vector.unname()unname() removes existing names (dimnames) setting NULL.Q2. dim() return applied 1-dimensional vector? might use NROW() NCOL()?A2. Dimensions 1-dimensional vector NULL. example,NROW() NCOL() helpful getting dimensions 1D vectors treating matrices dataframes.Q3. describe following three objects? makes different 1:5?A3. x1, x2, x3 one-dimensional arrays, different “orientations”, mentally visualize .x1 5 entries third dimension, x2 second dimension, x1 first dimension.Q4. early draft used code illustrate structure():print object don’t see comment attribute. ? attribute missing, something else special ? (Hint: try using help.)A4. ?attributes (emphasis mine):Note attributes (namely class, comment, dim, dimnames, names, row.names tsp) treated specially restrictions values can set.","code":"\nsetNames\n#> function (object = nm, nm) \n#> {\n#>     names(object) <- nm\n#>     object\n#> }\n#> <bytecode: 0x55c0caac1880>\n#> <environment: namespace:stats>\nsetNames(, c(\"a\", \"b\"))\n#>   a   b \n#> \"a\" \"b\"\n\nsetNames(c(1, 2), c(\"a\", \"b\"))\n#> a b \n#> 1 2\nunname\n#> function (obj, force = FALSE) \n#> {\n#>     if (!is.null(names(obj))) \n#>         names(obj) <- NULL\n#>     if (!is.null(dimnames(obj)) && (force || !is.data.frame(obj))) \n#>         dimnames(obj) <- NULL\n#>     obj\n#> }\n#> <bytecode: 0x55c0c9ecc578>\n#> <environment: namespace:base>\nunname(setNames(, c(\"a\", \"b\")))\n#> [1] \"a\" \"b\"\ndim(c(1, 2))\n#> NULL\n# example-1\nx <- character(0)\n\ndim(x)\n#> NULL\n\nnrow(x)\n#> NULL\nNROW(x)\n#> [1] 0\n\nncol(x)\n#> NULL\nNCOL(x)\n#> [1] 1\n\n# example-2\ny <- 1:4\n\ndim(y)\n#> NULL\n\nnrow(y)\n#> NULL\nNROW(y)\n#> [1] 4\n\nncol(y)\n#> NULL\nNCOL(y)\n#> [1] 1\nx1 <- array(1:5, c(1, 1, 5))\nx2 <- array(1:5, c(1, 5, 1))\nx3 <- array(1:5, c(5, 1, 1))\nstructure(1:5, comment = \"my attribute\")\n#> [1] 1 2 3 4 5\nstructure(1:5, x = \"my attribute\")\n#> [1] 1 2 3 4 5\n#> attr(,\"x\")\n#> [1] \"my attribute\"\n\nstructure(1:5, comment = \"my attribute\")\n#> [1] 1 2 3 4 5"},{"path":"vectors.html","id":"s3-atomic-vectors-exercises-3.4.5","chapter":"3 Vectors","heading":"3.3 S3 atomic vectors (Exercises 3.4.5)","text":"Q1. sort object table() return? type? attributes ? dimensionality change tabulate variables?A1. table() returns array integer type dimensions scale number variables present.Q2. happens factor modify levels?A2. levels change underlying integer values remain .Q3. code ? f2 f3 differ f1?A3. code:f2: underlying integers reversed, levels remain unchanged.f3: levels underlying integers reversed.","code":"\n(x <- table(mtcars$am))\n#> \n#>  0  1 \n#> 19 13\n(y <- table(mtcars$am, mtcars$cyl))\n#>    \n#>      4  6  8\n#>   0  3  4 12\n#>   1  8  3  2\n(z <- table(mtcars$am, mtcars$cyl, mtcars$vs))\n#> , ,  = 0\n#> \n#>    \n#>      4  6  8\n#>   0  0  0 12\n#>   1  1  3  2\n#> \n#> , ,  = 1\n#> \n#>    \n#>      4  6  8\n#>   0  3  4  0\n#>   1  7  0  0\n\n# type\npurrr::map(list(x, y, z), typeof)\n#> [[1]]\n#> [1] \"integer\"\n#> \n#> [[2]]\n#> [1] \"integer\"\n#> \n#> [[3]]\n#> [1] \"integer\"\n\n# attributes\npurrr::map(list(x, y, z), attributes)\n#> [[1]]\n#> [[1]]$dim\n#> [1] 2\n#> \n#> [[1]]$dimnames\n#> [[1]]$dimnames[[1]]\n#> [1] \"0\" \"1\"\n#> \n#> \n#> [[1]]$class\n#> [1] \"table\"\n#> \n#> \n#> [[2]]\n#> [[2]]$dim\n#> [1] 2 3\n#> \n#> [[2]]$dimnames\n#> [[2]]$dimnames[[1]]\n#> [1] \"0\" \"1\"\n#> \n#> [[2]]$dimnames[[2]]\n#> [1] \"4\" \"6\" \"8\"\n#> \n#> \n#> [[2]]$class\n#> [1] \"table\"\n#> \n#> \n#> [[3]]\n#> [[3]]$dim\n#> [1] 2 3 2\n#> \n#> [[3]]$dimnames\n#> [[3]]$dimnames[[1]]\n#> [1] \"0\" \"1\"\n#> \n#> [[3]]$dimnames[[2]]\n#> [1] \"4\" \"6\" \"8\"\n#> \n#> [[3]]$dimnames[[3]]\n#> [1] \"0\" \"1\"\n#> \n#> \n#> [[3]]$class\n#> [1] \"table\"\nf1 <- factor(letters)\nlevels(f1) <- rev(levels(f1))\nf1 <- factor(letters)\nf1\n#>  [1] a b c d e f g h i j k l m n o p q r s t u v w x y z\n#> 26 Levels: a b c d e f g h i j k l m n o p q r s t u ... z\nas.integer(f1)\n#>  [1]  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18\n#> [19] 19 20 21 22 23 24 25 26\n\nlevels(f1) <- rev(levels(f1))\nf1\n#>  [1] z y x w v u t s r q p o n m l k j i h g f e d c b a\n#> 26 Levels: z y x w v u t s r q p o n m l k j i h g f ... a\nas.integer(f1)\n#>  [1]  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18\n#> [19] 19 20 21 22 23 24 25 26\nf2 <- rev(factor(letters))\nf3 <- factor(letters, levels = rev(letters))\nf2 <- rev(factor(letters))\nf2\n#>  [1] z y x w v u t s r q p o n m l k j i h g f e d c b a\n#> 26 Levels: a b c d e f g h i j k l m n o p q r s t u ... z\nas.integer(f2)\n#>  [1] 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10  9\n#> [19]  8  7  6  5  4  3  2  1\nf3 <- factor(letters, levels = rev(letters))\nf3\n#>  [1] a b c d e f g h i j k l m n o p q r s t u v w x y z\n#> 26 Levels: z y x w v u t s r q p o n m l k j i h g f ... a\nas.integer(f3)\n#>  [1] 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10  9\n#> [19]  8  7  6  5  4  3  2  1"},{"path":"vectors.html","id":"lists-exercises-3.5.4","chapter":"3 Vectors","heading":"3.4 Lists (Exercises 3.5.4)","text":"Q1. List ways list differs atomic vector.A1. table comparison:Q2. need use unlist() convert list atomic vector? doesn’t .vector() work?A2. list already (generic) vector, .vector() going change anything, .atomic.vector. Thus, need use unlist().Q3. Compare contrast c() unlist() combining date date-time single vector.A3. Let’s first create date datetime objectAnd check attributes underlying double representation:Behavior c()Since S3 method c() dispatches first argument, resulting class vector going first argument. , attributes lost.Behavior unlist()removes attributes left underlying double representations objects.","code":"\nx <- list(a = 1, b = 2)\n\nis.vector(x)\n#> [1] TRUE\nis.atomic(x)\n#> [1] FALSE\n\n# still a list\nas.vector(x)\n#> $a\n#> [1] 1\n#> \n#> $b\n#> [1] 2\n\n# now a vector\nunlist(x)\n#> a b \n#> 1 2\ndate <- as.Date(\"1947-08-15\")\ndatetime <- as.POSIXct(\"1950-01-26 00:01\", tz = \"UTC\")\nattributes(date)\n#> $class\n#> [1] \"Date\"\nattributes(datetime)\n#> $class\n#> [1] \"POSIXct\" \"POSIXt\" \n#> \n#> $tzone\n#> [1] \"UTC\"\n\nas.double(date) # number of days since the Unix epoch 1970-01-01\n#> [1] -8175\nas.double(datetime) # number of seconds since then\n#> [1] -628991940\nc(date, datetime)\n#> [1] \"1947-08-15\" \"1950-01-26\"\n\nattributes(c(date, datetime))\n#> $class\n#> [1] \"Date\"\n\nc(datetime, date)\n#> [1] \"1950-01-26 00:01:00 UTC\" \"1947-08-15 00:00:00 UTC\"\n\nattributes(c(datetime, date))\n#> $class\n#> [1] \"POSIXct\" \"POSIXt\" \n#> \n#> $tzone\n#> [1] \"UTC\"\nunlist(list(date, datetime))\n#> [1]      -8175 -628991940\n\nunlist(list(datetime, date))\n#> [1] -628991940      -8175"},{"path":"vectors.html","id":"data-frames-and-tibbles-exercises-3.6.8","chapter":"3 Vectors","heading":"3.5 Data frames and tibbles (Exercises 3.6.8)","text":"Q1. Can data frame zero rows? zero columns?A1. Data frame 0 rows possible. basically list vector length 0.Data frame 0 columns also possible. empty list., finally, data frame 0 rows columns also possible:Although, might common create data frames, can results subsetting. example,Q2. happens attempt set rownames unique?A2. attempt set data frame rownames unique, work.Q3. df data frame, can say t(df), t(t(df))? Perform experiments, making sure try different column types.A3. Transposing data frame:transforms matrixcoerces elements typeQ4. .matrix() applied data frame columns different types? differ data.matrix()?A4. return type .matrix() depends data frame column types.docs .matrix() mention:method data frames return character matrix atomic columns non-(numeric/logical/complex) column, applying .vector factors format non-character columns. Otherwise usual coercion hierarchy (logical < integer < double < complex) used, e.g. -logical data frames coerced logical matrix, mixed logical-integer give integer matrix, etc.Let’s experiment:hand, data.matrix() always returns numeric matrix.documentation data.matrix():Return matrix obtained converting variables data frame numeric mode binding together columns matrix. Factors ordered factors replaced internal codes. […] Character columns first converted factors integers.Let’s experiment:","code":"\ndata.frame(x = numeric(0))\n#> [1] x\n#> <0 rows> (or 0-length row.names)\ndata.frame(row.names = 1)\n#> data frame with 0 columns and 1 row\ndata.frame()\n#> data frame with 0 columns and 0 rows\n\ndim(data.frame())\n#> [1] 0 0\nBOD[0, ]\n#> [1] Time   demand\n#> <0 rows> (or 0-length row.names)\n\nBOD[, 0]\n#> data frame with 0 columns and 6 rows\n\nBOD[0, 0]\n#> data frame with 0 columns and 0 rows\ndata.frame(row.names = c(1, 1))\n#> Error in data.frame(row.names = c(1, 1)): duplicate row.names: 1\n# original\n(df <- head(iris))\n#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species\n#> 1          5.1         3.5          1.4         0.2  setosa\n#> 2          4.9         3.0          1.4         0.2  setosa\n#> 3          4.7         3.2          1.3         0.2  setosa\n#> 4          4.6         3.1          1.5         0.2  setosa\n#> 5          5.0         3.6          1.4         0.2  setosa\n#> 6          5.4         3.9          1.7         0.4  setosa\n\n# transpose\nt(df)\n#>              1        2        3        4        5       \n#> Sepal.Length \"5.1\"    \"4.9\"    \"4.7\"    \"4.6\"    \"5.0\"   \n#> Sepal.Width  \"3.5\"    \"3.0\"    \"3.2\"    \"3.1\"    \"3.6\"   \n#> Petal.Length \"1.4\"    \"1.4\"    \"1.3\"    \"1.5\"    \"1.4\"   \n#> Petal.Width  \"0.2\"    \"0.2\"    \"0.2\"    \"0.2\"    \"0.2\"   \n#> Species      \"setosa\" \"setosa\" \"setosa\" \"setosa\" \"setosa\"\n#>              6       \n#> Sepal.Length \"5.4\"   \n#> Sepal.Width  \"3.9\"   \n#> Petal.Length \"1.7\"   \n#> Petal.Width  \"0.4\"   \n#> Species      \"setosa\"\n\n# transpose of a transpose\nt(t(df))\n#>   Sepal.Length Sepal.Width Petal.Length Petal.Width\n#> 1 \"5.1\"        \"3.5\"       \"1.4\"        \"0.2\"      \n#> 2 \"4.9\"        \"3.0\"       \"1.4\"        \"0.2\"      \n#> 3 \"4.7\"        \"3.2\"       \"1.3\"        \"0.2\"      \n#> 4 \"4.6\"        \"3.1\"       \"1.5\"        \"0.2\"      \n#> 5 \"5.0\"        \"3.6\"       \"1.4\"        \"0.2\"      \n#> 6 \"5.4\"        \"3.9\"       \"1.7\"        \"0.4\"      \n#>   Species \n#> 1 \"setosa\"\n#> 2 \"setosa\"\n#> 3 \"setosa\"\n#> 4 \"setosa\"\n#> 5 \"setosa\"\n#> 6 \"setosa\"\n\n# is it a dataframe?\nis.data.frame(df)\n#> [1] TRUE\nis.data.frame(t(df))\n#> [1] FALSE\nis.data.frame(t(t(df)))\n#> [1] FALSE\n\n# check type\ntypeof(df)\n#> [1] \"list\"\ntypeof(t(df))\n#> [1] \"character\"\ntypeof(t(t(df)))\n#> [1] \"character\"\n\n# check dimensions\ndim(df)\n#> [1] 6 5\ndim(t(df))\n#> [1] 5 6\ndim(t(t(df)))\n#> [1] 6 5\n# example with mixed types (coerced to character)\n(df <- head(iris))\n#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species\n#> 1          5.1         3.5          1.4         0.2  setosa\n#> 2          4.9         3.0          1.4         0.2  setosa\n#> 3          4.7         3.2          1.3         0.2  setosa\n#> 4          4.6         3.1          1.5         0.2  setosa\n#> 5          5.0         3.6          1.4         0.2  setosa\n#> 6          5.4         3.9          1.7         0.4  setosa\n\nas.matrix(df)\n#>   Sepal.Length Sepal.Width Petal.Length Petal.Width\n#> 1 \"5.1\"        \"3.5\"       \"1.4\"        \"0.2\"      \n#> 2 \"4.9\"        \"3.0\"       \"1.4\"        \"0.2\"      \n#> 3 \"4.7\"        \"3.2\"       \"1.3\"        \"0.2\"      \n#> 4 \"4.6\"        \"3.1\"       \"1.5\"        \"0.2\"      \n#> 5 \"5.0\"        \"3.6\"       \"1.4\"        \"0.2\"      \n#> 6 \"5.4\"        \"3.9\"       \"1.7\"        \"0.4\"      \n#>   Species \n#> 1 \"setosa\"\n#> 2 \"setosa\"\n#> 3 \"setosa\"\n#> 4 \"setosa\"\n#> 5 \"setosa\"\n#> 6 \"setosa\"\n\nstr(as.matrix(df))\n#>  chr [1:6, 1:5] \"5.1\" \"4.9\" \"4.7\" \"4.6\" \"5.0\" \"5.4\" ...\n#>  - attr(*, \"dimnames\")=List of 2\n#>   ..$ : chr [1:6] \"1\" \"2\" \"3\" \"4\" ...\n#>   ..$ : chr [1:5] \"Sepal.Length\" \"Sepal.Width\" \"Petal.Length\" \"Petal.Width\" ...\n\n# another example (no such coercion)\nBOD\n#>   Time demand\n#> 1    1    8.3\n#> 2    2   10.3\n#> 3    3   19.0\n#> 4    4   16.0\n#> 5    5   15.6\n#> 6    7   19.8\n\nas.matrix(BOD)\n#>      Time demand\n#> [1,]    1    8.3\n#> [2,]    2   10.3\n#> [3,]    3   19.0\n#> [4,]    4   16.0\n#> [5,]    5   15.6\n#> [6,]    7   19.8\ndata.matrix(df)\n#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species\n#> 1          5.1         3.5          1.4         0.2       1\n#> 2          4.9         3.0          1.4         0.2       1\n#> 3          4.7         3.2          1.3         0.2       1\n#> 4          4.6         3.1          1.5         0.2       1\n#> 5          5.0         3.6          1.4         0.2       1\n#> 6          5.4         3.9          1.7         0.4       1\n\nstr(data.matrix(df))\n#>  num [1:6, 1:5] 5.1 4.9 4.7 4.6 5 5.4 3.5 3 3.2 3.1 ...\n#>  - attr(*, \"dimnames\")=List of 2\n#>   ..$ : chr [1:6] \"1\" \"2\" \"3\" \"4\" ...\n#>   ..$ : chr [1:5] \"Sepal.Length\" \"Sepal.Width\" \"Petal.Length\" \"Petal.Width\" ..."},{"path":"vectors.html","id":"session-information-1","chapter":"3 Vectors","heading":"3.6 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  purrr         1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"subsetting.html","id":"subsetting","chapter":"4 Subsetting","heading":"4 Subsetting","text":"Attaching needed libraries:","code":"\nlibrary(tibble)"},{"path":"subsetting.html","id":"selecting-multiple-elements-exercises-4.2.6","chapter":"4 Subsetting","heading":"4.1 Selecting multiple elements (Exercises 4.2.6)","text":"Q1. Fix following common data frame subsetting errors:A1. Fixed versions commands:Q2. following code yield five missing values?A2. two reasons:default type NA R logical type.R recycles indexes match length vector.Q3. upper.tri() return? subsetting matrix work? need additional subsetting rules describe behaviour?A3. documentation upper.tri() states-Returns matrix logicals size given matrix entries TRUE upper triangleWhen used matrix subsetting, elements corresponding TRUE subsetting matrix selected. , instead matrix, returns vector:Q4. mtcars[1:20] return error? differ similar mtcars[1:20, ]?A4. indexed like list, data frame columns given indices selected.mtcars[1:20] doesn’t work 11 columns mtcars dataset.hand, mtcars[1:20, ] indexes dataframe like matrix, indeed 20 rows mtcars, columns rows selected.Q5. Implement function extracts diagonal entries matrix (behave like diag(x) x matrix).A5. can combine existing functions advantage:Q6. df[.na(df)] <- 0 ? work?A6. expression replaces every instance NA df 0..na(df) produces matrix logical values, provides way subsetting.","code":"mtcars[mtcars$cyl = 4, ]\nmtcars[-1:4, ]\nmtcars[mtcars$cyl <= 5]\nmtcars[mtcars$cyl == 4 | 6, ]\n# `==` instead of `=`\nmtcars[mtcars$cyl == 4, ]\n\n# `-(1:4)` instead of `-1:4`\nmtcars[-(1:4), ]\n\n# `,` was missing\nmtcars[mtcars$cyl <= 5, ]\n\n# correct subsetting syntax\nmtcars[mtcars$cyl == 4 | mtcars$cyl == 6, ]\nmtcars[mtcars$cyl %in% c(4, 6), ]\nx <- 1:5\nx[NA]\n#> [1] NA NA NA NA NA\ntypeof(NA)\n#> [1] \"logical\"\nx <- 1:5\nx[c(TRUE, FALSE)] # recycled to c(TRUE, FALSE, TRUE, FALSE, TRUE)\n#> [1] 1 3 5\nx <- outer(1:5, 1:5, FUN = \"*\")\nx[upper.tri(x)]\n(x <- outer(1:5, 1:5, FUN = \"*\"))\n#>      [,1] [,2] [,3] [,4] [,5]\n#> [1,]    1    2    3    4    5\n#> [2,]    2    4    6    8   10\n#> [3,]    3    6    9   12   15\n#> [4,]    4    8   12   16   20\n#> [5,]    5   10   15   20   25\n\nupper.tri(x)\n#>       [,1]  [,2]  [,3]  [,4]  [,5]\n#> [1,] FALSE  TRUE  TRUE  TRUE  TRUE\n#> [2,] FALSE FALSE  TRUE  TRUE  TRUE\n#> [3,] FALSE FALSE FALSE  TRUE  TRUE\n#> [4,] FALSE FALSE FALSE FALSE  TRUE\n#> [5,] FALSE FALSE FALSE FALSE FALSE\nx[upper.tri(x)]\n#>  [1]  2  3  6  4  8 12  5 10 15 20\nhead(mtcars[1:2])\n#>                    mpg cyl\n#> Mazda RX4         21.0   6\n#> Mazda RX4 Wag     21.0   6\n#> Datsun 710        22.8   4\n#> Hornet 4 Drive    21.4   6\n#> Hornet Sportabout 18.7   8\n#> Valiant           18.1   6\nnrow(mtcars[1:20, ])\n#> [1] 20\nx[!upper.tri(x) & !lower.tri(x)]\n#> [1]  1  4  9 16 25\n\ndiag(x)\n#> [1]  1  4  9 16 25\n(df <- tibble(x = c(1, 2, NA), y = c(NA, 5, NA)))\n#> # A tibble: 3 × 2\n#>       x     y\n#>   <dbl> <dbl>\n#> 1     1    NA\n#> 2     2     5\n#> 3    NA    NA\n\nis.na(df)\n#>          x     y\n#> [1,] FALSE  TRUE\n#> [2,] FALSE FALSE\n#> [3,]  TRUE  TRUE\n\nclass(is.na(df))\n#> [1] \"matrix\" \"array\""},{"path":"subsetting.html","id":"selecting-a-single-element-exercises-4.3.5","chapter":"4 Subsetting","heading":"4.2 Selecting a single element (Exercises 4.3.5)","text":"Q1. Brainstorm many ways possible extract third value cyl variable mtcars dataset.A1. Possible ways extract third value cyl variable mtcars dataset:Q2. Given linear model, e.g., mod <- lm(mpg ~ wt, data = mtcars), extract residual degrees freedom. extract R squared model summary (summary(mod))A2. Given objects class lm lists, can use subsetting operators extract elements want.extracting residual degrees freedomextracting R squared model summary","code":"\nmtcars[[\"cyl\"]][[3]]\n#> [1] 4\nmtcars[[c(2, 3)]]\n#> [1] 4\nmtcars[3, ][[\"cyl\"]]\n#> [1] 4\nmtcars[3, ]$cyl\n#> [1] 4\nmtcars[3, \"cyl\"]\n#> [1] 4\nmtcars[, \"cyl\"][[3]]\n#> [1] 4\nmtcars[3, 2]\n#> [1] 4\nmtcars$cyl[[3]]\n#> [1] 4\nmod <- lm(mpg ~ wt, data = mtcars)\nclass(mod)\n#> [1] \"lm\"\ntypeof(mod)\n#> [1] \"list\"\nmod$df.residual \n#> [1] 30\nmod[[\"df.residual\"]]\n#> [1] 30\nsummary(mod)$r.squared\n#> [1] 0.7528328\nsummary(mod)[[\"r.squared\"]]\n#> [1] 0.7528328"},{"path":"subsetting.html","id":"applications-exercises-4.5.9","chapter":"4 Subsetting","heading":"4.3 Applications (Exercises 4.5.9)","text":"Q1. randomly permute columns data frame? (important technique random forests.) Can simultaneously permute rows columns one step?A1. Let’s create small data frame work .randomly permute columns data frame, can combine [ sample() follows:randomly permute columnsrandomly permute rowsrandomly permute columns rowsQ2. select random sample m rows data frame? sample contiguous (.e., initial row, final row, every row )?A2. Let’s create small data frame work .select random sample m rows data frame, can combine [ sample() follows:random non-contiguous sample m rows data framerandom contiguous sample m rows data frameQ3. put columns data frame alphabetical order?A3. can sort columns data frame alphabetical order using [ order():","code":"\ndf <- head(mtcars)\n\n# original\ndf\n#>                    mpg cyl disp  hp drat    wt  qsec vs am\n#> Mazda RX4         21.0   6  160 110 3.90 2.620 16.46  0  1\n#> Mazda RX4 Wag     21.0   6  160 110 3.90 2.875 17.02  0  1\n#> Datsun 710        22.8   4  108  93 3.85 2.320 18.61  1  1\n#> Hornet 4 Drive    21.4   6  258 110 3.08 3.215 19.44  1  0\n#> Hornet Sportabout 18.7   8  360 175 3.15 3.440 17.02  0  0\n#> Valiant           18.1   6  225 105 2.76 3.460 20.22  1  0\n#>                   gear carb\n#> Mazda RX4            4    4\n#> Mazda RX4 Wag        4    4\n#> Datsun 710           4    1\n#> Hornet 4 Drive       3    1\n#> Hornet Sportabout    3    2\n#> Valiant              3    1\ndf[sample.int(ncol(df))]\n#>                   drat    wt carb am  qsec vs  hp  mpg disp\n#> Mazda RX4         3.90 2.620    4  1 16.46  0 110 21.0  160\n#> Mazda RX4 Wag     3.90 2.875    4  1 17.02  0 110 21.0  160\n#> Datsun 710        3.85 2.320    1  1 18.61  1  93 22.8  108\n#> Hornet 4 Drive    3.08 3.215    1  0 19.44  1 110 21.4  258\n#> Hornet Sportabout 3.15 3.440    2  0 17.02  0 175 18.7  360\n#> Valiant           2.76 3.460    1  0 20.22  1 105 18.1  225\n#>                   cyl gear\n#> Mazda RX4           6    4\n#> Mazda RX4 Wag       6    4\n#> Datsun 710          4    4\n#> Hornet 4 Drive      6    3\n#> Hornet Sportabout   8    3\n#> Valiant             6    3\ndf[sample.int(nrow(df)), ]\n#>                    mpg cyl disp  hp drat    wt  qsec vs am\n#> Datsun 710        22.8   4  108  93 3.85 2.320 18.61  1  1\n#> Mazda RX4 Wag     21.0   6  160 110 3.90 2.875 17.02  0  1\n#> Mazda RX4         21.0   6  160 110 3.90 2.620 16.46  0  1\n#> Hornet Sportabout 18.7   8  360 175 3.15 3.440 17.02  0  0\n#> Hornet 4 Drive    21.4   6  258 110 3.08 3.215 19.44  1  0\n#> Valiant           18.1   6  225 105 2.76 3.460 20.22  1  0\n#>                   gear carb\n#> Datsun 710           4    1\n#> Mazda RX4 Wag        4    4\n#> Mazda RX4            4    4\n#> Hornet Sportabout    3    2\n#> Hornet 4 Drive       3    1\n#> Valiant              3    1\ndf[sample.int(nrow(df)), sample.int(ncol(df))]\n#>                    qsec vs gear am    wt drat carb disp  hp\n#> Mazda RX4         16.46  0    4  1 2.620 3.90    4  160 110\n#> Hornet 4 Drive    19.44  1    3  0 3.215 3.08    1  258 110\n#> Datsun 710        18.61  1    4  1 2.320 3.85    1  108  93\n#> Mazda RX4 Wag     17.02  0    4  1 2.875 3.90    4  160 110\n#> Valiant           20.22  1    3  0 3.460 2.76    1  225 105\n#> Hornet Sportabout 17.02  0    3  0 3.440 3.15    2  360 175\n#>                    mpg cyl\n#> Mazda RX4         21.0   6\n#> Hornet 4 Drive    21.4   6\n#> Datsun 710        22.8   4\n#> Mazda RX4 Wag     21.0   6\n#> Valiant           18.1   6\n#> Hornet Sportabout 18.7   8\ndf <- head(mtcars)\n\n# original\ndf\n#>                    mpg cyl disp  hp drat    wt  qsec vs am\n#> Mazda RX4         21.0   6  160 110 3.90 2.620 16.46  0  1\n#> Mazda RX4 Wag     21.0   6  160 110 3.90 2.875 17.02  0  1\n#> Datsun 710        22.8   4  108  93 3.85 2.320 18.61  1  1\n#> Hornet 4 Drive    21.4   6  258 110 3.08 3.215 19.44  1  0\n#> Hornet Sportabout 18.7   8  360 175 3.15 3.440 17.02  0  0\n#> Valiant           18.1   6  225 105 2.76 3.460 20.22  1  0\n#>                   gear carb\n#> Mazda RX4            4    4\n#> Mazda RX4 Wag        4    4\n#> Datsun 710           4    1\n#> Hornet 4 Drive       3    1\n#> Hornet Sportabout    3    2\n#> Valiant              3    1\n\n# number of rows to sample\nm <- 2L\ndf[sample(nrow(df), m), ]\n#>                mpg cyl disp  hp drat    wt  qsec vs am gear\n#> Valiant       18.1   6  225 105 2.76 3.460 20.22  1  0    3\n#> Mazda RX4 Wag 21.0   6  160 110 3.90 2.875 17.02  0  1    4\n#>               carb\n#> Valiant          1\n#> Mazda RX4 Wag    4\n# select a random starting position from available number of rows\nstart_row <- sample(nrow(df) - m + 1, size = 1)\n\n# adjust ending position while avoiding off-by-one error\nend_row <- start_row + m - 1\n\ndf[start_row:end_row, ]\n#>               mpg cyl disp  hp drat    wt  qsec vs am gear\n#> Mazda RX4      21   6  160 110  3.9 2.620 16.46  0  1    4\n#> Mazda RX4 Wag  21   6  160 110  3.9 2.875 17.02  0  1    4\n#>               carb\n#> Mazda RX4        4\n#> Mazda RX4 Wag    4\n# columns in original order\nnames(mtcars)\n#>  [1] \"mpg\"  \"cyl\"  \"disp\" \"hp\"   \"drat\" \"wt\"   \"qsec\" \"vs\"  \n#>  [9] \"am\"   \"gear\" \"carb\"\n\n# columns in alphabetical order\nnames(mtcars[order(names(mtcars))])\n#>  [1] \"am\"   \"carb\" \"cyl\"  \"disp\" \"drat\" \"gear\" \"hp\"   \"mpg\" \n#>  [9] \"qsec\" \"vs\"   \"wt\""},{"path":"subsetting.html","id":"session-information-2","chapter":"4 Subsetting","heading":"4.4 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble      * 3.2.1   2023-03-20 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"control-flow.html","id":"control-flow","chapter":"5 Control flow","heading":"5 Control flow","text":"","code":""},{"path":"control-flow.html","id":"choices-exercises-5.2.4","chapter":"5 Control flow","heading":"5.1 Choices (Exercises 5.2.4)","text":"Q1. type vector following calls ifelse() return?Read documentation write rules words.A1. rules call ifelse() might return:type unstable, .e. type return depend type condition true (yes , .e.):works cases test argument evaluates logical type:test argument logical type, NA, return NA:test argument doesn’t resolve logical type, try coerce output logical type:also clarified docs function:vector length attributes (including dimensions \"class\") test data values values yes . mode answer coerced logical accommodate first values taken yes values taken .Q2. following code work?A2. code works conditional expressions () - even though numeric type - can successfully coerced logical type.","code":"\nifelse(TRUE, 1, \"no\")\nifelse(FALSE, 1, \"no\")\nifelse(NA, 1, \"no\")\nifelse(TRUE, 1, \"no\") # `numeric` returned\n#> [1] 1\nifelse(FALSE, 1, \"no\") # `character` returned\n#> [1] \"no\"\nifelse(NA_real_, 1, \"no\")\n#> [1] NA\nifelse(NaN, 1, \"no\")\n#> [1] NA\nifelse(NA, 1, \"no\")\n#> [1] NA\n# will work\nifelse(\"TRUE\", 1, \"no\")\n#> [1] 1\nifelse(\"false\", 1, \"no\")\n#> [1] \"no\"\n\n# won't work\nifelse(\"tRuE\", 1, \"no\")\n#> [1] NA\nifelse(NaN, 1, \"no\")\n#> [1] NA\nx <- 1:10\nif (length(x)) \"not empty\" else \"empty\"\n#> [1] \"not empty\"\n\nx <- numeric()\nif (length(x)) \"not empty\" else \"empty\"\n#> [1] \"empty\"\nas.logical(length(1:10))\n#> [1] TRUE\n\nas.logical(length(numeric()))\n#> [1] FALSE"},{"path":"control-flow.html","id":"loops-exercises-5.3.3","chapter":"5 Control flow","heading":"5.2 Loops (Exercises 5.3.3)","text":"Q1. code succeed without errors warnings?A1. works 1:length(x) works positive negative directions.case, since x length 0, go 1 0.Additionally, since --bound (OOB) value atomic vectors NA, related operations OOB values also produce NA.way avoid unintended behavior use seq_along() instead:Q2. following code evaluated, can say vector iterated?A2. iterator variable x initially takes values vector xs. can check printing x iteration:worth noting x updated iteration; otherwise, take increasingly bigger values xs, loop never end executing.Q3. following code tell index updated?A3. () loop index updated beginning iteration. Otherwise, encounter infinite loop.Also, worth contrasting behavior () loop () loop:","code":"\nx <- numeric()\nout <- vector(\"list\", length(x))\nfor (i in 1:length(x)) {\n  out[i] <- x[i]^2\n}\nout\n1:2\n#> [1] 1 2\n1:0\n#> [1] 1 0\n1:-3\n#> [1]  1  0 -1 -2 -3\nx <- numeric()\nout <- vector(\"list\", length(x))\n\nfor (i in 1:length(x)) {\n  print(paste(\"i:\", i, \", x[i]:\", x[i], \", out[i]:\", out[i]))\n\n  out[i] <- x[i]^2\n}\n#> [1] \"i: 1 , x[i]: NA , out[i]: NULL\"\n#> [1] \"i: 0 , x[i]:  , out[i]: \"\n\nout\n#> [[1]]\n#> [1] NA\nx <- numeric()\nout <- vector(\"list\", length(x))\n\nfor (i in seq_along(x)) {\n  out[i] <- x[i]^2\n}\n\nout\n#> list()\nxs <- c(1, 2, 3)\nfor (x in xs) {\n  xs <- c(xs, x * 2)\n}\nxs\n#> [1] 1 2 3 2 4 6\nxs <- c(1, 2, 3)\nfor (x in xs) {\n  cat(\"x:\", x, \"\\n\")\n  xs <- c(xs, x * 2)\n  cat(\"xs:\", paste(xs), \"\\n\")\n}\n#> x: 1 \n#> xs: 1 2 3 2 \n#> x: 2 \n#> xs: 1 2 3 2 4 \n#> x: 3 \n#> xs: 1 2 3 2 4 6\nfor (i in 1:3) {\n  i <- i * 2\n  print(i)\n}\n#> [1] 2\n#> [1] 4\n#> [1] 6\nfor (i in 1:3) {\n  cat(\"before: \", i, \"\\n\")\n  i <- i * 2\n  cat(\"after:  \", i, \"\\n\")\n}\n#> before:  1 \n#> after:   2 \n#> before:  2 \n#> after:   4 \n#> before:  3 \n#> after:   6\ni <- 1\nwhile (i < 4) {\n  cat(\"before: \", i, \"\\n\")\n  i <- i * 2\n  cat(\"after:  \", i, \"\\n\")\n}\n#> before:  1 \n#> after:   2 \n#> before:  2 \n#> after:   4"},{"path":"control-flow.html","id":"session-information-3","chapter":"5 Control flow","heading":"5.3 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"functions.html","id":"functions","chapter":"6 Functions","heading":"6 Functions","text":"Attaching needed libraries:","code":"\nlibrary(tidyverse, warn.conflicts = FALSE)"},{"path":"functions.html","id":"function-fundamentals-exercises-6.2.5","chapter":"6 Functions","heading":"6.1 Function fundamentals (Exercises 6.2.5)","text":"Q1. Given name, like \"mean\", match.fun() lets find function. Given function, can find name? doesn’t make sense R?A1. Given name, match.fun() lets find function., given function, doesn’t make sense find name can multiple names bound function.Q2. ’s possible (although typically useful) call anonymous function. two approaches correct? ?A2. first expression correct since function evaluate 3(), syntactically allowed since literals can’t treated like functions.correct way call anonymous function.Q3. good rule thumb anonymous function fit one line shouldn’t need use {}. Review code. used anonymous function instead named function? used named function instead anonymous function?A3. Self activity.Q4. function allows tell object function? function allows tell function primitive function?A4. Use .function() check object function:Use .primitive() check function primitive:Q5. code makes list functions base package.Use answer following questions:base function arguments?base function arguments?many base functions arguments? ’s special functions?many base functions arguments? ’s special functions?adapt code find primitive functions?adapt code find primitive functions?A5. provided code following:base function arguments?can use formals() extract number arguments, function returns NULL primitive functions.Therefore, focus non-primitive functions.scan() function arguments.many base functions arguments? ’s special functions?time writing, 47 base (non-primitive) functions arguments.adapt code find primitive functions?Q6. three important components function?A6. Except primitive functions, functions 3 important components:formals()body()environment()Q7. printing function show environment created ?A7. package functions print environment:two exceptions enclosing environment won’t printed:primitive functionsfunctions created global environment","code":"\nmatch.fun(\"mean\")\n#> function (x, ...) \n#> UseMethod(\"mean\")\n#> <bytecode: 0x55734d1505a8>\n#> <environment: namespace:base>\nf1 <- function(x) mean(x)\nf2 <- f1\n\nmatch.fun(\"f1\")\n#> function(x) mean(x)\n\nmatch.fun(\"f2\")\n#> function(x) mean(x)\nfunction(x) 3()\n#> function(x) 3()\n(function(x) 3)()\n#> [1] 3\nf <- (function(x) 3())\nf\n#> function(x) 3()\nf()\n#> Error in f(): attempt to apply non-function\n\nrlang::is_syntactic_literal(3)\n#> [1] TRUE\ng <- (function(x) 3)\ng\n#> function(x) 3\ng()\n#> [1] 3\n# these are functions\nf <- function(x) 3\nis.function(mean)\n#> [1] TRUE\nis.function(f)\n#> [1] TRUE\n\n# these aren't\nis.function(\"x\")\n#> [1] FALSE\nis.function(new.env())\n#> [1] FALSE\n# primitive\nis.primitive(sum)\n#> [1] TRUE\nis.primitive(`+`)\n#> [1] TRUE\n\n# not primitive\nis.primitive(mean)\n#> [1] FALSE\nis.primitive(read.csv)\n#> [1] FALSE\nobjs <- mget(ls(\"package:base\", all = TRUE), inherits = TRUE)\nfuns <- Filter(is.function, objs)\nobjs <- mget(ls(\"package:base\", all = TRUE), inherits = TRUE)\nfuns <- Filter(is.function, objs)\nformals(\"!\")\n#> NULL\n\nlength(formals(\"!\"))\n#> [1] 0\nfuns <- purrr::discard(funs, is.primitive)\ndf_formals <- purrr::map_df(funs, ~ length(formals(.))) %>%\n  tidyr::pivot_longer(\n    cols = dplyr::everything(),\n    names_to = \"function\",\n    values_to = \"argumentCount\"\n  ) %>%\n  dplyr::arrange(desc(argumentCount))\n\ndf_formals\n#> # A tibble: 1,137 × 2\n#>    `function`       argumentCount\n#>    <chr>                    <int>\n#>  1 scan                        22\n#>  2 format.default              16\n#>  3 source                      16\n#>  4 formatC                     15\n#>  5 library                     13\n#>  6 merge.data.frame            13\n#>  7 prettyNum                   13\n#>  8 system2                     11\n#>  9 print.default               10\n#> 10 save                        10\n#> # ℹ 1,127 more rows\ndplyr::filter(df_formals, argumentCount == 0)\n#> # A tibble: 47 × 2\n#>    `function`               argumentCount\n#>    <chr>                            <int>\n#>  1 .First.sys                           0\n#>  2 .NotYetImplemented                   0\n#>  3 .OptRequireMethods                   0\n#>  4 .standard_regexps                    0\n#>  5 .tryResumeInterrupt                  0\n#>  6 closeAllConnections                  0\n#>  7 contributors                         0\n#>  8 Cstack_info                          0\n#>  9 default.stringsAsFactors             0\n#> 10 extSoftVersion                       0\n#> # ℹ 37 more rows\nobjs <- mget(ls(\"package:base\", all = TRUE), inherits = TRUE)\nfuns <- Filter(is.function, objs)\nprimitives <- Filter(is.primitive, funs)\n\nlength(primitives)\n#> [1] 205\n\nnames(primitives)\n#>   [1] \"-\"                    \":\"                   \n#>   [3] \"::\"                   \":::\"                 \n#>   [5] \"!\"                    \"!=\"                  \n#>   [7] \"...elt\"               \"...length\"           \n#>   [9] \"...names\"             \".C\"                  \n#>  [11] \".cache_class\"         \".Call\"               \n#>  [13] \".Call.graphics\"       \".class2\"             \n#>  [15] \".External\"            \".External.graphics\"  \n#>  [17] \".External2\"           \".Fortran\"            \n#>  [19] \".Internal\"            \".isMethodsDispatchOn\"\n#>  [21] \".Primitive\"           \".primTrace\"          \n#>  [23] \".primUntrace\"         \".subset\"             \n#>  [25] \".subset2\"             \"(\"                   \n#>  [27] \"[\"                    \"[[\"                  \n#>  [29] \"[[<-\"                 \"[<-\"                 \n#>  [31] \"{\"                    \"@\"                   \n#>  [33] \"@<-\"                  \"*\"                   \n#>  [35] \"/\"                    \"&\"                   \n#>  [37] \"&&\"                   \"%*%\"                 \n#>  [39] \"%/%\"                  \"%%\"                  \n#>  [41] \"^\"                    \"+\"                   \n#>  [43] \"<\"                    \"<-\"                  \n#>  [45] \"<<-\"                  \"<=\"                  \n#>  [47] \"=\"                    \"==\"                  \n#>  [49] \">\"                    \">=\"                  \n#>  [51] \"|\"                    \"||\"                  \n#>  [53] \"~\"                    \"$\"                   \n#>  [55] \"$<-\"                  \"abs\"                 \n#>  [57] \"acos\"                 \"acosh\"               \n#>  [59] \"all\"                  \"any\"                 \n#>  [61] \"anyNA\"                \"Arg\"                 \n#>  [63] \"as.call\"              \"as.character\"        \n#>  [65] \"as.complex\"           \"as.double\"           \n#>  [67] \"as.environment\"       \"as.integer\"          \n#>  [69] \"as.logical\"           \"as.numeric\"          \n#>  [71] \"as.raw\"               \"asin\"                \n#>  [73] \"asinh\"                \"atan\"                \n#>  [75] \"atanh\"                \"attr\"                \n#>  [77] \"attr<-\"               \"attributes\"          \n#>  [79] \"attributes<-\"         \"baseenv\"             \n#>  [81] \"break\"                \"browser\"             \n#>  [83] \"c\"                    \"call\"                \n#>  [85] \"ceiling\"              \"class\"               \n#>  [87] \"class<-\"              \"Conj\"                \n#>  [89] \"cos\"                  \"cosh\"                \n#>  [91] \"cospi\"                \"cummax\"              \n#>  [93] \"cummin\"               \"cumprod\"             \n#>  [95] \"cumsum\"               \"digamma\"             \n#>  [97] \"dim\"                  \"dim<-\"               \n#>  [99] \"dimnames\"             \"dimnames<-\"          \n#> [101] \"emptyenv\"             \"enc2native\"          \n#> [103] \"enc2utf8\"             \"environment<-\"       \n#> [105] \"exp\"                  \"expm1\"               \n#> [107] \"expression\"           \"floor\"               \n#> [109] \"for\"                  \"forceAndCall\"        \n#> [111] \"function\"             \"gamma\"               \n#> [113] \"gc.time\"              \"globalenv\"           \n#> [115] \"if\"                   \"Im\"                  \n#> [117] \"interactive\"          \"invisible\"           \n#> [119] \"is.array\"             \"is.atomic\"           \n#> [121] \"is.call\"              \"is.character\"        \n#> [123] \"is.complex\"           \"is.double\"           \n#> [125] \"is.environment\"       \"is.expression\"       \n#> [127] \"is.finite\"            \"is.function\"         \n#> [129] \"is.infinite\"          \"is.integer\"          \n#> [131] \"is.language\"          \"is.list\"             \n#> [133] \"is.logical\"           \"is.matrix\"           \n#> [135] \"is.na\"                \"is.name\"             \n#> [137] \"is.nan\"               \"is.null\"             \n#> [139] \"is.numeric\"           \"is.object\"           \n#> [141] \"is.pairlist\"          \"is.raw\"              \n#> [143] \"is.recursive\"         \"is.single\"           \n#> [145] \"is.symbol\"            \"isS4\"                \n#> [147] \"lazyLoadDBfetch\"      \"length\"              \n#> [149] \"length<-\"             \"levels<-\"            \n#> [151] \"lgamma\"               \"list\"                \n#> [153] \"log\"                  \"log10\"               \n#> [155] \"log1p\"                \"log2\"                \n#> [157] \"max\"                  \"min\"                 \n#> [159] \"missing\"              \"Mod\"                 \n#> [161] \"names\"                \"names<-\"             \n#> [163] \"nargs\"                \"next\"                \n#> [165] \"nzchar\"               \"oldClass\"            \n#> [167] \"oldClass<-\"           \"on.exit\"             \n#> [169] \"pos.to.env\"           \"proc.time\"           \n#> [171] \"prod\"                 \"quote\"               \n#> [173] \"range\"                \"Re\"                  \n#> [175] \"rep\"                  \"repeat\"              \n#> [177] \"retracemem\"           \"return\"              \n#> [179] \"round\"                \"seq_along\"           \n#> [181] \"seq_len\"              \"seq.int\"             \n#> [183] \"sign\"                 \"signif\"              \n#> [185] \"sin\"                  \"sinh\"                \n#> [187] \"sinpi\"                \"sqrt\"                \n#> [189] \"standardGeneric\"      \"storage.mode<-\"      \n#> [191] \"substitute\"           \"sum\"                 \n#> [193] \"switch\"               \"tan\"                 \n#> [195] \"tanh\"                 \"tanpi\"               \n#> [197] \"tracemem\"             \"trigamma\"            \n#> [199] \"trunc\"                \"unCfillPOSIXlt\"      \n#> [201] \"unclass\"              \"untracemem\"          \n#> [203] \"UseMethod\"            \"while\"               \n#> [205] \"xtfrm\"\n# base\nmean\n#> function (x, ...) \n#> UseMethod(\"mean\")\n#> <bytecode: 0x55734d1505a8>\n#> <environment: namespace:base>\n\n# other package function\npurrr::map\n#> function (.x, .f, ..., .progress = FALSE) \n#> {\n#>     map_(\"list\", .x, .f, ..., .progress = .progress)\n#> }\n#> <bytecode: 0x55734f135980>\n#> <environment: namespace:purrr>\nsum\n#> function (..., na.rm = FALSE)  .Primitive(\"sum\")\nf <- function(x) mean(x)\nf\n#> function(x) mean(x)"},{"path":"functions.html","id":"lexical-scoping-exercises-6.4.5","chapter":"6 Functions","heading":"6.2 Lexical scoping (Exercises 6.4.5)","text":"Q1. following code return? ? Describe three c’s interpreted.A1. c(c = c):first c interpreted function call c()second c name vector elementthird c variable value 10You can also see lexical analysis expression:Q2. four principles govern R looks values?A2. Principles govern R looks values:Name masking (names defined inside function mask names defined outside function)Name masking (names defined inside function mask names defined outside function)Functions vs. variables (rule also applies function names)Functions vs. variables (rule also applies function names)fresh start (every time function called, new environment created host execution)fresh start (every time function called, new environment created host execution)Dynamic look-(R looks values function run, function created)Dynamic look-(R looks values function run, function created)Q3. following function return? Make prediction running code .A3. Correctly predicted 😉Although multiple f() functions, order evaluation goes inside outside x^2 evaluated first f(x) * 2 evaluated last. results 202 (= ((10 ^ 2) + 1) * 2).","code":"\nc <- 10\nc(c = c)\nc <- 10\nc(c = c)\n#>  c \n#> 10\np_expr <- parse(text = \"c(c = c)\", keep.source = TRUE)\ngetParseData(p_expr) %>% select(token, text)\n#>                   token text\n#> 12                 expr     \n#> 1  SYMBOL_FUNCTION_CALL    c\n#> 3                  expr     \n#> 2                   '('    (\n#> 4            SYMBOL_SUB    c\n#> 5                EQ_SUB    =\n#> 6                SYMBOL    c\n#> 8                  expr     \n#> 7                   ')'    )\nf <- function(x) {\n  f <- function(x) {\n    f <- function() {\n      x^2\n    }\n    f() + 1\n  }\n  f(x) * 2\n}\nf(10)\nf <- function(x) {\n  f <- function(x) {\n    f <- function() {\n      x^2\n    }\n    f() + 1\n  }\n  f(x) * 2\n}\n\nf(10)\n#> [1] 202"},{"path":"functions.html","id":"lazy-evaluation-exercises-6.5.4","chapter":"6 Functions","heading":"6.3 Lazy evaluation (Exercises 6.5.4)","text":"Q1. important property && makes x_ok() work?different code? behaviour undesirable ?A1. && evaluates left right short-circuit evaluation, .e., first operand TRUE, R short-circuit even look second operand.Replacing && & undesirable performs element-wise logical comparisons returns vector values always useful decision (TRUE, FALSE, NA).Q2. function return? ? principle illustrate?A2. function returns 100 due lazy evaluation.function execution environment encounters x, evaluates argument x = z since name z already bound value 100 environment, x also bound value.can check looking memory addresses:Q3. function return? ? principle illustrate?A3. Let’s first look function returns:name masking. function call c(x, y), x accessed function environment, following promise evaluated function environment:, thus y gets assigned 1, x 2, since last value scope.Therefore, neither promise y = 0 global assignment y <- 10 ever consulted find value y.Q4. hist(), default value xlim range(breaks), default value breaks \"Sturges\", andExplain hist() works get correct xlim value.A4. xlim defines range histogram’s x-axis.default xlim = range(breaks) breaks = \"Sturges\" arguments reveal function uses Sturges’ algorithm compute number breaks.see implementation, run sloop::s3_get_method(\"hist.default\").hist() ensures chosen algorithm returns numeric vector containing least two unique elements xlim computed.Q5. Explain function works. confusing?A5. Let’s take step--step.function argument x missing function call. means stop(\"Error!\") evaluated function environment, global environment., due lazy evaluation, promise stop(\"Error!\") evaluated x accessed. happens print(x) called.print(x) leads x evaluated, evaluates stop function environment. , function environment, base::stop() masked locally defined stop() function, returns Sys.time() output.Q6. many arguments required calling library()?A6. Going solely signature,looks like following arguments required:, reality, one argument required: package. function internally checks arguments missing adjusts accordingly.better arguments NULL instead missing; avoid confusion.","code":"\nx_ok <- function(x) {\n  !is.null(x) && length(x) == 1 && x > 0\n}\n\nx_ok(NULL)\nx_ok(1)\nx_ok(1:3)\nx_ok <- function(x) {\n  !is.null(x) & length(x) == 1 & x > 0\n}\n\nx_ok(NULL)\nx_ok(1)\nx_ok(1:3)\nx_ok <- function(x) {\n  !is.null(x) && length(x) == 1 && x > 0\n}\n\nx_ok(NULL)\n#> [1] FALSE\n\nx_ok(1)\n#> [1] TRUE\n\nx_ok(1:3)\n#> [1] FALSE\nx_ok <- function(x) {\n  !is.null(x) & length(x) == 1 & x > 0\n}\n\nx_ok(NULL)\n#> logical(0)\n\nx_ok(1)\n#> [1] TRUE\n\nx_ok(1:3)\n#> [1] FALSE FALSE FALSE\nf2 <- function(x = z) {\n  z <- 100\n  x\n}\nf2()\nf2 <- function(x = z) {\n  z <- 100\n  print(lobstr::obj_addrs(list(x, z)))\n  x\n}\n\nf2()\n#> [1] \"0x55734d8a0d10\" \"0x55734d8a0d10\"\n#> [1] 100\ny <- 10\nf1 <- function(x =\n                 {\n                   y <- 1\n                   2\n                 },\n               y = 0) {\n  c(x, y)\n}\nf1()\ny\ny <- 10\nf1 <- function(x =\n                 {\n                   y <- 1\n                   2\n                 },\n               y = 0) {\n  c(x, y)\n}\nf1()\n#> [1] 2 1\ny\n#> [1] 10\nx <- {\n  y <- 1\n  2\n}\nrange(\"Sturges\")\n#> [1] \"Sturges\" \"Sturges\"\nhist(mtcars$wt, xlim = c(1, 6))\nnclass.Sturges(mtcars$wt)\n#> [1] 6\nshow_time <- function(x = stop(\"Error!\")) {\n  stop <- function(...) Sys.time()\n  print(x)\n}\n\nshow_time()\n#> [1] \"2024-01-07 00:43:46 UTC\"\nformals(library)\n#> $package\n#> \n#> \n#> $help\n#> \n#> \n#> $pos\n#> [1] 2\n#> \n#> $lib.loc\n#> NULL\n#> \n#> $character.only\n#> [1] FALSE\n#> \n#> $logical.return\n#> [1] FALSE\n#> \n#> $warn.conflicts\n#> \n#> \n#> $quietly\n#> [1] FALSE\n#> \n#> $verbose\n#> getOption(\"verbose\")\n#> \n#> $mask.ok\n#> \n#> \n#> $exclude\n#> \n#> \n#> $include.only\n#> \n#> \n#> $attach.required\n#> missing(include.only)\nformals(library) %>%\n  purrr::discard(is.null) %>%\n  purrr::map_lgl(~ .x == \"\") %>%\n  purrr::keep(~ isTRUE(.x)) %>%\n  names()\n#> [1] \"package\"        \"help\"           \"warn.conflicts\"\n#> [4] \"mask.ok\"        \"exclude\"        \"include.only\""},{"path":"functions.html","id":"dot-dot-dot-exercises-6.6.1","chapter":"6 Functions","heading":"6.4 ... (dot-dot-dot) (Exercises 6.6.1)","text":"Q1. Explain following results:A1. Let’s look arguments functions:can seen, sum() function doesn’t na.omit argument. , input na.omit = TRUE treated 1 (logical implicitly coerced numeric), thus results. , expression evaluates sum(1, 2, 3, 1).mean() function, one parameter (x) ’s matched first argument (1). , expression evaluates mean(1).Q2. Explain find documentation named arguments following function call:A2. Typing ?plot console, see documentation, also shows signature:Since ... passed par(), can look ?par docs:.docs parameters interest reside .Q3. plot(1:10, col = \"red\") colour points, axes labels? Read source code plot.default() find .A3. Source code can found .plot.default() passes ... localTitle(), passes title().title() four parts: main, sub, xlab, ylab.single argument col work ambiguous element apply argument .","code":"\nsum(1, 2, 3)\n#> [1] 6\nmean(1, 2, 3)\n#> [1] 1\n\nsum(1, 2, 3, na.omit = TRUE)\n#> [1] 7\nmean(1, 2, 3, na.omit = TRUE)\n#> [1] 1\nstr(sum)\n#> function (..., na.rm = FALSE)\nstr(mean)\n#> function (x, ...)\nplot(1:10, col = \"red\", pch = 20, xlab = \"x\", col.lab = \"blue\")#> function (x, y, ...)#> function (..., no.readonly = FALSE)\nlocalTitle <- function(..., col, bg, pch, cex, lty, lwd) title(...)\n\ntitle <- function(main = NULL, sub = NULL, xlab = NULL, ylab = NULL,\n                  line = NA, outer = FALSE, ...) {\n  main <- as.graphicsAnnot(main)\n  sub <- as.graphicsAnnot(sub)\n  xlab <- as.graphicsAnnot(xlab)\n  ylab <- as.graphicsAnnot(ylab)\n  .External.graphics(C_title, main, sub, xlab, ylab, line, outer, ...)\n  invisible()\n}"},{"path":"functions.html","id":"exiting-a-function-exercises-6.7.5","chapter":"6 Functions","heading":"6.5 Exiting a function (Exercises 6.7.5)","text":"Q1. load() return? don’t normally see values?A1. load() function reloads datasets saved using save() function:normally don’t see value function loads datasets invisibly.can change setting verbose = TRUE:Q2. write.table() return? useful?A2. write.table() writes data frame file returns NULL invisibly.helpful function invisibly returned actual object written file, used.Q3. chdir parameter source() compare with_dir()? might prefer one ?A3. chdir parameter source() described :TRUE file pathname, R working directory temporarily changed directory containing file evaluatingThat , chdir allows changing working directory temporarily directory containing file sourced:withr::with_dir() temporarily changes current working directory:importantly, parameters dir allows temporarily changing working directory directory.Q4. Write function opens graphics device, runs supplied code, closes graphics device (always, regardless whether plotting code works).A4. function opens graphics device, runs supplied code, closes graphics device:Q5. can use .exit() implement simple version capture.output().Compare capture.output() capture.output2(). functions differ? features removed make key ideas easier see? rewritten key ideas ’re easier understand?A5. capture.output() significantly complex, can seen definition:key differences:capture.output() uses print() function print console:capture.output() can capture messages well:capture.output() takes account visibility expression:","code":"\nsave(iris, file = \"my_iris.rda\")\nload(\"my_iris.rda\")\nload(\"my_iris.rda\", verbose = TRUE)\n#> Loading objects:\n#>   iris\n\n# cleanup\nunlink(\"my_iris.rda\")\nwrite.table(BOD, file = \"BOD.csv\")\n# cleanup\nunlink(\"BOD.csv\")\nwithr::with_dir\n#> function (new, code) \n#> {\n#>     old <- setwd(dir = new)\n#>     on.exit(setwd(old))\n#>     force(code)\n#> }\n#> <bytecode: 0x55735689cc20>\n#> <environment: namespace:withr>\nwith_png_device <- function(filename, code, ...) {\n  grDevices::png(filename = filename, ...)\n  on.exit(grDevices::dev.off(), add = TRUE)\n\n  force(code)\n}\ncapture.output2 <- function(code) {\n  temp <- tempfile()\n  on.exit(file.remove(temp), add = TRUE, after = TRUE)\n\n  sink(temp)\n  on.exit(sink(), add = TRUE, after = TRUE)\n\n  force(code)\n  readLines(temp)\n}\n\ncapture.output2(cat(\"a\", \"b\", \"c\", sep = \"\\n\"))\n#> [1] \"a\" \"b\" \"c\"\ncapture.output\n#> function (..., file = NULL, append = FALSE, type = c(\"output\", \n#>     \"message\"), split = FALSE) \n#> {\n#>     type <- match.arg(type)\n#>     rval <- NULL\n#>     closeit <- TRUE\n#>     if (is.null(file)) \n#>         file <- textConnection(\"rval\", \"w\", local = TRUE)\n#>     else if (is.character(file)) \n#>         file <- file(file, if (append) \n#>             \"a\"\n#>         else \"w\")\n#>     else if (inherits(file, \"connection\")) {\n#>         if (!isOpen(file)) \n#>             open(file, if (append) \n#>                 \"a\"\n#>             else \"w\")\n#>         else closeit <- FALSE\n#>     }\n#>     else stop(\"'file' must be NULL, a character string or a connection\")\n#>     sink(file, type = type, split = split)\n#>     on.exit({\n#>         sink(type = type, split = split)\n#>         if (closeit) close(file)\n#>     })\n#>     for (i in seq_len(...length())) {\n#>         out <- withVisible(...elt(i))\n#>         if (out$visible) \n#>             print(out$value)\n#>     }\n#>     on.exit()\n#>     sink(type = type, split = split)\n#>     if (closeit) \n#>         close(file)\n#>     if (is.null(rval)) \n#>         invisible(NULL)\n#>     else rval\n#> }\n#> <bytecode: 0x557356d2cc20>\n#> <environment: namespace:utils>\ncapture.output(1)\n#> [1] \"[1] 1\"\n\ncapture.output2(1)\n#> character(0)\ncapture.output(message(\"Hi there!\"), \"a\", type = \"message\")\n#> Hi there!\n#> [1] \"a\"\n#> character(0)\ncapture.output(1, invisible(2), 3)\n#> [1] \"[1] 1\" \"[1] 3\""},{"path":"functions.html","id":"function-forms-exercises-6.8.6","chapter":"6 Functions","heading":"6.6 Function forms (Exercises 6.8.6)","text":"Q1. Rewrite following code snippets prefix form:A1. Prefix forms code snippets:Q2. Clarify following list odd function calls:A2. functions don’t dots (...) parameters, argument matching takes place following steps:exact matching named argumentspartial matchingposition-basedQ3. Explain following code fails:A3. provided book, replacement function defined :Let’s re-write provided code prefix format understand doesn’t work:Although works:following doesn’t code evaluates :get<- function R.Q4. Create replacement function modifies random location vector.A4. replacement function modifies random location vector:Let’s try :Q5. Write version + pastes inputs together character vectors behaves usual otherwise. words, make code work:A5. Infix operator re-create desired output:Q6. Create list replacement functions found base package. ones primitive functions? (Hint: use apropos().)A6. Replacement functions always <- end names., using apropos(), can find replacement functions search paths filter ones don’t belong {base} package:primitive replacement functions can listed using .primitive():Q7. valid names user-created infix functions?A7. mentioned respective section book:names infix functions flexible regular R functions: can contain sequence characters except %.Q8. Create infix xor() operator.A8. Exclusive logical operation TRUE arguments differ (one TRUE, FALSE).can create infix operator exclusive like :function vectorized inputs underlying logical operators vectorized.Q9. Create infix versions set functions intersect(), union(), setdiff(). might call %n%, %u%, %/% match conventions mathematics.A9. required infix operators can created following:can check outputs agree underlying functions:","code":"\n1 + 2 + 3\n\n1 + (2 + 3)\n\nif (length(x) <= 5) x[[5]] else x[[n]]\n# The binary `+`  operator has left to right associative property.\n`+`(`+`(1, 2), 3)\n\n`+`(1, `(`(`+`(2, 3)))\n\n`if`(cond = `<=`(length(x), 5), cons.expr = `[[`(x, 5), alt.expr = `[[`(x, n))\nx <- sample(replace = TRUE, 20, x = c(1:10, NA))\ny <- runif(min = 0, max = 1, 20)\ncor(m = \"k\", y = y, u = \"p\", x = x)\nmodify(get(\"x\"), 1) <- 10\n#> Error: target of assignment expands to non-language object\n`modify<-` <- function(x, position, value) {\n  x[position] <- value\n  x\n}\nget(\"x\") <- `modify<-`(x = get(\"x\"), position = 1, value = 10)\nx <- 5\n`modify<-`(x = get(\"x\"), position = 1, value = 10)\n#> [1] 10\n`get<-`(\"x\", 10)\n#> Error in `get<-`(\"x\", 10): could not find function \"get<-\"\n`random_modify<-` <- function(x, value) {\n  random_index <- sample(seq_along(x), size = 1)\n  x[random_index] <- value\n  return(x)\n}\nx1 <- rep(\"a\", 10)\nrandom_modify(x1) <- \"X\"\nx1\n#>  [1] \"a\" \"a\" \"a\" \"a\" \"X\" \"a\" \"a\" \"a\" \"a\" \"a\"\n\nx2 <- rep(\"a\", 10)\nrandom_modify(x2) <- \"Y\"\nx2\n#>  [1] \"a\" \"a\" \"a\" \"a\" \"a\" \"Y\" \"a\" \"a\" \"a\" \"a\"\n\nx3 <- rep(0, 15)\nrandom_modify(x3) <- -4\nx3\n#>  [1]  0  0  0  0 -4  0  0  0  0  0  0  0  0  0  0\n\nx4 <- rep(0, 15)\nrandom_modify(x4) <- -1\nx4\n#>  [1]  0  0  0  0  0  0  0  0  0  0  0  0 -1  0  0\n1 + 2\n#> [1] 3\n\n\"a\" + \"b\"\n#> [1] \"ab\"\n`+` <- function(x, y) {\n  if (is.character(x) || is.character(y)) {\n    paste0(x, y)\n  } else {\n    base::`+`(x, y)\n  }\n}\n\n1 + 2\n#> [1] 3\n\n\"a\" + \"b\"\n#> [1] \"ab\"\n\nrm(\"+\", envir = .GlobalEnv)\nls_replacement <- apropos(\"<-\", where = TRUE, mode = \"function\")\n\nbase_index <- which(grepl(\"base\", searchpaths()))\n\nls_replacement <- ls_replacement[which(names(ls_replacement) == as.character(base_index))]\n\nunname(ls_replacement)\n#>  [1] \".rowNamesDF<-\"           \"[[<-\"                   \n#>  [3] \"[[<-.data.frame\"         \"[[<-.factor\"            \n#>  [5] \"[[<-.numeric_version\"    \"[[<-.POSIXlt\"           \n#>  [7] \"[<-\"                     \"[<-.data.frame\"         \n#>  [9] \"[<-.Date\"                \"[<-.difftime\"           \n#> [11] \"[<-.factor\"              \"[<-.numeric_version\"    \n#> [13] \"[<-.POSIXct\"             \"[<-.POSIXlt\"            \n#> [15] \"@<-\"                     \"<-\"                     \n#> [17] \"<<-\"                     \"$<-\"                    \n#> [19] \"$<-.data.frame\"          \"$<-.POSIXlt\"            \n#> [21] \"attr<-\"                  \"attributes<-\"           \n#> [23] \"body<-\"                  \"class<-\"                \n#> [25] \"colnames<-\"              \"comment<-\"              \n#> [27] \"diag<-\"                  \"dim<-\"                  \n#> [29] \"dimnames<-\"              \"dimnames<-.data.frame\"  \n#> [31] \"Encoding<-\"              \"environment<-\"          \n#> [33] \"formals<-\"               \"is.na<-\"                \n#> [35] \"is.na<-.default\"         \"is.na<-.factor\"         \n#> [37] \"is.na<-.numeric_version\" \"length<-\"               \n#> [39] \"length<-.Date\"           \"length<-.difftime\"      \n#> [41] \"length<-.factor\"         \"length<-.POSIXct\"       \n#> [43] \"length<-.POSIXlt\"        \"levels<-\"               \n#> [45] \"levels<-.factor\"         \"mode<-\"                 \n#> [47] \"mostattributes<-\"        \"names<-\"                \n#> [49] \"names<-.POSIXlt\"         \"oldClass<-\"             \n#> [51] \"parent.env<-\"            \"regmatches<-\"           \n#> [53] \"row.names<-\"             \"row.names<-.data.frame\" \n#> [55] \"row.names<-.default\"     \"rownames<-\"             \n#> [57] \"split<-\"                 \"split<-.data.frame\"     \n#> [59] \"split<-.default\"         \"storage.mode<-\"         \n#> [61] \"substr<-\"                \"substring<-\"            \n#> [63] \"units<-\"                 \"units<-.difftime\"\nmget(ls_replacement, envir = baseenv()) %>%\n  purrr::keep(is.primitive) %>%\n  names()\n#>  [1] \"[[<-\"           \"[<-\"            \"@<-\"           \n#>  [4] \"<-\"             \"<<-\"            \"$<-\"           \n#>  [7] \"attr<-\"         \"attributes<-\"   \"class<-\"       \n#> [10] \"dim<-\"          \"dimnames<-\"     \"environment<-\" \n#> [13] \"length<-\"       \"levels<-\"       \"names<-\"       \n#> [16] \"oldClass<-\"     \"storage.mode<-\"\nlv1 <- c(TRUE, FALSE, TRUE, FALSE)\nlv2 <- c(TRUE, TRUE, FALSE, FALSE)\n\nxor(lv1, lv2)\n#> [1] FALSE  TRUE  TRUE FALSE\n`%xor%` <- function(x, y) {\n  !((x & y) | !(x | y))\n}\n\nlv1 %xor% lv2\n#> [1] FALSE  TRUE  TRUE FALSE\n\nTRUE %xor% TRUE\n#> [1] FALSE\n`%n%` <- function(x, y) {\n  intersect(x, y)\n}\n\n`%u%` <- function(x, y) {\n  union(x, y)\n}\n\n`%/%` <- function(x, y) {\n  setdiff(x, y)\n}\n(x <- c(sort(sample(1:20, 9)), NA))\n#>  [1]  4  7  8  9 11 13 15 16 20 NA\n(y <- c(sort(sample(3:23, 7)), NA))\n#> [1]  9 10 13 15 17 19 20 NA\n\nidentical(intersect(x, y), x %n% y)\n#> [1] TRUE\nidentical(union(x, y), x %u% y)\n#> [1] TRUE\nidentical(setdiff(x, y), x %/% y)\n#> [1] TRUE"},{"path":"functions.html","id":"session-information-4","chapter":"6 Functions","heading":"6.7 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  colorspace    2.1-0   2023-01-23 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr       * 1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  forcats     * 1.0.0   2023-01-29 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  ggplot2     * 3.4.4   2023-10-12 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  grid          4.3.2   2023-11-01 [3] local\n#>  gtable        0.3.4   2023-08-21 [1] RSPM\n#>  highr         0.10    2022-12-22 [1] RSPM\n#>  hms           1.1.3   2023-03-21 [1] RSPM\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  lobstr        1.1.2   2022-06-22 [1] RSPM\n#>  lubridate   * 1.9.3   2023-09-27 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  munsell       0.5.0   2018-06-12 [1] RSPM\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  purrr       * 1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  readr       * 2.1.4   2023-02-10 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  scales        1.3.0   2023-11-28 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  stringi       1.8.3   2023-12-11 [1] RSPM\n#>  stringr     * 1.5.1   2023-11-14 [1] RSPM\n#>  tibble      * 3.2.1   2023-03-20 [1] RSPM\n#>  tidyr       * 1.3.0   2023-01-24 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tidyverse   * 2.0.0   2023-02-22 [1] RSPM\n#>  timechange    0.2.0   2023-01-11 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  tzdb          0.4.0   2023-05-12 [1] RSPM\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"environments.html","id":"environments","chapter":"7 Environments","heading":"7 Environments","text":"Loading needed libraries:","code":"\nlibrary(rlang, warn.conflicts = FALSE)"},{"path":"environments.html","id":"environment-basics-exercises-7.2.7","chapter":"7 Environments","heading":"7.1 Environment basics (Exercises 7.2.7)","text":"Q1. List three ways environment differs list.A1. mentioned book, ways environments differ lists:Q2. Create environment illustrated picture.A2. Creating environment illustrated picture:binding loop memory address environment e:Q3. Create pair environments illustrated picture.A3. Creating specified environment:Q4. Explain e[[1]] e[c(\"\", \"b\")] don’t make sense e environment.A4. environment non-linear data structure, concept ordered elements. Therefore, indexing (e.g. e[[1]]) doesn’t make sense.Subsetting list vector returns subset underlying data structure. example, subsetting vector returns another vector. ’s unclear subsetting environment (e.g. e[c(\"\", \"b\")]) return data structure contain returns. can’t another environment since environments reference semantics.Q5. Create version env_poke() bind new names, never re-bind old names. programming languages , known single assignment languages.A5. Create version env_poke() doesn’t allow re-binding old names:Making sure behaves expected:Contrast behavior following:Q6. function ? differ <<- might prefer ?A6. downside <<- create new binding doesn’t exist given environment, something may wish:rebind() function let us know binding doesn’t exist, much safer:","code":"\nlibrary(rlang)\n\ne <- env()\ne$loop <- e\nenv_print(e)\n#> <environment: 0x5592b255c2e8>\n#> Parent: <environment: global>\n#> Bindings:\n#> • loop: <env>\nlobstr::ref(e$loop)\n#> █ [1:0x5592b255c2e8] <env> \n#> └─loop = [1:0x5592b255c2e8]\ne1 <- env()\ne2 <- env()\n\ne1$loop <- e2\ne2$deloop <- e1\n\n# following should be the same\nlobstr::obj_addrs(list(e1, e2$deloop))\n#> [1] \"0x5592b4bbb1e8\" \"0x5592b4bbb1e8\"\nlobstr::obj_addrs(list(e2, e1$loop))\n#> [1] \"0x5592b4c08948\" \"0x5592b4c08948\"\nenv_poke2 <- function(env, nm, value) {\n  if (env_has(env, nm)) {\n    abort(\"Can't re-bind existing names.\")\n  }\n\n  env_poke(env, nm, value)\n}\ne <- env(a = 1, b = 2, c = 3)\n\n# re-binding old names not allowed\nenv_poke2(e, \"b\", 4)\n#> Error in `env_poke2()`:\n#> ! Can't re-bind existing names.\n\n# binding new names allowed\nenv_poke2(e, \"d\", 8)\ne$d\n#> [1] 8\ne <- env(a = 1, b = 2, c = 3)\n\ne$b\n#> [1] 2\n\n# re-binding old names allowed\nenv_poke(e, \"b\", 4)\ne$b\n#> [1] 4\nrebind <- function(name, value, env = caller_env()) {\n  if (identical(env, empty_env())) {\n    stop(\"Can't find `\", name, \"`\", call. = FALSE)\n  } else if (env_has(env, name)) {\n    env_poke(env, name, value)\n  } else {\n    rebind(name, value, env_parent(env))\n  }\n}\nrebind(\"a\", 10)\n#> Error: Can't find `a`\na <- 5\nrebind(\"a\", 10)\na\n#> [1] 10\n# `x` doesn't exist\nexists(\"x\")\n#> [1] FALSE\n\n# so `<<-` will create one for us\n{\n  x <<- 5\n}\n\n# in the global environment\nenv_has(global_env(), \"x\")\n#>    x \n#> TRUE\nx\n#> [1] 5\nrebind <- function(name, value, env = caller_env()) {\n  if (identical(env, empty_env())) {\n    stop(\"Can't find `\", name, \"`\", call. = FALSE)\n  } else if (env_has(env, name)) {\n    env_poke(env, name, value)\n  } else {\n    rebind(name, value, env_parent(env))\n  }\n}\n\n# doesn't exist\nexists(\"abc\")\n#> [1] FALSE\n\n# so function will produce an error instead of creating it for us\nrebind(\"abc\", 10)\n#> Error: Can't find `abc`\n\n# but it will work as expected when the variable already exists\nabc <- 5\nrebind(\"abc\", 10)\nabc\n#> [1] 10"},{"path":"environments.html","id":"recursing-over-environments-exercises-7.3.1","chapter":"7 Environments","heading":"7.2 Recursing over environments (Exercises 7.3.1)","text":"Q1. Modify () return environments contain binding name. Carefully think type object function need return.A1. modified version () returns environments contain binding name.Since anticipate one environment, dynamically update list time environment specified binding found. important initialize empty list since signifies given binding found environments.Let’s try :Q2. Write function called fget() finds function objects. two arguments, name env, obey regular scoping rules functions: ’s object matching name ’s function, look parent. added challenge, also add inherits argument controls whether function recurses parents looks one environment.A2. function recursively looks function objects:Let’s try :","code":"\nwhere <- function(name, env = caller_env()) {\n  env_list <- list()\n\n  while (!identical(env, empty_env())) {\n    if (env_has(env, name)) {\n      env_list <- append(env_list, env)\n    }\n\n    env <- env_parent(env)\n  }\n\n  return(env_list)\n}\nwhere(\"yyy\")\n#> list()\n\nx <- 5\nwhere(\"x\")\n#> [[1]]\n#> <environment: R_GlobalEnv>\n\nwhere(\"mean\")\n#> [[1]]\n#> <environment: base>\n\nlibrary(dplyr, warn.conflicts = FALSE)\nwhere(\"filter\")\n#> [[1]]\n#> <environment: package:dplyr>\n#> attr(,\"name\")\n#> [1] \"package:dplyr\"\n#> attr(,\"path\")\n#> [1] \"/home/runner/work/_temp/Library/dplyr\"\n#> \n#> [[2]]\n#> <environment: package:stats>\n#> attr(,\"name\")\n#> [1] \"package:stats\"\n#> attr(,\"path\")\n#> [1] \"/opt/R/4.3.2/lib/R/library/stats\"\ndetach(\"package:dplyr\")\nfget <- function(name, env = caller_env(), inherits = FALSE) {\n  # we need only function objects\n  f_value <- mget(name,\n    envir = env,\n    mode = \"function\",\n    inherits = FALSE, # since we have our custom argument\n    ifnotfound = list(NULL)\n  )\n\n  if (!is.null(f_value[[1]])) {\n    # success case\n    f_value[[1]]\n  } else {\n    if (inherits && !identical(env, empty_env())) {\n      # recursive case\n      env <- env_parent(env)\n      fget(name, env, inherits = TRUE)\n    } else {\n      # base case\n      stop(\"No function objects with matching name was found.\", call. = FALSE)\n    }\n  }\n}\nfget(\"mean\", inherits = FALSE)\n#> Error: No function objects with matching name was found.\n\nfget(\"mean\", inherits = TRUE)\n#> function (x, ...) \n#> UseMethod(\"mean\")\n#> <bytecode: 0x5592aff915a8>\n#> <environment: namespace:base>\n\nmean <- 5\nfget(\"mean\", inherits = FALSE)\n#> Error: No function objects with matching name was found.\n\nmean <- function() NULL\nfget(\"mean\", inherits = FALSE)\n#> function() NULL\nrm(\"mean\")"},{"path":"environments.html","id":"special-environments-exercises-7.4.5","chapter":"7 Environments","heading":"7.3 Special environments (Exercises 7.4.5)","text":"Q1. search_envs() different env_parents(global_env())?A1. search_envs() lists chain environments currently attached search path contains exported functions packages. search path always ends {base} package environment. search path also includes global environment.env_parents() lists parent environments empty environment. course, global environment included list.Q2. Draw diagram shows enclosing environments function:A2. don’t access graphics software used create diagrams book, linking diagram official solutions manual, also find detailed description figure:Q3. Write enhanced version str() provides information functions. Show function found environment defined .A3. write required function, can first re-purpose fget() function wrote return environment found enclosing environment:Let’s try :can now write new version str() wrapper around function. need foresee users might enter function name either symbol string.Let’s first try base::mean():variant present global environment:","code":"\nsearch_envs()\n#>  [[1]] $ <env: global>\n#>  [[2]] $ <env: package:rlang>\n#>  [[3]] $ <env: package:magrittr>\n#>  [[4]] $ <env: package:stats>\n#>  [[5]] $ <env: package:graphics>\n#>  [[6]] $ <env: package:grDevices>\n#>  [[7]] $ <env: package:utils>\n#>  [[8]] $ <env: package:datasets>\n#>  [[9]] $ <env: package:methods>\n#> [[10]] $ <env: Autoloads>\n#> [[11]] $ <env: package:base>\nenv_parents(global_env())\n#>  [[1]] $ <env: package:rlang>\n#>  [[2]] $ <env: package:magrittr>\n#>  [[3]] $ <env: package:stats>\n#>  [[4]] $ <env: package:graphics>\n#>  [[5]] $ <env: package:grDevices>\n#>  [[6]] $ <env: package:utils>\n#>  [[7]] $ <env: package:datasets>\n#>  [[8]] $ <env: package:methods>\n#>  [[9]] $ <env: Autoloads>\n#> [[10]] $ <env: package:base>\n#> [[11]] $ <env: empty>\nf1 <- function(x1) {\n  f2 <- function(x2) {\n    f3 <- function(x3) {\n      x1 + x2 + x3\n    }\n    f3(3)\n  }\n  f2(2)\n}\nf1(1)\nfget2 <- function(name, env = caller_env()) {\n  # we need only function objects\n  f_value <- mget(name,\n    envir = env,\n    mode = \"function\",\n    inherits = FALSE,\n    ifnotfound = list(NULL)\n  )\n\n  if (!is.null(f_value[[1]])) {\n    # success case\n    list(\n      \"where\" = env,\n      \"enclosing\" = fn_env(f_value[[1]])\n    )\n  } else {\n    if (!identical(env, empty_env())) {\n      # recursive case\n      env <- env_parent(env)\n      fget2(name, env)\n    } else {\n      # base case\n      stop(\"No function objects with matching name was found.\", call. = FALSE)\n    }\n  }\n}\nfget2(\"mean\")\n#> $where\n#> <environment: base>\n#> \n#> $enclosing\n#> <environment: namespace:base>\n\nmean <- function() NULL\nfget2(\"mean\")\n#> $where\n#> <environment: R_GlobalEnv>\n#> \n#> $enclosing\n#> <environment: R_GlobalEnv>\nrm(\"mean\")\nstr_function <- function(.f) {\n  fget2(as_string(ensym(.f)))\n}\nstr_function(mean)\n#> $where\n#> <environment: base>\n#> \n#> $enclosing\n#> <environment: namespace:base>\n\nstr_function(\"mean\")\n#> $where\n#> <environment: base>\n#> \n#> $enclosing\n#> <environment: namespace:base>\nmean <- function() NULL\n\nstr_function(mean)\n#> $where\n#> <environment: R_GlobalEnv>\n#> \n#> $enclosing\n#> <environment: R_GlobalEnv>\n\nstr_function(\"mean\")\n#> $where\n#> <environment: R_GlobalEnv>\n#> \n#> $enclosing\n#> <environment: R_GlobalEnv>\n\nrm(\"mean\")"},{"path":"environments.html","id":"call-stacks-exercises-7.5.5","chapter":"7 Environments","heading":"7.4 Call stacks (Exercises 7.5.5)","text":"Q1. Write function lists variables defined environment called. return results ls().A1. function lists variables defined environment called:workhorse rlang::caller_env(), let’s also look definition:Let’s try :global environment:function environment:","code":"\n# let's first remove everything that exists in the global environment right now\n# to test with only newly defined objects\nrm(list = ls())\nrm(.Random.seed, envir = globalenv())\n\nls_env <- function(env = rlang::caller_env()) {\n  sort(rlang::env_names(env))\n}\nrlang::caller_env\n#> function (n = 1) \n#> {\n#>     parent.frame(n + 1)\n#> }\n#> <bytecode: 0x5592b0ab08d0>\n#> <environment: namespace:rlang>\nx <- \"a\"\ny <- 1\n\nls_env()\n#> [1] \"ls_env\" \"x\"      \"y\"\n\nls()\n#> [1] \"ls_env\" \"x\"      \"y\"\nfoo <- function() {\n  a <- \"x\"\n  b <- 2\n\n  print(ls_env())\n\n  print(ls())\n}\n\nfoo()\n#> [1] \"a\" \"b\"\n#> [1] \"a\" \"b\""},{"path":"environments.html","id":"session-information-5","chapter":"7 Environments","heading":"7.5 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  crayon        1.5.2   2022-09-29 [1] RSPM\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr         1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  highr         0.10    2022-12-22 [1] RSPM\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  lobstr        1.1.2   2022-06-22 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  png           0.1-8   2022-11-29 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"conditions.html","id":"conditions","chapter":"8 Conditions","heading":"8 Conditions","text":"Attaching needed libraries:","code":"\nlibrary(rlang, warn.conflicts = FALSE)\nlibrary(testthat, warn.conflicts = FALSE)"},{"path":"conditions.html","id":"signalling-conditions-exercises-8.2.4","chapter":"8 Conditions","heading":"8.1 Signalling conditions (Exercises 8.2.4)","text":"Q1. Write wrapper around file.remove() throws error file deleted exist.A1. Let’s first create wrapper function around file.remove() throws error file deleted exist.Let’s first create file can delete immediately.function fail files provided don’t exist:work expected file exists:Q2. appendLF argument message() ? related cat()?A2. mentioned docs message(), appendLF argument decides:messages given character string newline appended?TRUE (default value), final newline regarded part message:FALSE, messages concatenated:hand, cat() converts arguments character vectors concatenates single character vector default:order get message()-like default behavior outputs, can set sep = \"\\n\":","code":"\nfileRemove <- function(...) {\n  existing_files <- fs::file_exists(...)\n\n  if (!all(existing_files)) {\n    stop(\n      cat(\n        \"The following files to be deleted don't exist:\",\n        names(existing_files[!existing_files]),\n        sep = \"\\n\"\n      ),\n      call. = FALSE\n    )\n  }\n\n  file.remove(...)\n}\nfs::file_create(\"random.R\")\nfileRemove(c(\"random.R\", \"XYZ.csv\"))\n#> The following files to be deleted don't exist:\n#> XYZ.csv\n#> Error:\nfileRemove(\"random.R\")\n#> [1] TRUE\nfoo <- function(appendLF) {\n  message(\"Beetle\", appendLF = appendLF)\n  message(\"Juice\", appendLF = appendLF)\n}\n\nfoo(appendLF = TRUE)\n#> Beetle\n#> Juice\nfoo <- function(appendLF) {\n  message(\"Beetle\", appendLF = appendLF)\n  message(\"Juice\", appendLF = appendLF)\n}\n\nfoo(appendLF = FALSE)\n#> BeetleJuice\nfoo <- function() {\n  cat(\"Beetle\")\n  cat(\"Juice\")\n}\n\nfoo()\n#> BeetleJuice\nfoo <- function() {\n  cat(\"Beetle\", sep = \"\\n\")\n  cat(\"Juice\", sep = \"\\n\")\n}\n\nfoo()\n#> Beetle\n#> Juice"},{"path":"conditions.html","id":"handling-conditions-exercises-8.4.5","chapter":"8 Conditions","heading":"8.2 Handling conditions (Exercises 8.4.5)","text":"Q1. extra information condition generated abort() contain compared condition generated stop() .e. ’s difference two objects? Read help ?abort learn .A1. Compared base::stop(), rlang::abort() contains two additional pieces information:trace: traceback capturing sequence calls lead current functionparent: Information another condition used parent create chained condition.Q2. Predict results evaluating following codeA2. Correctly predicted 😉first three pieces code straightforward:last piece code challenging one illustrates tryCatch() works. docs:several handlers supplied single tryCatch first one considered recent second.Q3. Explain results running code:A3. surprising part output b last c.happens inner calling handler doesn’t handle message, bubbles outer calling handler.Q4. Read source code catch_cnd() explain works.A4. Let’s look source code catch_cnd():mentioned function docs:small wrapper around tryCatch() captures condition signalled evaluating argument.classes argument allows character vector condition classes catch, complex tidy evaluation generates necessary condition (; otherwise NULL).Q5. rewrite show_condition() use single handler?A5. source code rlang::catch_cond() gives us clue can .Conditions also class attribute, can use determine handler match condition.Let’s try new version examples used original version:","code":"\ncatch_cnd(stop(\"An error\"))\ncatch_cnd(abort(\"An error\"))\nlibrary(rlang)\n\nstopInfo <- catch_cnd(stop(\"An error\"))\nabortInfo <- catch_cnd(abort(\"An error\"))\n\nstr(stopInfo)\n#> List of 2\n#>  $ message: chr \"An error\"\n#>  $ call   : language force(expr)\n#>  - attr(*, \"class\")= chr [1:3] \"simpleError\" \"error\" \"condition\"\n\nstr(abortInfo)\n#> List of 5\n#>  $ message: chr \"An error\"\n#>  $ trace  :Classes 'rlang_trace', 'rlib_trace', 'tbl' and 'data.frame':  8 obs. of  6 variables:\n#>   ..$ call       :List of 8\n#>   .. ..$ : language catch_cnd(abort(\"An error\"))\n#>   .. ..$ : language eval_bare(rlang::expr(tryCatch(!!!handlers, {     force(expr) ...\n#>   .. ..$ : language tryCatch(condition = `<fn>`, {     force(expr) ...\n#>   .. ..$ : language tryCatchList(expr, classes, parentenv, handlers)\n#>   .. ..$ : language tryCatchOne(expr, names, parentenv, handlers[[1L]])\n#>   .. ..$ : language doTryCatch(return(expr), name, parentenv, handler)\n#>   .. ..$ : language force(expr)\n#>   .. ..$ : language abort(\"An error\")\n#>   ..$ parent     : int [1:8] 0 1 1 3 4 5 1 0\n#>   ..$ visible    : logi [1:8] FALSE FALSE FALSE FALSE FALSE FALSE ...\n#>   ..$ namespace  : chr [1:8] \"rlang\" \"rlang\" \"base\" \"base\" ...\n#>   ..$ scope      : chr [1:8] \"::\" \"::\" \"::\" \"local\" ...\n#>   ..$ error_frame: logi [1:8] FALSE FALSE FALSE FALSE FALSE FALSE ...\n#>   ..- attr(*, \"version\")= int 2\n#>  $ parent : NULL\n#>  $ rlang  :List of 1\n#>   ..$ inherit: logi TRUE\n#>  $ call   : NULL\n#>  - attr(*, \"class\")= chr [1:3] \"rlang_error\" \"error\" \"condition\"\nshow_condition <- function(code) {\n  tryCatch(\n    error = function(cnd) \"error\",\n    warning = function(cnd) \"warning\",\n    message = function(cnd) \"message\",\n    {\n      code\n      NULL\n    }\n  )\n}\n\nshow_condition(stop(\"!\"))\nshow_condition(10)\nshow_condition(warning(\"?!\"))\nshow_condition({\n  10\n  message(\"?\")\n  warning(\"?!\")\n})\nshow_condition <- function(code) {\n  tryCatch(\n    error = function(cnd) \"error\",\n    warning = function(cnd) \"warning\",\n    message = function(cnd) \"message\",\n    {\n      code\n      NULL\n    }\n  )\n}\n\nshow_condition(stop(\"!\"))\n#> [1] \"error\"\nshow_condition(10)\n#> NULL\nshow_condition(warning(\"?!\"))\n#> [1] \"warning\"\nshow_condition({\n  10\n  message(\"?\")\n  warning(\"?!\")\n})\n#> [1] \"message\"\nwithCallingHandlers(\n  message = function(cnd) message(\"b\"),\n  withCallingHandlers(\n    message = function(cnd) message(\"a\"),\n    message(\"c\")\n  )\n)\n#> b\n#> a\n#> b\n#> c\nrlang::catch_cnd\n#> function (expr, classes = \"condition\") \n#> {\n#>     stopifnot(is_character(classes))\n#>     handlers <- rep_named(classes, list(identity))\n#>     eval_bare(rlang::expr(tryCatch(!!!handlers, {\n#>         force(expr)\n#>         return(NULL)\n#>     })))\n#> }\n#> <bytecode: 0x55f2c19fa068>\n#> <environment: namespace:rlang>\ncatch_cnd(10)\n#> NULL\n\ncatch_cnd(abort(message = \"an error\", class = \"class1\"))\n#> <error/class1>\n#> Error:\n#> ! an error\n#> ---\n#> Backtrace:\n#> ▆\nshow_condition2 <- function(code) {\n  tryCatch(\n    condition = function(cnd) {\n      if (inherits(cnd, \"error\")) {\n        return(\"error\")\n      }\n      if (inherits(cnd, \"warning\")) {\n        return(\"warning\")\n      }\n      if (inherits(cnd, \"message\")) {\n        return(\"message\")\n      }\n    },\n    {\n      code\n      NULL\n    }\n  )\n}\nshow_condition2(stop(\"!\"))\n#> [1] \"error\"\nshow_condition2(10)\n#> NULL\nshow_condition2(warning(\"?!\"))\n#> [1] \"warning\"\nshow_condition2({\n  10\n  message(\"?\")\n  warning(\"?!\")\n})\n#> [1] \"message\""},{"path":"conditions.html","id":"custom-conditions-exercises-8.5.4","chapter":"8 Conditions","heading":"8.3 Custom conditions (Exercises 8.5.4)","text":"Q1. Inside package, ’s occasionally useful check package installed using . Write function checks package installed (requireNamespace(\"pkg\", quietly = FALSE)) , throws custom condition includes package name metadata.A1. desired function:reference, also see source code following functions:rlang::is_installed()insight::check_if_installed()Q2. Inside package often need stop error something right. packages depend package might tempted check errors unit tests. help packages avoid relying error message part user interface rather API might change without notice?A2. example, let’s say another package developer wanted use check_if_pkg_installed() function just wrote.developer using package can write unit test like :dissuade developers rely error messages check errors, can instead provide custom condition, can used unit testing instead:unit test :test wouldn’t fail even decided change exact message.","code":"\nabort_missing_package <- function(pkg) {\n  msg <- glue::glue(\"Problem loading `{pkg}` package, which is missing and must be installed.\")\n\n  abort(\"error_missing_package\",\n    message = msg,\n    pkg = pkg\n  )\n}\n\ncheck_if_pkg_installed <- function(pkg) {\n  if (!requireNamespace(pkg, quietly = TRUE)) {\n    abort_missing_package(pkg)\n  }\n\n  TRUE\n}\n\ncheck_if_pkg_installed(\"xyz123\")\n#> Error in `abort_missing_package()`:\n#> ! Problem loading `xyz123` package, which is missing and must be installed.\ncheck_if_pkg_installed(\"dplyr\")\n#> [1] TRUE\nexpect_error(\n  check_if_pkg_installed(\"xyz123\"),\n  \"Problem loading `xyz123` package, which is missing and must be installed.\"\n)\ne <- catch_cnd(check_if_pkg_installed(\"xyz123\"))\n\ninherits(e, \"error_missing_package\")\n#> [1] TRUE\nexpect_s3_class(e, \"error_missing_package\")"},{"path":"conditions.html","id":"applications-exercises-8.6.6","chapter":"8 Conditions","heading":"8.4 Applications (Exercises 8.6.6)","text":"Q1. Create suppressConditions() works like suppressMessages() suppressWarnings() suppresses everything. Think carefully handle errors.A1. create desired suppressConditions(), just need create equivalent suppressWarnings() suppressMessages() errors. suppress error message, can handle errors within tryCatch() return error object invisibly:Let’s try see works expected:condition messages now suppressed, note assign error object variable, can still extract useful information debugging:Q2. Compare following two implementations message2error(). main advantage withCallingHandlers() scenario? (Hint: look carefully traceback.)A2. withCallingHandlers(), condition handler called signaling function , , therefore, provides detailed call stack.tryCatch(), signalling function terminates condition raised, doesn’t provide detailed call stack.Q3. modify catch_cnds() definition wanted recreate original intermingling warnings messages?A3. Actually, won’t modify anything function defined chapter, since supports box.nothing additional here5! 😅Q4. catching interrupts dangerous? Run code find .A4. function catches interrupt way stop bottles_of_beer(), way usually stop using interrupt!RStudio IDE, can snap loop terminating R session.shows catching interrupt dangerous can result poor user experience.","code":"\nsuppressErrors <- function(expr) {\n  tryCatch(\n    error = function(cnd) invisible(cnd),\n    expr\n  )\n}\n\nsuppressConditions <- function(expr) {\n  suppressErrors(suppressWarnings(suppressMessages(expr)))\n}\nsuppressConditions(1)\n#> [1] 1\n\nsuppressConditions({\n  message(\"I'm messaging you\")\n  warning(\"I'm warning you\")\n})\n\nsuppressConditions({\n  stop(\"I'm stopping this\")\n})\ne <- suppressConditions({\n  stop(\"I'm stopping this\")\n})\n\ne\n#> <simpleError in withCallingHandlers(expr, message = function(c) if (inherits(c,     classes)) tryInvokeRestart(\"muffleMessage\")): I'm stopping this>\nmessage2error <- function(code) {\n  withCallingHandlers(code, message = function(e) stop(e))\n}\nmessage2error <- function(code) {\n  tryCatch(code, message = function(e) stop(e))\n}\nmessage2error1 <- function(code) {\n  withCallingHandlers(code, message = function(e) stop(\"error\"))\n}\n\nmessage2error1({\n  1\n  message(\"hidden error\")\n  NULL\n})\n#> Error in (function (e) : error\n\ntraceback()\n#> 9: stop(\"error\") at #2\n#> 8: (function (e)\n#>    stop(\"error\"))(list(message = \"hidden error\\n\",\n#>      call = message(\"hidden error\")))\n#> 7: signalCondition(cond)\n#> 6: doWithOneRestart(return(expr), restart)\n#> 5: withOneRestart(expr, restarts[[1L]])\n#> 4: withRestarts({\n#>        signalCondition(cond)\n#>        defaultHandler(cond)\n#>    }, muffleMessage = function() NULL)\n#> 3: message(\"hidden error\") at #1\n#> 2: withCallingHandlers(code,\n#>      message = function(e) stop(\"error\")) at #2\n#> 1: message2error1({\n#>        1\n#>        message(\"hidden error\")\n#>        NULL\n#>    })\nmessage2error2 <- function(code) {\n  tryCatch(code, message = function(e) (stop(\"error\")))\n}\n\nmessage2error2({\n  1\n  stop(\"hidden error\")\n  NULL\n})\n#> Error in value[[3L]](cond) : error\n\ntraceback()\n#> 6: stop(\"error\") at #2\n#> 5: value[[3L]](cond)\n#> 4: tryCatchOne(expr, names, parentenv, handlers[[1L]])\n#> 3: tryCatchList(expr, classes, parentenv, handlers)\n#> 2: tryCatch(code, message = function(e) (stop(\"error\"))) at #2\n#> 1: message2error2({\n#>        1\n#>        message(\"hidden error\")\n#>        NULL\n#>    })\ncatch_cnds <- function(expr) {\n  conds <- list()\n  add_cond <- function(cnd) {\n    conds <<- append(conds, list(cnd))\n    cnd_muffle(cnd)\n  }\n\n  withCallingHandlers(\n    message = add_cond,\n    warning = add_cond,\n    expr\n  )\n\n  conds\n}\n\ncatch_cnds({\n  inform(\"a\")\n  warn(\"b\")\n  inform(\"c\")\n})\n#> [[1]]\n#> <message/rlang_message>\n#> Message:\n#> a\n#> \n#> [[2]]\n#> <warning/rlang_warning>\n#> Warning:\n#> b\n#> \n#> [[3]]\n#> <message/rlang_message>\n#> Message:\n#> c\nbottles_of_beer <- function(i = 99) {\n  message(\n    \"There are \", i, \" bottles of beer on the wall, \",\n    i, \" bottles of beer.\"\n  )\n  while (i > 0) {\n    tryCatch(\n      Sys.sleep(1),\n      interrupt = function(err) {\n        i <<- i - 1\n        if (i > 0) {\n          message(\n            \"Take one down, pass it around, \", i,\n            \" bottle\", if (i > 1) \"s\", \" of beer on the wall.\"\n          )\n        }\n      }\n    )\n  }\n  message(\n    \"No more bottles of beer on the wall, \",\n    \"no more bottles of beer.\"\n  )\n}\nbottles_of_beer()\n#> There are 99 bottles of beer on the wall, 99 bottles of beer.\n#> Take one down, pass it around, 98 bottles of beer on the wall.\n#> Take one down, pass it around, 97 bottles of beer on the wall.\n#> Take one down, pass it around, 96 bottles of beer on the wall.\n#> Take one down, pass it around, 95 bottles of beer on the wall.\n#> Take one down, pass it around, 94 bottles of beer on the wall.\n#> Take one down, pass it around, 93 bottles of beer on the wall.\n#> Take one down, pass it around, 92 bottles of beer on the wall.\n#> Take one down, pass it around, 91 bottles of beer on the wall.\n#> ..."},{"path":"conditions.html","id":"session-information-6","chapter":"8 Conditions","heading":"8.5 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  brio          1.1.4   2023-12-10 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  desc          1.4.3   2023-12-10 [1] RSPM\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr         1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  pkgload       1.3.3   2023-09-22 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  rprojroot     2.0.4   2023-11-05 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  testthat    * 3.2.1   2023-12-02 [1] RSPM\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"functionals.html","id":"functionals","chapter":"9 Functionals","heading":"9 Functionals","text":"Attaching needed libraries:","code":"\nlibrary(purrr, warn.conflicts = FALSE)"},{"path":"functionals.html","id":"my-first-functional-map-exercises-9.2.6","chapter":"9 Functionals","heading":"9.1 My first functional: map() (Exercises 9.2.6)","text":"Q1. Use as_mapper() explore purrr generates anonymous functions integer, character, list helpers. helper allows extract attributes? Read documentation find .A1. Let’s handle two parts question separately.as_mapper() purrr-generated anonymous functions:Looking experimentation map() as_mapper(), can see , depending type input, as_mapper() creates extractor function using pluck().can extract attributes using purrr::attr_getter():Q2. map(1:3, ~ runif(2)) useful pattern generating random numbers, map(1:3, runif(2)) . ? Can explain returns result ?A2. shown as_mapper() outputs , second call appropriate generating random numbers translates pluck() function indices plucking taken randomly generated numbers, valid accessors get NULLs return.Q3. Use appropriate map() function :A3. Using appropriate map() function :Compute standard deviation every column numeric data frame:Compute standard deviation every numeric column mixed data frame:Compute number levels every factor data frame:Q4. following code simulates performance t-test non-normal data. Extract p-value test, visualise.A4.Extract p-value test:Visualise extracted p-values:Q5. following code uses map nested inside another map apply function every element nested list. fail, need make work?A5. function fails call effectively evaluates following:triple() one parameter (x), execution fails.fixed version:Q6. Use map() fit linear models mtcars dataset using formulas stored list:A6. Fitting linear models mtcars dataset using provided formulas:Q7. Fit model mpg ~ disp bootstrap replicates mtcars list , extract \\(R^2\\) model fit (Hint: can compute \\(R^2\\) summary().)A7. can done using map_dbl():","code":"\n# mapping by position -----------------------\n\nx <- list(1, list(2, 3, list(1, 2)))\n\nmap(x, 1)\n#> [[1]]\n#> [1] 1\n#> \n#> [[2]]\n#> [1] 2\nas_mapper(1)\n#> function (x, ...) \n#> pluck_raw(x, list(1), .default = NULL)\n#> <environment: 0x55d958cd1430>\n\nmap(x, list(2, 1))\n#> [[1]]\n#> NULL\n#> \n#> [[2]]\n#> [1] 3\nas_mapper(list(2, 1))\n#> function (x, ...) \n#> pluck_raw(x, list(2, 1), .default = NULL)\n#> <environment: 0x55d958db8158>\n\n# mapping by name -----------------------\n\ny <- list(\n  list(m = \"a\", list(1, m = \"mo\")),\n  list(n = \"b\", list(2, n = \"no\"))\n)\n\nmap(y, \"m\")\n#> [[1]]\n#> [1] \"a\"\n#> \n#> [[2]]\n#> NULL\nas_mapper(\"m\")\n#> function (x, ...) \n#> pluck_raw(x, list(\"m\"), .default = NULL)\n#> <environment: 0x55d958f06330>\n\n# mixing position and name\nmap(y, list(2, \"m\"))\n#> [[1]]\n#> [1] \"mo\"\n#> \n#> [[2]]\n#> NULL\nas_mapper(list(2, \"m\"))\n#> function (x, ...) \n#> pluck_raw(x, list(2, \"m\"), .default = NULL)\n#> <environment: 0x55d958fec178>\n\n# compact functions ----------------------------\n\nmap(y, ~ length(.x))\n#> [[1]]\n#> [1] 2\n#> \n#> [[2]]\n#> [1] 2\nas_mapper(~ length(.x))\n#> <lambda>\n#> function (..., .x = ..1, .y = ..2, . = ..1) \n#> length(.x)\n#> attr(,\"class\")\n#> [1] \"rlang_lambda_function\" \"function\"\npluck(Titanic, attr_getter(\"class\"))\n#> [1] \"table\"\nmap(1:3, ~ runif(2))\n#> [[1]]\n#> [1] 0.2180892 0.9876342\n#> \n#> [[2]]\n#> [1] 0.3484619 0.3810470\n#> \n#> [[3]]\n#> [1] 0.02098596 0.74972687\nas_mapper(~ runif(2))\n#> <lambda>\n#> function (..., .x = ..1, .y = ..2, . = ..1) \n#> runif(2)\n#> attr(,\"class\")\n#> [1] \"rlang_lambda_function\" \"function\"\n\nmap(1:3, runif(2))\n#> [[1]]\n#> [1] 1\n#> \n#> [[2]]\n#> [1] 2\n#> \n#> [[3]]\n#> [1] 3\nas_mapper(runif(2))\n#> function (x, ...) \n#> pluck_raw(x, list(0.597890264587477, 0.587997315218672), .default = NULL)\n#> <environment: 0x55d959f7d6f8>a) Compute the standard deviation of every column in a numeric data frame.\n\na) Compute the standard deviation of every numeric column in a mixed data frame. (Hint: you'll need to do it in two steps.)\n\na) Compute the number of levels for every factor in a data frame.\nmap_dbl(mtcars, sd)\n#>         mpg         cyl        disp          hp        drat \n#>   6.0269481   1.7859216 123.9386938  68.5628685   0.5346787 \n#>          wt        qsec          vs          am        gear \n#>   0.9784574   1.7869432   0.5040161   0.4989909   0.7378041 \n#>        carb \n#>   1.6152000\nkeep(iris, is.numeric) %>%\n  map_dbl(sd)\n#> Sepal.Length  Sepal.Width Petal.Length  Petal.Width \n#>    0.8280661    0.4358663    1.7652982    0.7622377\nmodify_if(dplyr::starwars, is.character, as.factor) %>%\n  keep(is.factor) %>%\n  map_int(~ length(levels(.)))\n#>       name hair_color skin_color  eye_color        sex \n#>         87         11         31         15          4 \n#>     gender  homeworld    species \n#>          2         48         37\ntrials <- map(1:100, ~ t.test(rpois(10, 10), rpois(7, 10)))\ntrials <- map(1:100, ~ t.test(rpois(10, 10), rpois(7, 10)))\n\n(p <- map_dbl(trials, \"p.value\"))\n#>   [1] 0.81695628 0.53177360 0.94750819 0.41026769 0.34655294\n#>   [6] 0.05300287 0.56479901 0.85936864 0.77517391 0.64321161\n#>  [11] 0.84462914 0.54144946 0.63070476 0.20325827 0.39824435\n#>  [16] 0.67052432 0.39932663 0.44437632 0.51645941 0.96578745\n#>  [21] 0.70219557 0.69931716 0.23946786 0.55100566 0.76028958\n#>  [26] 0.38105366 0.64544126 0.15379307 0.86945196 0.09965658\n#>  [31] 0.96425489 0.54239108 0.38985789 0.59019282 0.96247907\n#>  [36] 0.54997487 0.66111391 0.30961551 0.10897334 0.55049635\n#>  [41] 0.93882405 0.14836866 0.44307287 0.61583610 0.37284284\n#>  [46] 0.38559622 0.42935767 0.26059293 0.07831619 0.93768396\n#>  [51] 0.48459268 0.73571291 0.30288560 0.68521609 0.06374636\n#>  [56] 0.11007808 0.98758443 0.17831882 0.94471538 0.19711729\n#>  [61] 0.02094185 0.12370745 0.23247837 0.93842382 0.19160550\n#>  [66] 0.49005550 0.98146240 0.09034183 0.94912080 0.55857523\n#>  [71] 0.24692070 0.63658206 0.14290966 0.10309770 0.89516449\n#>  [76] 0.25660092 0.16943034 0.41199780 0.82721280 0.74017418\n#>  [81] 0.43724631 0.55944024 0.93615100 0.68788872 0.01416627\n#>  [86] 0.60120497 0.54125910 0.91581929 0.78949327 0.57887371\n#>  [91] 0.83217542 0.90108906 0.97474727 0.99129282 0.54436155\n#>  [96] 0.74159859 0.06534957 0.10834529 0.19737786 0.93750342\nplot(p)\n\nhist(p)\nx <- list(\n  list(1, c(3, 9)),\n  list(c(3, 6), 7, c(4, 7, 6))\n)\n\ntriple <- function(x) x * 3\nmap(x, map, .f = triple)\n#> Error in `map()`:\n#> ℹ In index: 1.\n#> Caused by error in `.f()`:\n#> ! unused argument (function (.x, .f, ..., .progress = FALSE) \n#> {\n#>     map_(\"list\", .x, .f, ..., .progress = .progress)\n#> })\nmap(.x = x, .f = ~ triple(x = .x, map))\nx <- list(\n  list(1, c(3, 9)),\n  list(c(3, 6), 7, c(4, 7, 6))\n)\n\ntriple <- function(x) x * 3\nmap(x, .f = ~ map(.x, ~ triple(.x)))\n#> [[1]]\n#> [[1]][[1]]\n#> [1] 3\n#> \n#> [[1]][[2]]\n#> [1]  9 27\n#> \n#> \n#> [[2]]\n#> [[2]][[1]]\n#> [1]  9 18\n#> \n#> [[2]][[2]]\n#> [1] 21\n#> \n#> [[2]][[3]]\n#> [1] 12 21 18\nformulas <- list(\n  mpg ~ disp,\n  mpg ~ I(1 / disp),\n  mpg ~ disp + wt,\n  mpg ~ I(1 / disp) + wt\n)\nformulas <- list(\n  mpg ~ disp,\n  mpg ~ I(1 / disp),\n  mpg ~ disp + wt,\n  mpg ~ I(1 / disp) + wt\n)\n\nmap(formulas, ~ lm(formula = ., data = mtcars))\n#> [[1]]\n#> \n#> Call:\n#> lm(formula = ., data = mtcars)\n#> \n#> Coefficients:\n#> (Intercept)         disp  \n#>    29.59985     -0.04122  \n#> \n#> \n#> [[2]]\n#> \n#> Call:\n#> lm(formula = ., data = mtcars)\n#> \n#> Coefficients:\n#> (Intercept)    I(1/disp)  \n#>       10.75      1557.67  \n#> \n#> \n#> [[3]]\n#> \n#> Call:\n#> lm(formula = ., data = mtcars)\n#> \n#> Coefficients:\n#> (Intercept)         disp           wt  \n#>    34.96055     -0.01772     -3.35083  \n#> \n#> \n#> [[4]]\n#> \n#> Call:\n#> lm(formula = ., data = mtcars)\n#> \n#> Coefficients:\n#> (Intercept)    I(1/disp)           wt  \n#>      19.024     1142.560       -1.798\nbootstrap <- function(df) {\n  df[sample(nrow(df), replace = TRUE), , drop = FALSE]\n}\n\nbootstraps <- map(1:10, ~ bootstrap(mtcars))\nbootstrap <- function(df) {\n  df[sample(nrow(df), replace = TRUE), , drop = FALSE]\n}\n\nbootstraps <- map(1:10, ~ bootstrap(mtcars))\n\nbootstraps %>%\n  map(~ lm(mpg ~ disp, data = .x)) %>%\n  map(summary) %>%\n  map_dbl(\"r.squared\")\n#>  [1] 0.7864562 0.8110818 0.7956331 0.7632399 0.7967824\n#>  [6] 0.7364226 0.7203027 0.6653252 0.7732780 0.6753329"},{"path":"functionals.html","id":"map-variants-exercises-9.4.6","chapter":"9 Functionals","heading":"9.2 Map variants (Exercises 9.4.6)","text":"Q1. Explain results modify(mtcars, 1).A1. modify() returns object type input. Since input data frame certain dimensions .f = 1 translates plucking first element column, returns data frame dimensions plucked element recycled across rows.Q2. Rewrite following code use iwalk() instead walk2(). advantages disadvantages?A2. Let’s first rewrite provided code using iwalk():advantage using iwalk() need now deal single variable (cyls) instead two (cyls paths).disadvantage code difficult reason :\nwalk2(), ’s explicit .x (= cyls) .y (= paths) correspond , iwalk() (.e., .x = cyls .y = names(cyls)) .y argument “invisible”.Q3. Explain following code transforms data frame using functions stored list.Compare contrast map2() approach map() approach:A3. map2() supplies functions stored trans anonymous functions via placeholder f, names columns specified mtcars[nm] supplied var argument anonymous function. Note function iterating indices vectors transformations column names.map() approach, function iterating indices vectors column names.latter approach can’t afford passing arguments placeholders anonymous function.Q4. write.csv() return, .e. happens use map2() instead walk2()?A4. use map2(), work, print NULLs console every list element.","code":"\nhead(modify(mtcars, 1))\n#>   mpg cyl disp  hp drat   wt  qsec vs am gear carb\n#> 1  21   6  160 110  3.9 2.62 16.46  0  1    4    4\n#> 2  21   6  160 110  3.9 2.62 16.46  0  1    4    4\n#> 3  21   6  160 110  3.9 2.62 16.46  0  1    4    4\n#> 4  21   6  160 110  3.9 2.62 16.46  0  1    4    4\n#> 5  21   6  160 110  3.9 2.62 16.46  0  1    4    4\n#> 6  21   6  160 110  3.9 2.62 16.46  0  1    4    4\ncyls <- split(mtcars, mtcars$cyl)\npaths <- file.path(temp, paste0(\"cyl-\", names(cyls), \".csv\"))\nwalk2(cyls, paths, write.csv)\ncyls <- split(mtcars, mtcars$cyl)\nnames(cyls) <- file.path(temp, paste0(\"cyl-\", names(cyls), \".csv\"))\niwalk(cyls, ~ write.csv(.x, .y))\ntrans <- list(\n  disp = function(x) x * 0.0163871,\n  am = function(x) factor(x, labels = c(\"auto\", \"manual\"))\n)\n\nnm <- names(trans)\nmtcars[nm] <- map2(trans, mtcars[nm], function(f, var) f(var))\nmtcars[nm] <- map(nm, ~ trans[[.x]](mtcars[[.x]]))\ntrans <- list(\n  disp = function(x) x * 0.0163871,\n  am = function(x) factor(x, labels = c(\"auto\", \"manual\"))\n)\n\nnm <- names(trans)\nmtcars[nm] <- map2(trans, mtcars[nm], function(f, var) f(var))\nmtcars[nm] <- map(nm, ~ trans[[.x]](mtcars[[.x]]))\nwithr::with_tempdir(\n  code = {\n    ls <- split(mtcars, mtcars$cyl)\n    nm <- names(ls)\n    map2(ls, nm, write.csv)\n  }\n)\n#> $`4`\n#> NULL\n#> \n#> $`6`\n#> NULL\n#> \n#> $`8`\n#> NULL"},{"path":"functionals.html","id":"predicate-functionals-exercises-9.6.3","chapter":"9 Functionals","heading":"9.3 Predicate functionals (Exercises 9.6.3)","text":"Q1. isn’t .na() predicate function? base R function closest predicate version .na()?A1. mentioned docs:predicate function returns single TRUE FALSE..na() function return logical scalar, instead returns vector thus isn’t predicate function.closest equivalent predicate function base-R anyNA() function.Q2. simple_reduce() problem x length 0 length 1. Describe source problem might go fixing .A2. supplied function struggles inputs length 0 1 function tries subscript --bound values.problem can solved adding init argument, supplies default initial value:Let’s try :Depending function, can provide different init argument:Q3. Implement span() function Haskell: given list x predicate function f, span(x, f) returns location longest sequential run elements predicate true. (Hint: might find rle() helpful.)A3. Implementation span():Testing :Testing twice:Q4. Implement arg_max(). take function vector inputs, return elements input function returns highest value. example, arg_max(-10:5, function(x) x ^ 2) return -10. arg_max(-5:5, function(x) x ^ 2) return c(-5, 5). Also implement matching arg_min() function.A4. implementations specified functions:Implementing arg_max()Implementing arg_min()Q5. function scales vector falls range [0, 1]. apply every column data frame? apply every numeric column data frame?A5. use purrr package apply function. Key thing keep mind data frame list atomic vectors equal length.Applying function every column data frame: use anscombe example since numeric columns.Applying function every numeric column data frame: use iris example since columns numeric type.","code":"\n# contrast the following behavior of predicate functions\nis.character(c(\"x\", 2))\n#> [1] TRUE\nis.null(c(3, NULL))\n#> [1] FALSE\n\n# with this behavior\nis.na(c(NA, 1))\n#> [1]  TRUE FALSE\nanyNA(c(NA, 1))\n#> [1] TRUE\nsimple_reduce <- function(x, f) {\n  out <- x[[1]]\n  for (i in seq(2, length(x))) {\n    out <- f(out, x[[i]])\n  }\n  out\n}\nsimple_reduce(numeric(), sum)\n#> Error in x[[1]]: subscript out of bounds\nsimple_reduce(1, sum)\n#> Error in x[[i]]: subscript out of bounds\nsimple_reduce(1:3, sum)\n#> [1] 6\nsimple_reduce2 <- function(x, f, init = 0) {\n  # initializer will become the first value\n  if (length(x) == 0L) {\n    return(init)\n  }\n\n  if (length(x) == 1L) {\n    return(x[[1L]])\n  }\n\n  out <- x[[1]]\n\n  for (i in seq(2, length(x))) {\n    out <- f(out, x[[i]])\n  }\n\n  out\n}\nsimple_reduce2(numeric(), sum)\n#> [1] 0\nsimple_reduce2(1, sum)\n#> [1] 1\nsimple_reduce2(1:3, sum)\n#> [1] 6\nsimple_reduce2(numeric(), `*`, init = 1)\n#> [1] 1\nsimple_reduce2(1, `*`, init = 1)\n#> [1] 1\nsimple_reduce2(1:3, `*`, init = 1)\n#> [1] 6\nspan <- function(x, f) {\n  running_lengths <- purrr::map_lgl(x, ~ f(.x)) %>% rle()\n\n  df <- dplyr::tibble(\n    \"lengths\" = running_lengths$lengths,\n    \"values\" = running_lengths$values\n  ) %>%\n    dplyr::mutate(rowid = dplyr::row_number()) %>%\n    dplyr::filter(values)\n\n  # no sequence where condition is `TRUE`\n  if (nrow(df) == 0L) {\n    return(integer())\n  }\n\n  # only single sequence where condition is `TRUE`\n  if (nrow(df) == 1L) {\n    return((df$rowid):(df$lengths - 1 + df$rowid))\n  }\n\n  # multiple sequences where condition is `TRUE`; select max one\n  if (nrow(df) > 1L) {\n    df <- dplyr::filter(df, lengths == max(lengths))\n    return((df$rowid):(df$lengths - 1 + df$rowid))\n  }\n}\nspan(c(0, 0, 0, 0, 0), is.na)\n#> integer(0)\nspan(c(NA, 0, NA, NA, NA), is.na)\n#> [1] 3 4 5\nspan(c(NA, 0, 0, 0, 0), is.na)\n#> [1] 1\nspan(c(NA, NA, 0, 0, 0), is.na)\n#> [1] 1 2\nspan(c(3, 1, 2, 4, 5, 6), function(x) x > 3)\n#> [1] 2 3 4\nspan(c(3, 1, 2, 4, 5, 6), function(x) x > 9)\n#> integer(0)\nspan(c(3, 1, 2, 4, 5, 6), function(x) x == 3)\n#> [1] 1\nspan(c(3, 1, 2, 4, 5, 6), function(x) x %in% c(2, 4))\n#> [1] 2 3\narg_max <- function(.x, .f) {\n  df <- dplyr::tibble(\n    original = .x,\n    transformed = purrr::map_dbl(.x, .f)\n  )\n\n  dplyr::filter(df, transformed == max(transformed))[[\"original\"]]\n}\n\narg_max(-10:5, function(x) x^2)\n#> [1] -10\narg_max(-5:5, function(x) x^2)\n#> [1] -5  5\narg_min <- function(.x, .f) {\n  df <- dplyr::tibble(\n    original = .x,\n    transformed = purrr::map_dbl(.x, .f)\n  )\n\n  dplyr::filter(df, transformed == min(transformed))[[\"original\"]]\n}\n\narg_min(-10:5, function(x) x^2)\n#> [1] 0\narg_min(-5:5, function(x) x^2)\n#> [1] 0\nscale01 <- function(x) {\n  rng <- range(x, na.rm = TRUE)\n  (x - rng[1]) / (rng[2] - rng[1])\n}\npurrr::map_df(head(anscombe), .f = scale01)\n#> # A tibble: 6 × 8\n#>      x1    x2    x3    x4    y1     y2     y3    y4\n#>   <dbl> <dbl> <dbl> <dbl> <dbl>  <dbl>  <dbl> <dbl>\n#> 1 0.333 0.333 0.333   NaN 0.362 0.897  0.116  0.266\n#> 2 0     0     0       NaN 0     0.0345 0      0    \n#> 3 0.833 0.833 0.833   NaN 0.209 0.552  1      0.633\n#> 4 0.167 0.167 0.167   NaN 0.618 0.578  0.0570 1    \n#> 5 0.5   0.5   0.5     NaN 0.458 1      0.174  0.880\n#> 6 1     1     1       NaN 1     0      0.347  0.416\npurrr::modify_if(head(iris), .p = is.numeric, .f = scale01)\n#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species\n#> 1        0.625   0.5555556         0.25           0  setosa\n#> 2        0.375   0.0000000         0.25           0  setosa\n#> 3        0.125   0.2222222         0.00           0  setosa\n#> 4        0.000   0.1111111         0.50           0  setosa\n#> 5        0.500   0.6666667         0.25           0  setosa\n#> 6        1.000   1.0000000         1.00           1  setosa"},{"path":"functionals.html","id":"base-functionals-exercises-9.7.3","chapter":"9 Functionals","heading":"9.4 Base functionals (Exercises 9.7.3)","text":"Q1. apply() arrange output? Read documentation perform experiments.A1. Let’s prepare array apply function different margins:can seen, apply() returns outputs organised first margins operated , results.Q2. eapply() rapply() ? purrr equivalents?A2. Let’s consider one--one.eapply()mentioned documentation:eapply() applies FUN named values environment returns results list.example:purrr doesn’t function iterate environments.rapply()rapply() recursive version lapply flexibility result structured (= “..”).example:purrr something similar modify_tree().Q3. Challenge: read fixed point algorithm. Complete exercises using R.A3. mentioned suggested reading material:number \\(x\\) called fixed point function \\(f\\) \\(x\\) satisfies equation \\(f(x) = x\\). functions \\(f\\) can locate fixed point beginning initial guess applying \\(f\\) repeatedly, \\(f(x), f(f(x)), f(f(f(x))), ...\\) value change much. Using idea, can devise procedure fixed-point takes inputs function initial guess produces approximation fixed point function.Let’s first implement fixed-point algorithm:Let’s check works expected:solve one exercise reading material. Rest beyond scope solution manual.Show golden ratio \\(\\phi\\) fixed point transformation \\(x \\mapsto 1 + 1/x\\), use fact compute \\(\\phi\\) means fixed-point procedure.","code":"\n(m <- as.array(table(mtcars$cyl, mtcars$am, mtcars$vs)))\n#> , ,  = 0\n#> \n#>    \n#>     auto manual\n#>   4    0      1\n#>   6    0      3\n#>   8   12      2\n#> \n#> , ,  = 1\n#> \n#>    \n#>     auto manual\n#>   4    3      7\n#>   6    4      0\n#>   8    0      0\n\n# rows\napply(m, 1, function(x) x^2)\n#>       \n#>         4  6   8\n#>   [1,]  0  0 144\n#>   [2,]  1  9   4\n#>   [3,]  9 16   0\n#>   [4,] 49  0   0\n\n# columns\napply(m, 2, function(x) x^2)\n#>       \n#>        auto manual\n#>   [1,]    0      1\n#>   [2,]    0      9\n#>   [3,]  144      4\n#>   [4,]    9     49\n#>   [5,]   16      0\n#>   [6,]    0      0\n\n# rows and columns\napply(m, c(1, 2), function(x) x^2)\n#> , ,  = auto\n#> \n#>    \n#>     4  6   8\n#>   0 0  0 144\n#>   1 9 16   0\n#> \n#> , ,  = manual\n#> \n#>    \n#>      4 6 8\n#>   0  1 9 4\n#>   1 49 0 0\nlibrary(rlang)\n#> \n#> Attaching package: 'rlang'\n#> The following objects are masked from 'package:purrr':\n#> \n#>     %@%, flatten, flatten_chr, flatten_dbl,\n#>     flatten_int, flatten_lgl, flatten_raw, invoke,\n#>     splice\n#> The following object is masked from 'package:magrittr':\n#> \n#>     set_names\n\ne <- env(\"x\" = 1, \"y\" = 2)\nrlang::env_print(e)\n#> <environment: 0x55d9572060a8>\n#> Parent: <environment: global>\n#> Bindings:\n#> • x: <dbl>\n#> • y: <dbl>\n\neapply(e, as.character)\n#> $x\n#> [1] \"1\"\n#> \n#> $y\n#> [1] \"2\"\nX <- list(list(a = TRUE, b = list(c = c(4L, 3.2))), d = 9.0)\n\nrapply(X, as.character, classes = \"numeric\", how = \"replace\")\n#> [[1]]\n#> [[1]]$a\n#> [1] TRUE\n#> \n#> [[1]]$b\n#> [[1]]$b$c\n#> [1] \"4\"   \"3.2\"\n#> \n#> \n#> \n#> $d\n#> [1] \"9\"\nX <- list(list(a = TRUE, b = list(c = c(4L, 3.2))), d = 9.0)\n\npurrr::modify_tree(X, leaf = length)\n#> [[1]]\n#> [[1]]$a\n#> [1] 1\n#> \n#> [[1]]$b\n#> [[1]]$b$c\n#> [1] 2\n#> \n#> \n#> \n#> $d\n#> [1] 1\nclose_enough <- function(x1, x2, tolerance = 0.001) {\n  if (abs(x1 - x2) < tolerance) {\n    return(TRUE)\n  } else {\n    return(FALSE)\n  }\n}\n\nfind_fixed_point <- function(.f, .guess, tolerance = 0.001) {\n  .next <- .f(.guess)\n  is_close_enough <- close_enough(.next, .guess, tol = tolerance)\n\n  if (is_close_enough) {\n    return(.next)\n  } else {\n    find_fixed_point(.f, .next, tolerance)\n  }\n}\nfind_fixed_point(cos, 1.0)\n#> [1] 0.7387603\n\n# cos(x) = x\ncos(find_fixed_point(cos, 1.0))\n#> [1] 0.7393039\ngolden_ratio_f <- function(x) 1 + (1 / x)\n\nfind_fixed_point(golden_ratio_f, 1.0)\n#> [1] 1.618182"},{"path":"functionals.html","id":"session-information-7","chapter":"9 Functionals","heading":"9.5 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr         1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  highr         0.10    2022-12-22 [1] RSPM\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  purrr       * 1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"function-factories.html","id":"function-factories","chapter":"10 Function factories","heading":"10 Function factories","text":"Attaching needed libraries:","code":"\nlibrary(rlang, warn.conflicts = FALSE)\nlibrary(ggplot2, warn.conflicts = FALSE)"},{"path":"function-factories.html","id":"factory-fundamentals-exercises-10.2.6","chapter":"10 Function factories","heading":"10.1 Factory fundamentals (Exercises 10.2.6)","text":"Q1. definition force() simple:better force(x) instead just x?A1. Due lazy evaluation, argument function won’t evaluated value needed. sometimes may want eager evaluation, using force() makes intent clearer.Q2. Base R contains two function factories, approxfun() ecdf(). Read documentation experiment figure functions return.A2. two function factories-approxfun()function factory returns function performing linear (constant) interpolation.ecdf()function factory computes empirical cumulative distribution function.Q3. Create function pick() takes index, , argument returns function argument x subsets x .A3. write desired function, just need make sure argument eagerly evaluated.Testing specified test cases:Q4. Create function creates functions compute ithcentral moment numeric vector. can test running following code:A4. following function satisfied specified requirements:Testing specified test cases:Q5. happens don’t use closure? Make predictions, verify code .A5. case closures used context, counts stored global variable, can modified processes even deleted.Q6. happens use <- instead <<-? Make predictions, verify code .A6. case, function always return 1.","code":"\nforce\n#> function (x) \n#> x\n#> <bytecode: 0x56479c5e3ab0>\n#> <environment: namespace:base>\nx <- 1:10\ny <- rnorm(10)\nf <- approxfun(x, y)\nf\n#> function (v) \n#> .approxfun(x, y, v, method, yleft, yright, f, na.rm)\n#> <bytecode: 0x5647a0854cb0>\n#> <environment: 0x5647a0853ba0>\nf(x)\n#>  [1] -0.7786629 -0.3894764 -2.0337983 -0.9823731  0.2478901\n#>  [6] -2.1038646 -0.3814180  2.0749198  1.0271384  0.4730142\ncurve(f(x), 0, 11)\nx <- rnorm(12)\nf <- ecdf(x)\nf\n#> Empirical CDF \n#> Call: ecdf(x)\n#>  x[1:12] = -1.8793, -1.3221, -1.2392,  ..., 1.1604, 1.7956\nf(seq(-2, 2, by = 0.1))\n#>  [1] 0.00000000 0.00000000 0.08333333 0.08333333 0.08333333\n#>  [6] 0.08333333 0.08333333 0.16666667 0.25000000 0.25000000\n#> [11] 0.33333333 0.33333333 0.33333333 0.41666667 0.41666667\n#> [16] 0.41666667 0.41666667 0.50000000 0.58333333 0.58333333\n#> [21] 0.66666667 0.75000000 0.75000000 0.75000000 0.75000000\n#> [26] 0.75000000 0.75000000 0.75000000 0.75000000 0.83333333\n#> [31] 0.83333333 0.83333333 0.91666667 0.91666667 0.91666667\n#> [36] 0.91666667 0.91666667 0.91666667 1.00000000 1.00000000\n#> [41] 1.00000000\npick(1)(x)\n# should be equivalent to\nx[[1]]\n\nlapply(mtcars, pick(5))\n# should be equivalent to\nlapply(mtcars, function(x) x[[5]])\npick <- function(i) {\n  force(i)\n  function(x) x[[i]]\n}\nx <- list(\"a\", \"b\", \"c\")\nidentical(x[[1]], pick(1)(x))\n#> [1] TRUE\n\nidentical(\n  lapply(mtcars, pick(5)),\n  lapply(mtcars, function(x) x[[5]])\n)\n#> [1] TRUE\nm1 <- moment(1)\nm2 <- moment(2)\nx <- runif(100)\nstopifnot(all.equal(m1(x), 0))\nstopifnot(all.equal(m2(x), var(x) * 99 / 100))\nmoment <- function(k) {\n  force(k)\n\n  function(x) (sum((x - mean(x))^k)) / length(x)\n}\nm1 <- moment(1)\nm2 <- moment(2)\nx <- runif(100)\n\nstopifnot(all.equal(m1(x), 0))\nstopifnot(all.equal(m2(x), var(x) * 99 / 100))\ni <- 0\nnew_counter2 <- function() {\n  i <<- i + 1\n  i\n}\nnew_counter2()\n#> [1] 1\n\nnew_counter2()\n#> [1] 2\n\nnew_counter2()\n#> [1] 3\n\ni <- 20\nnew_counter2()\n#> [1] 21\nnew_counter3 <- function() {\n  i <- 0\n  function() {\n    i <- i + 1\n    i\n  }\n}\nnew_counter3()\n#> function() {\n#>     i <- i + 1\n#>     i\n#>   }\n#> <environment: 0x5647a09f9e68>\n\nnew_counter3()\n#> function() {\n#>     i <- i + 1\n#>     i\n#>   }\n#> <bytecode: 0x5647a0bbe600>\n#> <environment: 0x5647a0a71c80>"},{"path":"function-factories.html","id":"graphical-factories-exercises-10.3.4","chapter":"10 Function factories","heading":"10.2 Graphical factories (Exercises 10.3.4)","text":"Q1. Compare contrast ggplot2::label_bquote() scales::number_format().A1. compare contrast, let’s first look source code functions:ggplot2::label_bquote()scales::number_format()functions return formatting functions used style facets labels labels desired format ggplot2 plots.example, using plotmath expression facet label:display axes labels desired format:ggplot2::label_bquote() adds additional class returned function.scales::number_format() function simple pass-method forces evaluation parameters passes underlying scales::number() function.","code":"\nggplot2::label_bquote\n#> function (rows = NULL, cols = NULL, default) \n#> {\n#>     cols_quoted <- substitute(cols)\n#>     rows_quoted <- substitute(rows)\n#>     call_env <- env_parent()\n#>     fun <- function(labels) {\n#>         quoted <- resolve_labeller(rows_quoted, cols_quoted, \n#>             labels)\n#>         if (is.null(quoted)) {\n#>             return(label_value(labels))\n#>         }\n#>         evaluate <- function(...) {\n#>             params <- list(...)\n#>             params <- as_environment(params, call_env)\n#>             eval(substitute(bquote(expr, params), list(expr = quoted)))\n#>         }\n#>         list(inject(mapply(evaluate, !!!labels, SIMPLIFY = FALSE)))\n#>     }\n#>     structure(fun, class = \"labeller\")\n#> }\n#> <bytecode: 0x5647a0d500b0>\n#> <environment: namespace:ggplot2>\nscales::number_format\n#> function (accuracy = NULL, scale = 1, prefix = \"\", suffix = \"\", \n#>     big.mark = \" \", decimal.mark = \".\", style_positive = c(\"none\", \n#>         \"plus\", \"space\"), style_negative = c(\"hyphen\", \"minus\", \n#>         \"parens\"), scale_cut = NULL, trim = TRUE, ...) \n#> {\n#>     force_all(accuracy, scale, prefix, suffix, big.mark, decimal.mark, \n#>         style_positive, style_negative, scale_cut, trim, ...)\n#>     function(x) {\n#>         number(x, accuracy = accuracy, scale = scale, prefix = prefix, \n#>             suffix = suffix, big.mark = big.mark, decimal.mark = decimal.mark, \n#>             style_positive = style_positive, style_negative = style_negative, \n#>             scale_cut = scale_cut, trim = trim, ...)\n#>     }\n#> }\n#> <bytecode: 0x5647a0fa0bc8>\n#> <environment: namespace:scales>\nlibrary(ggplot2)\n\np <- ggplot(mtcars, aes(wt, mpg)) +\n  geom_point()\np + facet_grid(. ~ vs, labeller = label_bquote(cols = alpha^.(vs)))\nlibrary(scales)\n\nggplot(mtcars, aes(wt, mpg)) +\n  geom_point() +\n  scale_y_continuous(labels = number_format(accuracy = 0.01, decimal.mark = \",\"))"},{"path":"function-factories.html","id":"statistical-factories-exercises-10.4.4","chapter":"10 Function factories","heading":"10.3 Statistical factories (Exercises 10.4.4)","text":"Q1. boot_model(), don’t need force evaluation df model?A1. don’t need force evaluation df model arguments automatically evaluated lm():Q2. might formulate Box-Cox transformation like ?A2. see formulate transformation like , can compare one mentioned book:Let’s look one example :can seen:boxcox2(), can vary x value lambda, whilein boxcox3(), can vary lambda vector.Thus, boxcox3() can handy exploring different transformations across inputs.Q3. don’t need worry boot_permute() stores copy data inside function generates?A3. look source code generated function factory, notice exact data frame (mtcars) referenced:don’t need worry copy made df function environment points memory address data frame. can confirm comparing memory addresses:can also check values bindings entered function factory:Q4. much time ll_poisson2() save compared ll_poisson1()? Use bench::mark() see much faster optimisation occurs. changing length x change results?A4. Let’s first compare performance functions example book:can seen, second version much faster first version.can also vary length vector confirm across wide range vector lengths, performance advantage observed.","code":"\nboot_model <- function(df, formula) {\n  mod <- lm(formula, data = df)\n  fitted <- unname(fitted(mod))\n  resid <- unname(resid(mod))\n  rm(mod)\n\n  function() {\n    fitted + sample(resid)\n  }\n}\nboxcox3 <- function(x) {\n  function(lambda) {\n    if (lambda == 0) {\n      log(x)\n    } else {\n      (x^lambda - 1) / lambda\n    }\n  }\n}\nboxcox2 <- function(lambda) {\n  if (lambda == 0) {\n    function(x) log(x)\n  } else {\n    function(x) (x^lambda - 1) / lambda\n  }\n}\nboxcox2(1)\n#> function(x) (x^lambda - 1) / lambda\n#> <environment: 0x56479c998358>\n\nboxcox3(mtcars$wt)\n#> function(lambda) {\n#>     if (lambda == 0) {\n#>       log(x)\n#>     } else {\n#>       (x^lambda - 1) / lambda\n#>     }\n#>   }\n#> <environment: 0x56479de7f188>\nboot_permute <- function(df, var) {\n  n <- nrow(df)\n  force(var)\n\n  function() {\n    col <- df[[var]]\n    col[sample(n, replace = TRUE)]\n  }\n}\n\nboot_permute(mtcars, \"mpg\")\n#> function() {\n#>     col <- df[[var]]\n#>     col[sample(n, replace = TRUE)]\n#>   }\n#> <environment: 0x56479fac0b28>\nboot_permute_env <- rlang::fn_env(boot_permute(mtcars, \"mpg\"))\nrlang::env_print(boot_permute_env)\n#> <environment: 0x5647a06ef728>\n#> Parent: <environment: global>\n#> Bindings:\n#> • n: <int>\n#> • df: <df[,11]>\n#> • var: <chr>\n\nidentical(\n  lobstr::obj_addr(boot_permute_env$df),\n  lobstr::obj_addr(mtcars)\n)\n#> [1] TRUE\nidentical(boot_permute_env$df, mtcars)\n#> [1] TRUE\nidentical(boot_permute_env$var, \"mpg\")\n#> [1] TRUE\nll_poisson1 <- function(x) {\n  n <- length(x)\n\n  function(lambda) {\n    log(lambda) * sum(x) - n * lambda - sum(lfactorial(x))\n  }\n}\n\nll_poisson2 <- function(x) {\n  n <- length(x)\n  sum_x <- sum(x)\n  c <- sum(lfactorial(x))\n\n  function(lambda) {\n    log(lambda) * sum_x - n * lambda - c\n  }\n}\n\nx1 <- c(41, 30, 31, 38, 29, 24, 30, 29, 31, 38)\n\nbench::mark(\n  \"LL1\" = optimise(ll_poisson1(x1), c(0, 100), maximum = TRUE),\n  \"LL2\" = optimise(ll_poisson2(x1), c(0, 100), maximum = TRUE)\n)\n#> # A tibble: 2 × 6\n#>   expression      min   median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1 LL1          28.3µs   30.2µs    31979.    12.9KB     41.6\n#> 2 LL2          14.6µs   15.7µs    60147.        0B     48.2\ngenerate_ll_benches <- function(n) {\n  x_vec <- sample.int(n, n)\n\n  bench::mark(\n    \"LL1\" = optimise(ll_poisson1(x_vec), c(0, 100), maximum = TRUE),\n    \"LL2\" = optimise(ll_poisson2(x_vec), c(0, 100), maximum = TRUE)\n  )[1:4] %>%\n    dplyr::mutate(length = n, .before = expression)\n}\n\n(df_bench <- purrr::map_dfr(\n  .x = c(10, 20, 50, 100, 1000),\n  .f = ~ generate_ll_benches(n = .x)\n))\n#> # A tibble: 10 × 5\n#>    length expression      min   median `itr/sec`\n#>     <dbl> <bch:expr> <bch:tm> <bch:tm>     <dbl>\n#>  1     10 LL1          39.9µs   42.1µs    22543.\n#>  2     10 LL2          17.3µs   18.3µs    53968.\n#>  3     20 LL1            43µs   44.6µs    22159.\n#>  4     20 LL2          17.1µs     18µs    54815.\n#>  5     50 LL1          46.8µs   48.3µs    20374.\n#>  6     50 LL2          16.3µs   17.4µs    56753.\n#>  7    100 LL1          62.4µs   64.1µs    15369.\n#>  8    100 LL2          17.6µs   18.4µs    53652.\n#>  9   1000 LL1         857.4µs    965µs     1081.\n#> 10   1000 LL2          54.6µs   56.2µs    17530.\n\nggplot(\n  df_bench,\n  aes(\n    x = as.numeric(length),\n    y = median,\n    group = as.character(expression),\n    color = as.character(expression)\n  )\n) +\n  geom_point() +\n  geom_line() +\n  labs(\n    x = \"Vector length\",\n    y = \"Median Execution Time\",\n    colour = \"Function used\"\n  )"},{"path":"function-factories.html","id":"function-factories-functionals-exercises-10.5.1","chapter":"10 Function factories","heading":"10.4 Function factories + functionals (Exercises 10.5.1)","text":"Q1. following commands equivalent (x, f(z))?A1. depends whether () used data frame list.Q2. Compare contrast effects env_bind() vs. attach() following code.A2. Let’s compare contrast effects env_bind() vs. attach().attach() adds funs search path. Since functions names functions {base} package, attached names mask ones {base} package.env_bind() adds functions funs global environment, instead masking names {base} package.Note \"funs\" output.","code":"(a) `x$f(x$z)`.\n(b) `f(x$z)`.\n(c) `x$f(z)`.\n(d) `f(z)`.\n(e) It depends.\nf <- mean\nz <- 1\nx <- list(f = mean, z = 1)\n\nidentical(with(x, f(z)), x$f(x$z))\n#> [1] TRUE\n\nidentical(with(x, f(z)), f(x$z))\n#> [1] TRUE\n\nidentical(with(x, f(z)), x$f(z))\n#> [1] TRUE\n\nidentical(with(x, f(z)), f(z))\n#> [1] TRUE\nfuns <- list(\n  mean = function(x) mean(x, na.rm = TRUE),\n  sum = function(x) sum(x, na.rm = TRUE)\n)\n\nattach(funs)\n#> The following objects are masked from package:base:\n#> \n#>     mean, sum\n\nmean\n#> function(x) mean(x, na.rm = TRUE)\nhead(search())\n#> [1] \".GlobalEnv\"       \"funs\"             \"package:scales\"  \n#> [4] \"package:ggplot2\"  \"package:rlang\"    \"package:magrittr\"\n\nmean <- function(x) stop(\"Hi!\")\nmean\n#> function(x) stop(\"Hi!\")\nhead(search())\n#> [1] \".GlobalEnv\"       \"funs\"             \"package:scales\"  \n#> [4] \"package:ggplot2\"  \"package:rlang\"    \"package:magrittr\"\n\ndetach(funs)\nenv_bind(globalenv(), !!!funs)\nmean\n#> function(x) mean(x, na.rm = TRUE)\n\nmean <- function(x) stop(\"Hi!\")\nmean\n#> function(x) stop(\"Hi!\")\nenv_unbind(globalenv(), names(funs))"},{"path":"function-factories.html","id":"session-information-8","chapter":"10 Function factories","heading":"10.5 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bench         1.1.3   2023-05-04 [1] RSPM\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  colorspace    2.1-0   2023-01-23 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr         1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  farver        2.1.1   2022-07-06 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  ggplot2     * 3.4.4   2023-10-12 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  grid          4.3.2   2023-11-01 [3] local\n#>  gtable        0.3.4   2023-08-21 [1] RSPM\n#>  highr         0.10    2022-12-22 [1] RSPM\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  labeling      0.4.3   2023-08-29 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  lobstr        1.1.2   2022-06-22 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  munsell       0.5.0   2018-06-12 [1] RSPM\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  profmem       0.6.0   2020-12-13 [1] RSPM\n#>  purrr         1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  scales      * 1.3.0   2023-11-28 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"function-operators.html","id":"function-operators","chapter":"11 Function operators","heading":"11 Function operators","text":"Attaching needed libraries:","code":"\nlibrary(purrr, warn.conflicts = FALSE)"},{"path":"function-operators.html","id":"existing-function-operators-exercises-11.2.3","chapter":"11 Function operators","heading":"11.1 Existing function operators (Exercises 11.2.3)","text":"Q1. Base R provides function operator form Vectorize(). ? might use ?A1. Vectorize() function creates function vectorizes action provided function specified arguments (.e., acts element vector). see utility solving problem otherwise difficult solve.problem find indices matching numeric values given threshold creating hybrid following functions:%% (doesn’t provide way provide tolerance comparing numeric values),dplyr::near() (vectorized element-wise thus expects two vectors equal length)Let’s use :Note needed create new function neither existing functions want.solved complex task using Vectorize() function!Q2. Read source code possibly(). work?A2. Let’s look source code function:Looking code, can see possibly():uses tryCatch() error handlinghas parameter otherwise specify default value case error occurshas parameter quiet suppress error message (needed)Q3. Read source code safely(). work?A3. Let’s look source code function:Looking code, can see safely():uses list save results (function executes successfully) error (fails)uses tryCatch() error handlinghas parameter otherwise specify default value case error occurshas parameter quiet suppress error message (needed)","code":"\nwhich_near <- function(x, y, tolerance) {\n  # Vectorize `dplyr::near()` function only over the `y` argument.\n  # `Vectorize()` is a function operator and will return a function.\n  customNear <- Vectorize(dplyr::near, vectorize.args = c(\"y\"), SIMPLIFY = FALSE)\n\n  # Apply the vectorized function to vector arguments and then check where the\n  # comparisons are equal (i.e. `TRUE`) using `which()`.\n  #\n  # Use `compact()` to remove empty elements from the resulting list.\n  index_list <- purrr::compact(purrr::map(customNear(x, y, tol = tolerance), which))\n\n  # If there are any matches, return the indices as an atomic vector of integers.\n  if (length(index_list) > 0L) {\n    index_vector <- purrr::simplify(index_list, \"integer\")\n    return(index_vector)\n  }\n\n  # If there are no matches\n  return(integer(0L))\n}\nx1 <- c(2.1, 3.3, 8.45, 8, 6)\nx2 <- c(6, 8.40, 3)\n\nwhich_near(x1, x2, tolerance = 0.1)\n#> [1] 5 3\nwhich(x1 %in% x2)\n#> [1] 5\n\nwhich(dplyr::near(x1, x2, tol = 0.1))\n#> Warning in x - y: longer object length is not a multiple of\n#> shorter object length\n#> integer(0)\npossibly\n#> function (.f, otherwise = NULL, quiet = TRUE) \n#> {\n#>     .f <- as_mapper(.f)\n#>     force(otherwise)\n#>     check_bool(quiet)\n#>     function(...) {\n#>         tryCatch(.f(...), error = function(e) {\n#>             if (!quiet) \n#>                 message(\"Error: \", conditionMessage(e))\n#>             otherwise\n#>         })\n#>     }\n#> }\n#> <bytecode: 0x55d5181c6328>\n#> <environment: namespace:purrr>\nsafely\n#> function (.f, otherwise = NULL, quiet = TRUE) \n#> {\n#>     .f <- as_mapper(.f)\n#>     force(otherwise)\n#>     check_bool(quiet)\n#>     function(...) capture_error(.f(...), otherwise, quiet)\n#> }\n#> <bytecode: 0x55d518363bb8>\n#> <environment: namespace:purrr>\n\npurrr:::capture_error\n#> function (code, otherwise = NULL, quiet = TRUE) \n#> {\n#>     tryCatch(list(result = code, error = NULL), error = function(e) {\n#>         if (!quiet) \n#>             message(\"Error: \", conditionMessage(e))\n#>         list(result = otherwise, error = e)\n#>     })\n#> }\n#> <bytecode: 0x55d5183be3f0>\n#> <environment: namespace:purrr>"},{"path":"function-operators.html","id":"case-study-creating-your-own-function-operators-exercises-11.3.1","chapter":"11 Function operators","heading":"11.2 Case study: Creating your own function operators (Exercises 11.3.1)","text":"Q1. Weigh pros cons download.file %>% dot_every(10) %>% delay_by(0.1) versus download.file %>% delay_by(0.1) %>% dot_every(10).A1. Although chains piped operations produce number dots need amount time, subtle difference .download.file %>% dot_every(10) %>% delay_by(0.1), printing dot also delayed, first dot printed 10th URL download starts.download.file %>% delay_by(0.1) %>% dot_every(10), first dot printed 9th download finished, 10th download starts short delay.Q2. memoise download.file()? ?A2. Since download.file() meant download files Internet, memoising recommended following reasons:Memoization helpful giving input function returns output. necessarily case webpages since constantly change, may continue “download” outdated version webpage.Memoization helpful giving input function returns output. necessarily case webpages since constantly change, may continue “download” outdated version webpage.Memoization works caching results, can take significant amount memory.Memoization works caching results, can take significant amount memory.Q3. Create function operator reports whenever file created deleted working directory, using dir() setdiff(). global function effects might want track?A3. First, let’s create helper functions compare print added removed filenames:can write function operator use create functions necessary tracking:Let’s try :global function effects might want track:working directoryenvironment variablesconnectionslibrary pathsgraphics devicesetc.Q4. Write function operator logs timestamp message file every time function run.A4. following function operator logs timestamp message file every time function run:Q5. Modify delay_by() instead delaying fixed amount time, ensures certain amount time elapsed since function last called. , called g <- delay_by(1, f); g(); Sys.sleep(2); g() shouldn’t extra delay.A5. Modified version function meeting specified requirements:","code":"\nprint_multiple_entries <- function(header, entries) {\n  message(paste0(header, \":\\n\"), paste0(entries, collapse = \"\\n\"))\n}\n\nfile_comparator <- function(old, new) {\n  if (setequal(old, new)) {\n    return()\n  }\n\n  removed <- setdiff(old, new)\n  added <- setdiff(new, old)\n\n  if (length(removed) > 0L) print_multiple_entries(\"- File removed\", removed)\n  if (length(added) > 0L) print_multiple_entries(\"- File added\", added)\n}\ndir_tracker <- function(f) {\n  force(f)\n  function(...) {\n    old_files <- dir()\n    on.exit(file_comparator(old_files, dir()), add = TRUE)\n\n    f(...)\n  }\n}\n\nfile_creation_tracker <- dir_tracker(file.create)\nfile_deletion_tracker <- dir_tracker(file.remove)\nfile_creation_tracker(c(\"a.txt\", \"b.txt\"))\n#> - File added:\n#> a.txt\n#> b.txt\n#> [1] TRUE TRUE\n\nfile_deletion_tracker(c(\"a.txt\", \"b.txt\"))\n#> - File removed:\n#> a.txt\n#> b.txt\n#> [1] TRUE TRUE\n# helper function to write to a file connection\nwrite_line <- function(filepath, ...) {\n  cat(..., \"\\n\", sep = \"\", file = filepath, append = TRUE)\n}\n\n# function operator\nlogger <- function(f, filepath) {\n  force(f)\n  force(filepath)\n\n  write_line(filepath, \"Function created at: \", as.character(Sys.time()))\n\n  function(...) {\n    write_line(filepath, \"Function called at:  \", as.character(Sys.time()))\n    f(...)\n  }\n}\n\n# check that the function works as expected with a tempfile\nwithr::with_tempfile(\"logfile\", code = {\n  logged_runif <- logger(runif, logfile)\n\n  Sys.sleep(sample.int(10, 1))\n  logged_runif(1)\n\n  Sys.sleep(sample.int(10, 1))\n  logged_runif(2)\n\n  Sys.sleep(sample.int(10, 1))\n  logged_runif(3)\n\n  cat(readLines(logfile), sep = \"\\n\")\n})\n#> Function created at: 2024-01-07 00:44:00.426587\n#> Function called at:  2024-01-07 00:44:05.433612\n#> Function called at:  2024-01-07 00:44:10.439013\n#> Function called at:  2024-01-07 00:44:18.447574\ndelay_by_atleast <- function(f, amount) {\n  force(f)\n  force(amount)\n\n  # the last time the function was run\n  last_time <- NULL\n\n  function(...) {\n    if (!is.null(last_time)) {\n      wait <- (last_time - Sys.time()) + amount\n      if (wait > 0) Sys.sleep(wait)\n    }\n\n    # update the time in the parent frame for the next run when the function finishes\n    on.exit(last_time <<- Sys.time())\n\n    f(...)\n  }\n}"},{"path":"function-operators.html","id":"session-information-9","chapter":"11 Function operators","heading":"11.3 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr         1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  purrr       * 1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"base-types.html","id":"base-types","chapter":"12 Base Types","heading":"12 Base Types","text":"exercises.","code":""},{"path":"s3.html","id":"s3","chapter":"13 S3","heading":"13 S3","text":"Attaching needed libraries:","code":"\nlibrary(sloop, warn.conflicts = FALSE)\nlibrary(dplyr, warn.conflicts = FALSE)\nlibrary(purrr, warn.conflicts = FALSE)"},{"path":"s3.html","id":"basics-exercises-13.2.1","chapter":"13 S3","heading":"13.1 Basics (Exercises 13.2.1)","text":"Q1. Describe difference t.test() t.data.frame(). function called?A1. difference t.test() t.data.frame() following:t.test() generic function perform t-test.t.test() generic function perform t-test.t.data.frame() method generic t() (matrix transform function) dispatched data.frame objects.t.data.frame() method generic t() (matrix transform function) dispatched data.frame objects.can also confirm function types using ftype():Q2. Make list commonly used base R functions contain . name S3 methods.A2. common R functions . S3 methods:.equal().* functions (like .data.frame(), .numeric(), etc.)install.packages().exit()\netc.example,Q3. .data.frame.data.frame() method ? confusing? avoid confusion code?A3. ’s S3 method generic .data.frame().can seen methods supported generic:Given number .s name, quite confusing figure name generic name class.Q4. Describe difference behaviour two calls.A4. difference behaviour specified calls.unclassing, mean generic dispatches .Date method:unclassing, mean generic dispatches .numeric method:Q5. class object following code return? base type built ? attributes use?A5. object based base type closure6, type function.class ecdf, superclasses.Apart class, following attributes:Q6. class object following code return? base type built ? attributes use?A6. object based base type integer.class table.Apart class, following attributes:","code":"\nftype(t.test)\n#> [1] \"S3\"      \"generic\"\nftype(t.data.frame)\n#> [1] \"S3\"     \"method\"\nftype(as.data.frame)\n#> [1] \"S3\"      \"generic\"\nftype(on.exit)\n#> [1] \"primitive\"\nftype(as.data.frame.data.frame)\n#> [1] \"S3\"     \"method\"\ns3_methods_generic(\"as.data.frame\") %>%\n  dplyr::filter(class == \"data.frame\")\n#> # A tibble: 1 × 4\n#>   generic       class      visible source\n#>   <chr>         <chr>      <lgl>   <chr> \n#> 1 as.data.frame data.frame TRUE    base\nset.seed(1014)\nsome_days <- as.Date(\"2017-01-31\") + sample(10, 5)\nmean(some_days)\n#> [1] \"2017-02-06\"\nmean(unclass(some_days))\n#> [1] 17203.4\nsome_days <- as.Date(\"2017-01-31\") + sample(10, 5)\n\nsome_days\n#> [1] \"2017-02-06\" \"2017-02-09\" \"2017-02-05\" \"2017-02-08\"\n#> [5] \"2017-02-07\"\n\ns3_dispatch(mean(some_days))\n#> => mean.Date\n#>  * mean.default\n\nmean(some_days)\n#> [1] \"2017-02-07\"\nunclass(some_days)\n#> [1] 17203 17206 17202 17205 17204\n\nmean(unclass(some_days))\n#> [1] 17204\n\ns3_dispatch(mean(unclass(some_days)))\n#>    mean.double\n#>    mean.numeric\n#> => mean.default\nx <- ecdf(rpois(100, 10))\nx\nx <- ecdf(rpois(100, 10))\nx\n#> Empirical CDF \n#> Call: ecdf(rpois(100, 10))\n#>  x[1:18] =      2,      3,      4,  ...,     18,     19\n\notype(x)\n#> [1] \"S3\"\ntypeof(x)\n#> [1] \"closure\"\ns3_class(x)\n#> [1] \"ecdf\"     \"stepfun\"  \"function\"\nattributes(x)\n#> $class\n#> [1] \"ecdf\"     \"stepfun\"  \"function\"\n#> \n#> $call\n#> ecdf(rpois(100, 10))\nx <- table(rpois(100, 5))\nx\nx <- table(rpois(100, 5))\nx\n#> \n#>  1  2  3  4  5  6  7  8  9 10 \n#>  7  7 18 13 14 14 16  4  4  3\n\notype(x)\n#> [1] \"S3\"\ntypeof(x)\n#> [1] \"integer\"\ns3_class(x)\n#> [1] \"table\"\nattributes(x)\n#> $dim\n#> [1] 10\n#> \n#> $dimnames\n#> $dimnames[[1]]\n#>  [1] \"1\"  \"2\"  \"3\"  \"4\"  \"5\"  \"6\"  \"7\"  \"8\"  \"9\"  \"10\"\n#> \n#> \n#> $class\n#> [1] \"table\""},{"path":"s3.html","id":"classes-exercises-13.3.4","chapter":"13 S3","heading":"13.2 Classes (Exercises 13.3.4)","text":"Q1. Write constructor data.frame objects. base type data frame built ? attributes use? restrictions placed individual elements? names?A1. data frame built top named list atomic vectors attributes row names:restriction imposed individual elements need length. Additionally, names need syntactically valid unique.Let’s try :Q2. Enhance factor() helper better behaviour one values found levels. base::factor() situation?A2. one values found levels, values converted NA base::factor():new constructor, can throw error inform user:Let’s try :Q3. Carefully read source code factor(). constructor ?A3. source code factor() can read .number ways base version flexible.allows labeling values:checks levels duplicated.levels argument can NULL.Q4. Factors optional “contrasts” attribute. Read help C(), briefly describe purpose attribute. type ? Rewrite new_factor() constructor include attribute.A4. Categorical variables typically encoded dummy variables regression models default level compared first factor level. Contrats provide flexible way comparisons.can set \"contrasts\" attribute factor using stats::C().Alternatively, can set \"contrasts\" attribute using matrix (?contrasts):[Contrasts] can matrix one row level factor suitable function like contr.poly character string giving name functionThe constructor provided book:can updated also support contrasts:Q5. Read documentation utils::.roman(). write constructor class? need validator? might helper ?A5. utils::.roman() converts Indo-Arabic numerals Roman numerals. Removing class also reveals implemented using base type integer:Therefore, can create simple constructor create new instance class:docs mention following:numbers 1 3899 unique representation roman numbers, hence others result .roman(NA).Therefore, can warn user return NA validator function:helper function can coerce entered input integer type convenience:Let’s try :","code":"\nunclass(data.frame())\n#> named list()\n#> attr(,\"row.names\")\n#> integer(0)\nnew_data_frame <- function(x = list(), row.names = character()) {\n  # row names should be character\n  if (!all(is.character(row.names))) {\n    stop(\"Row name should be of `chracter` type.\", call. = FALSE)\n  }\n\n  # all elements should have the same length\n  unique_element_lengths <- unique(purrr::map_int(x, length))\n  if (length(unique_element_lengths) > 1L) {\n    stop(\"All list elements in `x` should have same length.\", call. = FALSE)\n  }\n\n  # if not provided, generate row names\n  # this is necessary if there is at least one element in the list\n  if (length(x) > 0L && length(row.names) == 0L) {\n    row.names <- .set_row_names(unique_element_lengths)\n  }\n\n  structure(x, class = \"data.frame\", row.names = row.names)\n}\nnew_data_frame(list(\"x\" = 1, \"y\" = c(2, 3)))\n#> Error: All list elements in `x` should have same length.\n\nnew_data_frame(list(\"x\" = 1, \"y\" = c(2)), row.names = 1L)\n#> Error: Row name should be of `chracter` type.\n\nnew_data_frame(list())\n#> data frame with 0 columns and 0 rows\n\nnew_data_frame(list(\"x\" = 1, \"y\" = 2))\n#>   x y\n#> 1 1 2\n\nnew_data_frame(list(\"x\" = 1, \"y\" = 2), row.names = \"row-1\")\n#>       x y\n#> row-1 1 2\nbase::factor(c(\"a\", \"b\", \"c\"), levels = c(\"a\", \"c\"))\n#> [1] a    <NA> c   \n#> Levels: a c\nnew_factor <- function(x = integer(), levels = character()) {\n  stopifnot(is.integer(x))\n  stopifnot(is.character(levels))\n\n  structure(\n    x,\n    levels = levels,\n    class = \"factor\"\n  )\n}\n\nvalidate_factor <- function(x) {\n  values <- unclass(x)\n  levels <- attr(x, \"levels\")\n\n  if (!all(!is.na(values) & values > 0)) {\n    stop(\n      \"All `x` values must be non-missing and greater than zero\",\n      call. = FALSE\n    )\n  }\n\n  if (length(levels) < max(values)) {\n    stop(\n      \"There must be at least as many `levels` as possible values in `x`\",\n      call. = FALSE\n    )\n  }\n\n  x\n}\n\ncreate_factor <- function(x = character(), levels = unique(x)) {\n  ind <- match(x, levels)\n\n  if (any(is.na(ind))) {\n    missing_values <- x[which(is.na(match(x, levels)))]\n\n    stop(\n      paste0(\n        \"Following values from `x` are not present in `levels`:\\n\",\n        paste0(missing_values, collapse = \"\\n\")\n      ),\n      call. = FALSE\n    )\n  }\n\n  validate_factor(new_factor(ind, levels))\n}\ncreate_factor(c(\"a\", \"b\", \"c\"), levels = c(\"a\", \"c\"))\n#> Error: Following values from `x` are not present in `levels`:\n#> b\n\ncreate_factor(c(\"a\", \"b\", \"c\"), levels = c(\"a\", \"b\", \"c\"))\n#> [1] a b c\n#> Levels: a b c\nx <- c(\"a\", \"b\", \"b\")\nlevels <- c(\"a\", \"b\", \"c\")\nlabels <- c(\"one\", \"two\", \"three\")\n\nfactor(x, levels = levels, labels = labels)\n#> [1] one two two\n#> Levels: one two three\nx <- c(\"a\", \"b\", \"b\")\nlevels <- c(\"a\", \"b\", \"b\")\n\nfactor(x, levels = levels)\n#> Error in `levels<-`(`*tmp*`, value = as.character(levels)): factor level [3] is duplicated\n\ncreate_factor(x, levels = levels)\n#> [1] a b b\n#> Levels: a b b\n#> Warning in print.factor(x): duplicated level [3] in factor\nx <- c(\"a\", \"b\", \"b\")\n\nfactor(x, levels = NULL)\n#> [1] <NA> <NA> <NA>\n#> Levels:\n\ncreate_factor(x, levels = NULL)\n#> Error: Following values from `x` are not present in `levels`:\n#> a\n#> b\n#> b\nnew_factor <- function(x = integer(), levels = character()) {\n  stopifnot(is.integer(x))\n  stopifnot(is.character(levels))\n\n  structure(\n    x,\n    levels = levels,\n    class = \"factor\"\n  )\n}\nnew_factor <- function(x = integer(),\n                       levels = character(),\n                       contrasts = NULL) {\n  stopifnot(is.integer(x))\n  stopifnot(is.character(levels))\n\n  if (!is.null(contrasts)) {\n    stopifnot(is.matrix(contrasts) && is.numeric(contrasts))\n  }\n\n  structure(\n    x,\n    levels = levels,\n    class = \"factor\",\n    contrasts = contrasts\n  )\n}\nas.roman(1)\n#> [1] I\n\ntypeof(unclass(as.roman(1)))\n#> [1] \"integer\"\nnew_roman <- function(x = integer()) {\n  stopifnot(is.integer(x))\n\n  structure(x, class = \"roman\")\n}\nas.roman(10000)\n#> [1] <NA>\nvalidate_new_roman <- function(x) {\n  int_values <- unclass(x)\n\n  if (any(int_values < 1L | int_values > 3899L)) {\n    warning(\n      \"Integer should be between 1 and 3899. Returning `NA` otherwise.\",\n      call. = FALSE\n    )\n  }\n\n  x\n}\nroman <- function(x = integer()) {\n  x <- as.integer(x)\n\n  validate_new_roman(new_roman(x))\n}\nroman(1)\n#> [1] I\n\nroman(c(5, 20, 100, 150, 100000))\n#> Warning: Integer should be between 1 and 3899. Returning\n#> `NA` otherwise.\n#> [1] V    XX   C    CL   <NA>"},{"path":"s3.html","id":"generics-and-methods-exercises-13.4.4","chapter":"13 S3","heading":"13.3 Generics and methods (Exercises 13.4.4)","text":"Q1. Read source code t() t.test() confirm t.test() S3 generic S3 method. happens create object class test call t() ? ?A1. Looking source code functions, can see generic, can confirm using sloop:Looking S3 dispatch, can see since R can’t find S3 method test class generic function t(), dispatches default method, converts structure matrix:behaviour can observed vector:Q2. generics table class methods ?A2. table class methods following generics:Q3. generics ecdf class methods ?A3. ecdf class methods following generics:Q4. base generic greatest number defined methods?A4. answer question, first, let’s list functions base retain generics.Now ’s simple matter counting number methods per generic ordering data frame descending order count:reveals base generic function methods print().Q5. Carefully read documentation UseMethod() explain following code returns results . two usual rules function evaluation UseMethod() violate?A5. called directly, g.default() method takes x value argument y global environment:, g() function called, takes x argument, comes function environment:docs ?UseMethod() clarify case:local variables defined call UseMethod retainedThat , UseMethod() calls g.default(), variables defined inside generic also available g.default() method. arguments supplied function passed , however, affected code inside generic.Two rules function evaluation violated UseMethod():Name maskingA fresh startQ6. arguments [? hard question answer?A6. difficult say many formal arguments subsetting [ operator generic function methods vectors, matrices, arrays, lists, etc., different methods different number arguments:can sample see wide variation number formal arguments:","code":"\nx <- structure(1:10, class = \"test\")\nt(x)\nt\n#> function (x) \n#> UseMethod(\"t\")\n#> <bytecode: 0x56119bea5208>\n#> <environment: namespace:base>\nsloop::is_s3_generic(\"t\")\n#> [1] TRUE\n\nt.test\n#> function (x, ...) \n#> UseMethod(\"t.test\")\n#> <bytecode: 0x561196f41140>\n#> <environment: namespace:stats>\nsloop::is_s3_generic(\"t.test\")\n#> [1] TRUE\nx <- structure(1:10, class = \"test\")\nt(x)\n#>      [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10]\n#> [1,]    1    2    3    4    5    6    7    8    9    10\n#> attr(,\"class\")\n#> [1] \"test\"\ns3_dispatch(t(x))\n#> => t.test\n#>  * t.default\nt(1:10)\n#>      [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10]\n#> [1,]    1    2    3    4    5    6    7    8    9    10\ns3_methods_class(\"table\")\n#> # A tibble: 11 × 4\n#>    generic       class visible source             \n#>    <chr>         <chr> <lgl>   <chr>              \n#>  1 [             table TRUE    base               \n#>  2 aperm         table TRUE    base               \n#>  3 as_tibble     table FALSE   registered S3method\n#>  4 as.data.frame table TRUE    base               \n#>  5 Axis          table FALSE   registered S3method\n#>  6 lines         table FALSE   registered S3method\n#>  7 plot          table FALSE   registered S3method\n#>  8 points        table FALSE   registered S3method\n#>  9 print         table TRUE    base               \n#> 10 summary       table TRUE    base               \n#> 11 tail          table FALSE   registered S3method\ns3_methods_class(\"ecdf\")\n#> # A tibble: 4 × 4\n#>   generic  class visible source             \n#>   <chr>    <chr> <lgl>   <chr>              \n#> 1 plot     ecdf  TRUE    stats              \n#> 2 print    ecdf  FALSE   registered S3method\n#> 3 quantile ecdf  FALSE   registered S3method\n#> 4 summary  ecdf  FALSE   registered S3method\n# getting all functions names\nobjs <- mget(ls(\"package:base\", all = TRUE), inherits = TRUE)\nfuns <- Filter(is.function, objs)\n\n# extracting only generics\ngenFuns <- names(funs) %>%\n  purrr::keep(~ sloop::is_s3_generic(.x))\npurrr::map_dfr(\n  genFuns,\n  ~ s3_methods_generic(.)\n) %>%\n  dplyr::group_by(generic) %>%\n  dplyr::tally() %>%\n  dplyr::arrange(desc(n))\n#> # A tibble: 122 × 2\n#>    generic           n\n#>    <chr>         <int>\n#>  1 print           289\n#>  2 format          130\n#>  3 [                53\n#>  4 summary          39\n#>  5 as.character     38\n#>  6 as.data.frame    32\n#>  7 plot             30\n#>  8 [[               26\n#>  9 [<-              17\n#> 10 $                15\n#> # ℹ 112 more rows\ng <- function(x) {\n  x <- 10\n  y <- 10\n  UseMethod(\"g\")\n}\ng.default <- function(x) c(x = x, y = y)\nx <- 1\ny <- 1\ng(x)\n#>  x  y \n#>  1 10\ng.default(x)\n#> x y \n#> 1 1\ng(x)\n#>  x  y \n#>  1 10\ns3_methods_generic(\"[\") %>%\n  dplyr::filter(source == \"base\")\n#> # A tibble: 17 × 4\n#>    generic class           visible source\n#>    <chr>   <chr>           <lgl>   <chr> \n#>  1 [       AsIs            TRUE    base  \n#>  2 [       data.frame      TRUE    base  \n#>  3 [       Date            TRUE    base  \n#>  4 [       difftime        TRUE    base  \n#>  5 [       Dlist           TRUE    base  \n#>  6 [       DLLInfoList     TRUE    base  \n#>  7 [       factor          TRUE    base  \n#>  8 [       hexmode         TRUE    base  \n#>  9 [       listof          TRUE    base  \n#> 10 [       noquote         TRUE    base  \n#> 11 [       numeric_version TRUE    base  \n#> 12 [       octmode         TRUE    base  \n#> 13 [       POSIXct         TRUE    base  \n#> 14 [       POSIXlt         TRUE    base  \n#> 15 [       simple.list     TRUE    base  \n#> 16 [       table           TRUE    base  \n#> 17 [       warnings        TRUE    base\n# table\nnames(formals(`[.table`))\n#> [1] \"x\"    \"i\"    \"j\"    \"...\"  \"drop\"\n\n# Date\nnames(formals(`[.Date`))\n#> [1] \"x\"    \"...\"  \"drop\"\n\n# data frame\nnames(formals(`[.data.frame`))\n#> [1] \"x\"    \"i\"    \"j\"    \"drop\"\n\n# etc."},{"path":"s3.html","id":"object-styles-exercises-13.5.1","chapter":"13 S3","heading":"13.4 Object styles (Exercises 13.5.1)","text":"Q1. Categorise objects returned lm(), factor(), table(), .Date(), .POSIXct() ecdf(), ordered(), () styles described .A1. Objects returned functions can categorized follows:Vector style objects (length represents . observations)factor()table().Date().POSIXct()ordered()Record style objects (equi-length vectors represent object components)None.Dataframe style objects (Record style two-dimensions)None.Scalar objects (list represent single thing)lm() (represent one regression model)ecdf() (represents one distribution)() special:\njust adds new class object indicate treated .Therefore, object style superclass’ object style.Q2. constructor function lm objects, new_lm(), look like? Use ?lm experimentation figure required fields types.A2. lm object scalar object, .e. object contains named list atomic vectors varying lengths types represent single thing (regression model).Based information, can write new constructor object:","code":"\nfactor_obj <- factor(c(\"a\", \"b\"))\nlength(factor_obj)\n#> [1] 2\nlength(unclass(factor_obj))\n#> [1] 2\ntab_object <- table(mtcars$am)\nlength(tab_object)\n#> [1] 2\nlength(unlist(tab_object))\n#> [1] 2\ndate_obj <- as.Date(\"02/27/92\", \"%m/%d/%y\")\nlength(date_obj)\n#> [1] 1\nlength(unclass(date_obj))\n#> [1] 1\nposix_obj <- as.POSIXct(1472562988, origin = \"1960-01-01\")\nlength(posix_obj)\n#> [1] 1\nlength(unclass(posix_obj))\n#> [1] 1\nordered_obj <- ordered(factor(c(\"a\", \"b\")))\nlength(ordered_obj)\n#> [1] 2\nlength(unclass(ordered_obj))\n#> [1] 2\nlm_obj <- lm(wt ~ mpg, mtcars)\nlength(lm_obj)\n#> [1] 12\nlength(unclass(lm_obj))\n#> [1] 12\necdf_obj <- ecdf(rnorm(12))\nlength(ecdf_obj)\n#> [1] 1\nlength(unclass(ecdf_obj))\n#> [1] 1\nx <- ecdf(rnorm(12))\nclass(x)\n#> [1] \"ecdf\"     \"stepfun\"  \"function\"\nclass(I(x))\n#> [1] \"AsIs\"     \"ecdf\"     \"stepfun\"  \"function\"\nmod <- lm(wt ~ mpg, mtcars)\n\ntypeof(mod)\n#> [1] \"list\"\n\nattributes(mod)\n#> $names\n#>  [1] \"coefficients\"  \"residuals\"     \"effects\"      \n#>  [4] \"rank\"          \"fitted.values\" \"assign\"       \n#>  [7] \"qr\"            \"df.residual\"   \"xlevels\"      \n#> [10] \"call\"          \"terms\"         \"model\"        \n#> \n#> $class\n#> [1] \"lm\"\n\npurrr::map_chr(unclass(mod), typeof)\n#>  coefficients     residuals       effects          rank \n#>      \"double\"      \"double\"      \"double\"     \"integer\" \n#> fitted.values        assign            qr   df.residual \n#>      \"double\"     \"integer\"        \"list\"     \"integer\" \n#>       xlevels          call         terms         model \n#>        \"list\"    \"language\"    \"language\"        \"list\"\n\npurrr::map_int(unclass(mod), length)\n#>  coefficients     residuals       effects          rank \n#>             2            32            32             1 \n#> fitted.values        assign            qr   df.residual \n#>            32             2             5             1 \n#>       xlevels          call         terms         model \n#>             0             3             3             2\nnew_lm <- function(coefficients,\n                   residuals,\n                   effects,\n                   rank,\n                   fitted.values,\n                   assign,\n                   qr,\n                   df.residual,\n                   xlevels,\n                   call,\n                   terms,\n                   model) {\n  stopifnot(\n    is.double(coefficients),\n    is.double(residuals),\n    is.double(effects),\n    is.integer(rank),\n    is.double(fitted.values),\n    is.integer(assign),\n    is.list(qr),\n    is.integer(df.residual),\n    is.list(xlevels),\n    is.language(call),\n    is.language(terms),\n    is.list(model)\n  )\n\n  structure(\n    list(\n      coefficients  = coefficients,\n      residuals     = residuals,\n      effects       = effects,\n      rank          = rank,\n      fitted.values = fitted.values,\n      assign        = assign,\n      qr            = qr,\n      df.residual   = df.residual,\n      xlevels       = xlevels,\n      call          = call,\n      terms         = terms,\n      model         = model\n    ),\n    class = \"lm\"\n  )\n}"},{"path":"s3.html","id":"inheritance-exercises-13.6.3","chapter":"13 S3","heading":"13.5 Inheritance (Exercises 13.6.3)","text":"Q1. [.Date support subclasses? fail support subclasses?A1. [.Date method defined follows:.Date function looks like :, oldClass class().Therefore, reading code, can surmise :[.Date supports subclasses preserving class input.[.Date fails support subclasses preserving attributes input.example,Q2. R two classes representing date time data, POSIXct POSIXlt, inherit POSIXt. generics different behaviours two classes? generics share behaviour?A2. First, let’s demonstrate POSIXct POSIXlt indeed subclasses POSIXt superclass.Remember way S3 method dispatch works, generic method superclass, method also inherited subclass.can extract vector generics supported sub- super-classes:Methods specific subclasses:Let’s see example:Methods inherited subclasses superclass:Let’s see one example generic:Q3. expect code return? actually return? ?A3. Naively, expect code return \"a1\", actually returns \"a2\":S3 dispatch explains :mentioned book, UseMethod() functiontracks list potential next methods special variable, means modifying object ’s dispatched upon impact method gets called next.special variable .Class:.Class character vector classes used find next method. NextMethod adds attribute “previous” .Class giving .Class last used dispatch, shifts .Class along used dispatch., can print .Class confirm adding new class x indeed doesn’t change .Class, therefore dispatch occurs \"a2\" class:","code":"\nsloop::s3_get_method(\"[.Date\")\n#> function (x, ..., drop = TRUE) \n#> {\n#>     .Date(NextMethod(\"[\"), oldClass(x))\n#> }\n#> <bytecode: 0x56119c235db0>\n#> <environment: namespace:base>\n.Date\n#> function (xx, cl = \"Date\") \n#> `class<-`(xx, cl)\n#> <bytecode: 0x561199d84c50>\n#> <environment: namespace:base>\nx <- structure(Sys.Date(), name = \"myName\", class = c(\"subDate\", \"Date\"))\n\n# `$name` is gone\nattributes(x[1])\n#> $class\n#> [1] \"subDate\" \"Date\"\n\nx[1]\n#> [1] \"2024-01-07\"\ndt_lt <- as.POSIXlt(Sys.time(), \"GMT\")\nclass(dt_lt)\n#> [1] \"POSIXlt\" \"POSIXt\"\n\ndt_ct <- as.POSIXct(Sys.time(), \"GMT\")\nclass(dt_ct)\n#> [1] \"POSIXct\" \"POSIXt\"\n\ndt_t <- structure(dt_ct, class = \"POSIXt\")\nclass(dt_t)\n#> [1] \"POSIXt\"\n(t_generics <- s3_methods_class(\"POSIXt\")$generic)\n#>  [1] \"-\"            \"+\"            \"all.equal\"   \n#>  [4] \"as.character\" \"Axis\"         \"cut\"         \n#>  [7] \"diff\"         \"hist\"         \"is.numeric\"  \n#> [10] \"julian\"       \"Math\"         \"months\"      \n#> [13] \"Ops\"          \"pretty\"       \"quantile\"    \n#> [16] \"quarters\"     \"round\"        \"seq\"         \n#> [19] \"str\"          \"trunc\"        \"weekdays\"\n\n(lt_generics <- s3_methods_class(\"POSIXlt\")$generic)\n#>  [1] \"[\"             \"[[\"            \"[[<-\"         \n#>  [4] \"[<-\"           \"$<-\"           \"anyNA\"        \n#>  [7] \"as.data.frame\" \"as.Date\"       \"as.double\"    \n#> [10] \"as.list\"       \"as.matrix\"     \"as.POSIXct\"   \n#> [13] \"as.vector\"     \"c\"             \"duplicated\"   \n#> [16] \"format\"        \"is.finite\"     \"is.infinite\"  \n#> [19] \"is.na\"         \"is.nan\"        \"length\"       \n#> [22] \"length<-\"      \"mean\"          \"names\"        \n#> [25] \"names<-\"       \"print\"         \"rep\"          \n#> [28] \"sort\"          \"summary\"       \"Summary\"      \n#> [31] \"unique\"        \"weighted.mean\" \"xtfrm\"\n\n(ct_generics <- s3_methods_class(\"POSIXct\")$generic)\n#>  [1] \"[\"             \"[[\"            \"[<-\"          \n#>  [4] \"as.data.frame\" \"as.Date\"       \"as.list\"      \n#>  [7] \"as.POSIXlt\"    \"c\"             \"format\"       \n#> [10] \"length<-\"      \"mean\"          \"print\"        \n#> [13] \"rep\"           \"split\"         \"summary\"      \n#> [16] \"Summary\"       \"weighted.mean\" \"xtfrm\"\nunion(lt_generics, ct_generics)\n#>  [1] \"[\"             \"[[\"            \"[[<-\"         \n#>  [4] \"[<-\"           \"$<-\"           \"anyNA\"        \n#>  [7] \"as.data.frame\" \"as.Date\"       \"as.double\"    \n#> [10] \"as.list\"       \"as.matrix\"     \"as.POSIXct\"   \n#> [13] \"as.vector\"     \"c\"             \"duplicated\"   \n#> [16] \"format\"        \"is.finite\"     \"is.infinite\"  \n#> [19] \"is.na\"         \"is.nan\"        \"length\"       \n#> [22] \"length<-\"      \"mean\"          \"names\"        \n#> [25] \"names<-\"       \"print\"         \"rep\"          \n#> [28] \"sort\"          \"summary\"       \"Summary\"      \n#> [31] \"unique\"        \"weighted.mean\" \"xtfrm\"        \n#> [34] \"as.POSIXlt\"    \"split\"\ns3_dispatch(is.na(dt_lt))\n#> => is.na.POSIXlt\n#>    is.na.POSIXt\n#>    is.na.default\n#>  * is.na (internal)\n\ns3_dispatch(is.na(dt_ct))\n#>    is.na.POSIXct\n#>    is.na.POSIXt\n#>    is.na.default\n#> => is.na (internal)\n\ns3_dispatch(is.na(dt_t))\n#>    is.na.POSIXt\n#>    is.na.default\n#> => is.na (internal)\nsetdiff(t_generics, union(lt_generics, ct_generics))\n#>  [1] \"-\"            \"+\"            \"all.equal\"   \n#>  [4] \"as.character\" \"Axis\"         \"cut\"         \n#>  [7] \"diff\"         \"hist\"         \"is.numeric\"  \n#> [10] \"julian\"       \"Math\"         \"months\"      \n#> [13] \"Ops\"          \"pretty\"       \"quantile\"    \n#> [16] \"quarters\"     \"round\"        \"seq\"         \n#> [19] \"str\"          \"trunc\"        \"weekdays\"\ns3_dispatch(is.numeric(dt_lt))\n#>    is.numeric.POSIXlt\n#> => is.numeric.POSIXt\n#>    is.numeric.default\n#>  * is.numeric (internal)\n\ns3_dispatch(is.numeric(dt_ct))\n#>    is.numeric.POSIXct\n#> => is.numeric.POSIXt\n#>    is.numeric.default\n#>  * is.numeric (internal)\n\ns3_dispatch(is.numeric(dt_t))\n#> => is.numeric.POSIXt\n#>    is.numeric.default\n#>  * is.numeric (internal)\ngeneric2 <- function(x) UseMethod(\"generic2\")\ngeneric2.a1 <- function(x) \"a1\"\ngeneric2.a2 <- function(x) \"a2\"\ngeneric2.b <- function(x) {\n  class(x) <- \"a1\"\n  NextMethod()\n}\n\ngeneric2(structure(list(), class = c(\"b\", \"a2\")))\ngeneric2 <- function(x) UseMethod(\"generic2\")\ngeneric2.a1 <- function(x) \"a1\"\ngeneric2.a2 <- function(x) \"a2\"\ngeneric2.b <- function(x) {\n  class(x) <- \"a1\"\n  NextMethod()\n}\n\ngeneric2(structure(list(), class = c(\"b\", \"a2\")))\n#> [1] \"a2\"\nsloop::s3_dispatch(generic2(structure(list(), class = c(\"b\", \"a2\"))))\n#> => generic2.b\n#> -> generic2.a2\n#>    generic2.default\ngeneric2.b <- function(x) {\n  message(paste0(\"before: \", paste0(.Class, collapse = \", \")))\n  class(x) <- \"a1\"\n  message(paste0(\"after: \", paste0(.Class, collapse = \", \")))\n\n  NextMethod()\n}\n\ninvisible(generic2(structure(list(), class = c(\"b\", \"a2\"))))\n#> before: b, a2\n#> after: b, a2"},{"path":"s3.html","id":"dispatch-details-exercises-13.7.5","chapter":"13 S3","heading":"13.6 Dispatch details (Exercises 13.7.5)","text":"Q1. Explain differences dispatch :A1. differences dispatch due classes arguments:x1 implicit class integer inherits numeric, x2 explicitly assigned class integer.Q2. classes method Math group generic base R? Read source code. methods work?A2. following classes method Math group generic base R:Reading source code methods:Math.factor() Math.Date() provide error message:Math.data.frame() defined follows (except first line code, deliberately added):can surmised code: method checks elements expected type., applies generic (tracked via environment variable .Generic) element list atomic vectors makes data frame:, produces error:Q3. Math.difftime() complicated described. ?A3. Math.difftime() source code looks like following:group generic bit complicated produces error generics, works others.","code":"\nlength.integer <- function(x) 10\n\nx1 <- 1:5\nclass(x1)\n#> [1] \"integer\"\ns3_dispatch(length(x1))\n#>  * length.integer\n#>    length.numeric\n#>    length.default\n#> => length (internal)\n\nx2 <- structure(x1, class = \"integer\")\nclass(x2)\n#> [1] \"integer\"\ns3_dispatch(length(x2))\n#> => length.integer\n#>    length.default\n#>  * length (internal)\ns3_class(x1)\n#> [1] \"integer\" \"numeric\"\n\ns3_class(x2)\n#> [1] \"integer\"\ns3_methods_generic(\"Math\") %>%\n  dplyr::filter(source == \"base\")\n#> # A tibble: 5 × 4\n#>   generic class      visible source\n#>   <chr>   <chr>      <lgl>   <chr> \n#> 1 Math    data.frame TRUE    base  \n#> 2 Math    Date       TRUE    base  \n#> 3 Math    difftime   TRUE    base  \n#> 4 Math    factor     TRUE    base  \n#> 5 Math    POSIXt     TRUE    base\nMath.factor <- function(x, ...) {\n  stop(gettextf(\"%s not meaningful for factors\", sQuote(.Generic)))\n}\n\nMath.Date <- function(x, ...) {\n  stop(gettextf(\"%s not defined for \\\"Date\\\" objects\", .Generic),\n    domain = NA\n  )\n}\nMath.data.frame <- function(x, ...) {\n  message(paste0(\"Environment variable `.Generic` set to: \", .Generic))\n\n  mode.ok <- vapply(x, function(x) {\n    is.numeric(x) || is.logical(x) || is.complex(x)\n  }, NA)\n\n  if (all(mode.ok)) {\n    x[] <- lapply(X = x, FUN = .Generic, ...)\n    return(x)\n  } else {\n    vnames <- names(x)\n    if (is.null(vnames)) vnames <- seq_along(x)\n    stop(\n      \"non-numeric-alike variable(s) in data frame: \",\n      paste(vnames[!mode.ok], collapse = \", \")\n    )\n  }\n}\ndf1 <- data.frame(x = 1:2, y = 3:4)\nsqrt(df1)\n#> Environment variable `.Generic` set to: sqrt\n#>          x        y\n#> 1 1.000000 1.732051\n#> 2 1.414214 2.000000\ndf2 <- data.frame(x = c(TRUE, FALSE), y = c(\"a\", \"b\"))\nabs(df2)\n#> Environment variable `.Generic` set to: abs\n#> Error in Math.data.frame(df2): non-numeric-alike variable(s) in data frame: y\nMath.difftime <- function(x, ...) {\n  switch(.Generic,\n    \"abs\" = ,\n    \"sign\" = ,\n    \"floor\" = ,\n    \"ceiling\" = ,\n    \"trunc\" = ,\n    \"round\" = ,\n    \"signif\" = {\n      units <- attr(x, \"units\")\n      .difftime(NextMethod(), units)\n    },\n    ### otherwise :\n    stop(gettextf(\"'%s' not defined for \\\"difftime\\\" objects\", .Generic),\n      domain = NA\n    )\n  )\n}"},{"path":"s3.html","id":"session-information-10","chapter":"13 S3","heading":"13.7 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  codetools     0.2-19  2023-02-01 [3] CRAN (R 4.3.2)\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  crayon        1.5.2   2022-09-29 [1] RSPM\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr       * 1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  purrr       * 1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  sloop       * 1.0.1   2019-02-17 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"r6.html","id":"r6","chapter":"14 R6","heading":"14 R6","text":"Loading needed libraries:","code":"\nlibrary(R6)"},{"path":"r6.html","id":"classes-and-methods-exercises-14.2.6","chapter":"14 R6","heading":"14.1 Classes and methods (Exercises 14.2.6)","text":"Q1. Create bank account R6 class stores balance allows deposit withdraw money. Create subclass throws error attempt go overdraft. Create another subclass allows go overdraft, charges fee. Create superclass make sure works expected.A1. First, let’s create bank account R6 class stores balance allows deposit withdraw money:Let’s try :Create subclass errors attempt overdraw:Let’s try :Now let’s create subclass charges fee account overdrawn:Let’s try :Q2. Create R6 class represents shuffled deck cards. able draw cards deck $draw(n), return cards deck reshuffle $reshuffle(). Use following code make vector cards.A2. Let’s create needed class represents shuffled deck cards:Let’s try :Q3. can’t model bank account deck cards S3 class?A3. can’t model bank account deck cards S3 class instances classes immutable.hand, R6 classes encapsulate data represent state, can change course object’s lifecycle. words, objects mutable well-suited model bank account.Q4. Create R6 class allows get set current time zone. can access current time zone Sys.timezone() set Sys.setenv(TZ = \"newtimezone\"). setting time zone, make sure new time zone list provided OlsonNames().A4. R6 class manages current time zone:Let’s try :Q5. Create R6 class manages current working directory. $get() $set() methods.A5. R6 class manages current working directory:Let’s create instance class check methods work expected:Q6. can’t model time zone current working directory S3 class?A6. answer Q3:Objects represent real-life entities need mutable S3 class instances mutable.Q7. base type R6 objects built top ? attributes ?A7. Let’s create example class create instance class:R6 objects built top environment:class attribute, character vector \"R6\" last element superclasses elements:","code":"\nlibrary(R6)\n\nbankAccount <- R6::R6Class(\n  \"bankAccount\",\n  public = list(\n    balance = 0,\n    initialize = function(balance) {\n      self$balance <- balance\n    },\n    deposit = function(amount) {\n      self$balance <- self$balance + amount\n      message(paste0(\"Current balance is: \", self$balance))\n      invisible(self)\n    },\n    withdraw = function(amount) {\n      self$balance <- self$balance - amount\n      message(paste0(\"Current balance is: \", self$balance))\n      invisible(self)\n    }\n  )\n)\nindra <- bankAccount$new(balance = 100)\n\nindra$deposit(20)\n#> Current balance is: 120\n\nindra$withdraw(10)\n#> Current balance is: 110\nbankAccountStrict <- R6::R6Class(\n  \"bankAccountStrict\",\n  inherit = bankAccount,\n  public = list(\n    withdraw = function(amount) {\n      if (self$balance - amount < 0) {\n        stop(\n          paste0(\"Can't withdraw more than your current balance: \", self$balance),\n          call. = FALSE\n        )\n      }\n\n      super$withdraw(amount)\n    }\n  )\n)\nPritesh <- bankAccountStrict$new(balance = 100)\n\nPritesh$deposit(20)\n#> Current balance is: 120\n\nPritesh$withdraw(150)\n#> Error: Can't withdraw more than your current balance: 120\nbankAccountFee <- R6::R6Class(\n  \"bankAccountFee\",\n  inherit = bankAccount,\n  public = list(\n    withdraw = function(amount) {\n      super$withdraw(amount)\n\n      if (self$balance) {\n        self$balance <- self$balance - 10\n        message(\"You're withdrawing more than your current balance. You will be charged a fee of 10 euros.\")\n      }\n    }\n  )\n)\nMangesh <- bankAccountFee$new(balance = 100)\n\nMangesh$deposit(20)\n#> Current balance is: 120\n\nMangesh$withdraw(150)\n#> Current balance is: -30\n#> You're withdrawing more than your current balance. You will be charged a fee of 10 euros.\nsuit <- c(\"♠\", \"♥\", \"♦\", \"♣\")\nvalue <- c(\"A\", 2:10, \"J\", \"Q\", \"K\")\ncards <- paste0(rep(value, 4), suit)\nsuit <- c(\"♠\", \"♥\", \"♦\", \"♣\")\nvalue <- c(\"A\", 2:10, \"J\", \"Q\", \"K\")\ncards <- paste(rep(value, 4), suit)\n\nDeck <- R6::R6Class(\n  \"Deck\",\n  public = list(\n    initialize = function(deck) {\n      private$cards <- sample(deck)\n    },\n    draw = function(n) {\n      if (n > length(private$cards)) {\n        stop(\n          paste0(\"Can't draw more than remaining number of cards: \", length(private$cards)),\n          call. = FALSE\n        )\n      }\n\n      drawn_cards <- sample(private$cards, n)\n      private$cards <- private$cards[-which(private$cards %in% drawn_cards)]\n      message(paste0(\"Remaining number of cards: \", length(private$cards)))\n\n      return(drawn_cards)\n    },\n    reshuffle = function() {\n      private$cards <- sample(private$cards)\n      invisible(self)\n    }\n  ),\n  private = list(\n    cards = NULL\n  )\n)\nmyDeck <- Deck$new(cards)\n\nmyDeck$draw(4)\n#> Remaining number of cards: 48\n#> [1] \"2 ♠\"  \"10 ♦\" \"9 ♦\"  \"3 ♦\"\n\nmyDeck$reshuffle()$draw(5)\n#> Remaining number of cards: 43\n#> [1] \"6 ♦\"  \"10 ♥\" \"2 ♥\"  \"A ♥\"  \"8 ♥\"\n\nmyDeck$draw(50)\n#> Error: Can't draw more than remaining number of cards: 43\nCurrentTimeZone <- R6::R6Class(\"CurrentTimeZone\",\n  public = list(\n    setTimeZone = function(tz) {\n      stopifnot(tz %in% OlsonNames())\n      Sys.setenv(TZ = tz)\n    },\n    getTimeZone = function() {\n      Sys.timezone()\n    }\n  )\n)\nmyCurrentTimeZone <- CurrentTimeZone$new()\n\nmyCurrentTimeZone$getTimeZone()\n#> [1] \"UTC\"\n\nmyCurrentTimeZone$setTimeZone(\"Asia/Kolkata\")\nmyCurrentTimeZone$getTimeZone()\n#> [1] \"Asia/Kolkata\"\n\nmyCurrentTimeZone$setTimeZone(\"Europe/Berlin\")\nManageDirectory <- R6::R6Class(\"ManageDirectory\",\n  public = list(\n    setWorkingDirectory = function(dir) {\n      setwd(dir)\n    },\n    getWorkingDirectory = function() {\n      getwd()\n    }\n  )\n)\nmyDirManager <- ManageDirectory$new()\n\n# current working directory\nmyDirManager$getWorkingDirectory()\n\n# change and check if that worked\nmyDirManager$setWorkingDirectory(\"..\")\nmyDirManager$getWorkingDirectory()\n\n# revert this change\nmyDirManager$setWorkingDirectory(\"/Advanced-R-exercises\")\nExample <- R6::R6Class(\"Example\")\nmyExample <- Example$new()\ntypeof(myExample)\n#> [1] \"environment\"\n\nrlang::env_print(myExample)\n#> <environment: 0x5565e0c0a290> [L]\n#> Parent: <environment: empty>\n#> Class: Example, R6\n#> Bindings:\n#> • .__enclos_env__: <env>\n#> • clone: <fn> [L]\nattributes(myExample)\n#> $class\n#> [1] \"Example\" \"R6\""},{"path":"r6.html","id":"controlling-access-exercises-14.3.3","chapter":"14 R6","heading":"14.2 Controlling access (Exercises 14.3.3)","text":"Q1. Create bank account class prevents directly setting account balance, can still withdraw deposit . Throw error attempt go overdraft.A1. bank account class satisfies specified requirements:Let’s check works expected:Q2. Create class write-$password field. $check_password(password) method returns TRUE FALSE, way view complete password.A2. implementation class needed properties:, course, everything possible:Q3. Extend Rando class another active binding allows access previous random value. Ensure active binding way access value.A3. modified version Rando class meet specified requirements:Let’s try :Q4. Can subclasses access private fields/methods parent? Perform experiment find .A4. Unlike common OOP languages (e.g. C++), R6 subclasses (derived classes) also access private methods superclass (base class).instance, following example, Duck class private method $quack(), subclass Mallard can access using super$quack().","code":"\nSafeBankAccount <- R6::R6Class(\n  classname = \"SafeBankAccount\",\n  public = list(\n    deposit = function(deposit_amount) {\n      private$.balance <- private$.balance + deposit_amount\n      print(paste(\"Current balance:\", private$.balance))\n\n      invisible(self)\n    },\n    withdraw = function(withdrawal_amount) {\n      if (withdrawal_amount > private$.balance) {\n        stop(\"You can't withdraw more than your current balance.\", call. = FALSE)\n      }\n\n      private$.balance <- private$.balance - withdrawal_amount\n      print(paste(\"Current balance:\", private$.balance))\n\n      invisible(self)\n    }\n  ),\n  private = list(\n    .balance = 0\n  )\n)\nmySafeBankAccount <- SafeBankAccount$new()\n\nmySafeBankAccount$deposit(100)\n#> [1] \"Current balance: 100\"\n\nmySafeBankAccount$withdraw(50)\n#> [1] \"Current balance: 50\"\n\nmySafeBankAccount$withdraw(100)\n#> Error: You can't withdraw more than your current balance.\nlibrary(R6)\n\ncheckCredentials <- R6Class(\n  \"checkCredentials\",\n  public = list(\n    # setter\n    set_password = function(password) {\n      private$.password <- password\n    },\n\n    # checker\n    check_password = function(password) {\n      if (is.null(private$.password)) {\n        stop(\"No password set to check against.\")\n      }\n\n      identical(password, private$.password)\n    },\n\n    # the default print method prints the private fields as well\n    print = function() {\n      cat(\"Password: XXXX\")\n\n      # for method chaining\n      invisible(self)\n    }\n  ),\n  private = list(\n    .password = NULL\n  )\n)\n\nmyCheck <- checkCredentials$new()\n\nmyCheck$set_password(\"1234\")\nprint(myCheck)\n#> Password: XXXX\n\nmyCheck$check_password(\"abcd\")\n#> [1] FALSE\nmyCheck$check_password(\"1234\")\n#> [1] TRUE\nmyCheck$.__enclos_env__$private$.password\n#> [1] \"1234\"\nRando <- R6::R6Class(\"Rando\",\n  active = list(\n    random = function(value) {\n      if (missing(value)) {\n        newValue <- runif(1)\n        private$.previousRandom <- private$.currentRandom\n        private$.currentRandom <- newValue\n        return(private$.currentRandom)\n      } else {\n        stop(\"Can't set `$random`\", call. = FALSE)\n      }\n    },\n    previousRandom = function(value) {\n      if (missing(value)) {\n        if (is.null(private$.previousRandom)) {\n          message(\"No random value has been generated yet.\")\n        } else {\n          return(private$.previousRandom)\n        }\n      } else {\n        stop(\"Can't set `$previousRandom`\", call. = FALSE)\n      }\n    }\n  ),\n  private = list(\n    .currentRandom = NULL,\n    .previousRandom = NULL\n  )\n)\nmyRando <- Rando$new()\n\n# first time\nmyRando$random\n#> [1] 0.5549124\nmyRando$previousRandom\n#> No random value has been generated yet.\n#> NULL\n\n# second time\nmyRando$random\n#> [1] 0.3482785\nmyRando$previousRandom\n#> [1] 0.5549124\n\n# third time\nmyRando$random\n#> [1] 0.2187275\nmyRando$previousRandom\n#> [1] 0.3482785\nDuck <- R6Class(\"Duck\",\n  private = list(quack = function() print(\"Quack Quack\"))\n)\n\nMallard <- R6Class(\"Mallard\",\n  inherit = Duck,\n  public = list(quack = function() super$quack())\n)\n\nmyMallard <- Mallard$new()\nmyMallard$quack()\n#> [1] \"Quack Quack\""},{"path":"r6.html","id":"reference-semantics-exercises-14.4.4","chapter":"14 R6","heading":"14.3 Reference semantics (Exercises 14.4.4)","text":"Q1. Create class allows write line specified file. open connection file $initialize(), append line using cat() $append_line(), close connection $finalize().A1. class allows write line specified file:Let’s check works expected:","code":"\nfileEditor <- R6Class(\n  \"fileEditor\",\n  public = list(\n    initialize = function(filePath) {\n      private$.connection <- file(filePath, open = \"wt\")\n    },\n    append_line = function(text) {\n      cat(\n        text,\n        file = private$.connection,\n        sep = \"\\n\",\n        append = TRUE\n      )\n    }\n  ),\n  private = list(\n    .connection = NULL,\n    # according to R6 docs, the destructor method should be private\n    finalize = function() {\n      print(\"Closing the file connection!\")\n      close(private$.connection)\n    }\n  )\n)\ngreetMom <- function() {\n  f <- tempfile()\n  myfileEditor <- fileEditor$new(f)\n\n  readLines(f)\n\n  myfileEditor$append_line(\"Hi mom!\")\n  myfileEditor$append_line(\"It's a beautiful day!\")\n\n  readLines(f)\n}\n\ngreetMom()\n#> [1] \"Hi mom!\"               \"It's a beautiful day!\"\n\n# force garbage collection\ngc()\n#> [1] \"Closing the file connection!\"\n#>           used (Mb) gc trigger (Mb) max used (Mb)\n#> Ncells  746818 39.9    1365334   73  1365334 73.0\n#> Vcells 1372826 10.5    8388608   64  2594870 19.8"},{"path":"r6.html","id":"session-information-11","chapter":"14 R6","heading":"14.4 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       Europe/Berlin\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  R6          * 2.5.1   2021-08-19 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"s4.html","id":"s4","chapter":"15 S4","heading":"15 S4","text":"","code":""},{"path":"s4.html","id":"basics-exercises-15.2.1","chapter":"15 S4","heading":"15.1 Basics (Exercises 15.2.1)","text":"Q1. lubridate::period() returns S4 class. slots ? class slot? accessors provide?A1. Let’s first create instance Period class:following slots:Additionally, base type slot (numeric) can seen str() output:lubridate package provides accessors slots:Q2. ways can find help method? Read ?\"?\" summarise details.A2. \"?\" operator allows access documentation three ways. demonstrate different ways access documentation, let’s define new S4 class.Ways access documentation:general documentation generic can found ?topic:expression type?topic look overall documentation methods function f.","code":"\nlibrary(lubridate)\nx <- lubridate::period(c(2, 43, 6), c(\"hour\", \"second\", \"minute\"))\nx\n#> [1] \"2H 6M 43S\"\nslotNames(x)\n#> [1] \".Data\"  \"year\"   \"month\"  \"day\"    \"hour\"   \"minute\"\nstr(x)\n#> Formal class 'Period' [package \"lubridate\"] with 6 slots\n#>   ..@ .Data : num 43\n#>   ..@ year  : num 0\n#>   ..@ month : num 0\n#>   ..@ day   : num 0\n#>   ..@ hour  : num 2\n#>   ..@ minute: num 6\nyear(x)\n#> [1] 0\nmonth(x)\n#> [1] 0\nday(x)\n#> [1] 0\nhour(x)\n#> [1] 2\nminute(x)\n#> [1] 6\nsecond(x)\n#> [1] 43\npow <- function(x, exp) c(x, exp)\nsetGeneric(\"pow\")\n#> [1] \"pow\"\nsetMethod(\"pow\", c(\"numeric\", \"numeric\"), function(x, exp) x^exp)\n?pow\n?pow # produces the function documentation\n\nmethods?pow # looks for the overall methods documentation"},{"path":"s4.html","id":"classes-exercises-15.3.6","chapter":"15 S4","heading":"15.2 Classes (Exercises 15.3.6)","text":"Q1. Extend Person class fields match utils::person(). Think slots need, class slot , ’ll need check validity method.A1. code extends Person class described book match closely utils::person().Let’s make sure validation works expected:Q2. happens define new S4 class doesn’t slots? (Hint: read virtual classes ?setClass.)A2. define new S4 class doesn’t slots, create virtual classes:can’t create instance class:useful? mentioned ?setClass docs:Classes exist actual objects can created, virtual classes.common useful form virtual class class union, virtual class defined call setClassUnion() rather call setClass().virtual classes can still inherited:addition specifying slots, another way create virtual classes:Calls setClass() also create virtual class, either Class argument supplied (slots superclasses) contains= argument includes special class name \"VIRTUAL\".Q3. Imagine going reimplement factors, dates, data frames S4. Sketch setClass() calls use define classes. Think appropriate slots prototype.A3. reimplementation following classes S4 might definitions like following.factorFor simplicity, won’t provide options factor() provides. Note x pseudo-class accept objects type.DateJust like base-R version, integer values.data.frameThe tricky part supporting ... argument data.frame(). , can let users pass (named) list.","code":"\nsetClass(\"Person\",\n  slots = c(\n    age     = \"numeric\",\n    given   = \"character\",\n    family  = \"character\",\n    middle  = \"character\",\n    email   = \"character\",\n    role    = \"character\",\n    comment = \"character\"\n  ),\n  prototype = list(\n    age     = NA_real_,\n    given   = NA_character_,\n    family  = NA_character_,\n    middle  = NA_character_,\n    email   = NA_character_,\n    role    = NA_character_,\n    comment = NA_character_\n  )\n)\n\n# Helper function to create an instance of the `Person` class\nPerson <- function(given,\n                   family,\n                   middle = NA_character_,\n                   age = NA_real_,\n                   email = NA_character_,\n                   role = NA_character_,\n                   comment = NA_character_) {\n  age <- as.double(age)\n\n  new(\"Person\",\n    age     = age,\n    given   = given,\n    family  = family,\n    middle  = middle,\n    email   = email,\n    role    = role,\n    comment = comment\n  )\n}\n\n# Validator to ensure that each slot is of length one and that the specified\n# role is one of the possible roles\nsetValidity(\"Person\", function(object) {\n  invalid_length <- NULL\n  slot_lengths <- c(\n    length(object@age),\n    length(object@given),\n    length(object@middle),\n    length(object@family),\n    length(object@email),\n    length(object@comment)\n  )\n\n  if (any(slot_lengths > 1L)) {\n    invalid_length <- \"\\nFollowing slots must be of length 1:\\n @age, @given, @family, @middle, @email, @comment\"\n  }\n\n  possible_roles <- c(\n    NA_character_, \"aut\", \"com\", \"cph\", \"cre\", \"ctb\", \"ctr\", \"dtc\", \"fnd\", \"rev\", \"ths\", \"trl\"\n  )\n\n  if (any(!object@role %in% possible_roles)) {\n    invalid_length <- paste(\n      invalid_length,\n      \"\\nSlot @role(s) must be one of the following:\\n\",\n      paste(possible_roles, collapse = \", \")\n    )\n  }\n\n  if (!is.null(invalid_length)) {\n    return(invalid_length)\n  } else {\n    return(TRUE)\n  }\n})\n#> Class \"Person\" [in \".GlobalEnv\"]\n#> \n#> Slots:\n#>                                                         \n#> Name:        age     given    family    middle     email\n#> Class:   numeric character character character character\n#>                           \n#> Name:       role   comment\n#> Class: character character\n# length of first argument not 1\nPerson(c(\"Indrajeet\", \"Surendra\"), \"Patil\")\n#> Error in validObject(.Object): invalid class \"Person\" object: \n#> Following slots must be of length 1:\n#>  @age, @given, @family, @middle, @email, @comment\n\n# role not recognized\nPerson(\"Indrajeet\", \"Patil\", role = \"xyz\")\n#> Error in validObject(.Object): invalid class \"Person\" object:  \n#> Slot @role(s) must be one of the following:\n#>  NA, aut, com, cph, cre, ctb, ctr, dtc, fnd, rev, ths, trl\n\n# all okay\nPerson(\"Indrajeet\", \"Patil\", role = c(\"aut\", \"cph\"))\n#> An object of class \"Person\"\n#> Slot \"age\":\n#> [1] NA\n#> \n#> Slot \"given\":\n#> [1] \"Indrajeet\"\n#> \n#> Slot \"family\":\n#> [1] \"Patil\"\n#> \n#> Slot \"middle\":\n#> [1] NA\n#> \n#> Slot \"email\":\n#> [1] NA\n#> \n#> Slot \"role\":\n#> [1] \"aut\" \"cph\"\n#> \n#> Slot \"comment\":\n#> [1] NA\nsetClass(\"Empty\")\n\nisVirtualClass(\"Empty\")\n#> [1] TRUE\nnew(\"Empty\")\n#> Error in new(\"Empty\"): trying to generate an object from a virtual class (\"Empty\")\nsetClass(\"Nothing\", contains = \"Empty\")\nsetClass(\"Factor\",\n  slots = c(\n    x       = \"ANY\",\n    levels  = \"character\",\n    ordered = \"logical\"\n  ),\n  prototype = list(\n    x       = character(),\n    levels  = character(),\n    ordered = FALSE\n  )\n)\n\nnew(\"Factor\", x = letters[1:3], levels = LETTERS[1:3])\n#> An object of class \"Factor\"\n#> Slot \"x\":\n#> [1] \"a\" \"b\" \"c\"\n#> \n#> Slot \"levels\":\n#> [1] \"A\" \"B\" \"C\"\n#> \n#> Slot \"ordered\":\n#> [1] FALSE\n\nnew(\"Factor\", x = 1:3, levels = letters[1:3])\n#> An object of class \"Factor\"\n#> Slot \"x\":\n#> [1] 1 2 3\n#> \n#> Slot \"levels\":\n#> [1] \"a\" \"b\" \"c\"\n#> \n#> Slot \"ordered\":\n#> [1] FALSE\n\nnew(\"Factor\", x = c(TRUE, FALSE, TRUE), levels = c(\"x\", \"y\", \"x\"))\n#> An object of class \"Factor\"\n#> Slot \"x\":\n#> [1]  TRUE FALSE  TRUE\n#> \n#> Slot \"levels\":\n#> [1] \"x\" \"y\" \"x\"\n#> \n#> Slot \"ordered\":\n#> [1] FALSE\nsetClass(\"Date2\",\n  slots = list(\n    data = \"integer\"\n  ),\n  prototype = list(\n    data = integer()\n  )\n)\n\nnew(\"Date2\", data = 1342L)\n#> An object of class \"Date2\"\n#> Slot \"data\":\n#> [1] 1342\nsetClass(\"DataFrame\",\n  slots = c(\n    data      = \"list\",\n    row.names = \"character\"\n  ),\n  prototype = list(\n    data      = list(),\n    row.names = character(0L)\n  )\n)\n\nnew(\"DataFrame\", data = list(x = c(\"a\", \"b\"), y = c(1L, 2L)))\n#> An object of class \"DataFrame\"\n#> Slot \"data\":\n#> $x\n#> [1] \"a\" \"b\"\n#> \n#> $y\n#> [1] 1 2\n#> \n#> \n#> Slot \"row.names\":\n#> character(0)"},{"path":"s4.html","id":"generics-and-methods-exercises-15.4.5","chapter":"15 S4","heading":"15.3 Generics and methods (Exercises 15.4.5)","text":"Q1. Add age() accessors Person class.A1. first define generic method class:Q2. definition generic, necessary repeat name generic twice?A2. Let’s look generic just defined; generic name \"age\" repeated twice.:\"age\" passed argument name provides name genericthe \"age\" passed argument def supplies method dispatchThis reminiscent defined S3 generic, also repeat name twice:Q3. show() method defined Section Show method use (object)[[1]]? (Hint: try printing employee subclass.)A3. wish define show() method specific class, need disregard super-/sub-classes.Always using first element ensures method defined class question:Q4. happens define method different argument names generic?A4. Let’s experiment method defined Q1. study behavior.original method worked expected since argument name generic method matched:case, either get warning get error depending many arguments specified:","code":"\nIndra <- Person(\"Indrajeet\", \"Patil\", role = c(\"aut\", \"cph\"), age = 34)\n\nsetGeneric(\"age\", function(x) standardGeneric(\"age\"))\n#> [1] \"age\"\nsetMethod(\"age\", \"Person\", function(x) x@age)\n\nage(Indra)\n#> [1] 34\nsetGeneric(name = \"age\", def = function(x) standardGeneric(\"age\"))\nage <- function(x) {\n  UseMethod(\"age\")\n}\nAlice <- new(\"Employee\")\n\nis(Alice)\n#> [1] \"Employee\" \"Person\"\n\nis(Alice)[[1]]\n#> [1] \"Employee\"\nsetMethod(\"age\", \"Person\", function(x) x@age)\nsetMethod(\"age\", \"Person\", function(object) object@age)\n#> Warning: For function 'age', signature 'Person': argument\n#> in method definition changed from (object) to (x)\n\nsetMethod(\"age\", \"Person\", function(object, x) object@age)\n#> Error in rematchDefinition(definition, fdef, mnames, fnames, signature): methods can add arguments to the generic 'age' only if '...' is an argument to the generic\n\nsetMethod(\"age\", \"Person\", function(...) ...elt(1)@age)\n#> Warning: For function 'age', signature 'Person': argument\n#> in method definition changed from (...) to (x)\n\nsetMethod(\"age\", \"Person\", function(x, ...) x@age)\n#> Error in rematchDefinition(definition, fdef, mnames, fnames, signature): methods can add arguments to the generic 'age' only if '...' is an argument to the generic"},{"path":"s4.html","id":"method-dispatch-exercises-15.5.5","chapter":"15 S4","heading":"15.4 Method dispatch (Exercises 15.5.5)","text":"Q1. Draw method graph f(😅, 😽).A1. don’t prepare visual illustrations used book, linking illustration official solution manual:Q2. Draw method graph f(😃, 😉, 😙).A2. don’t prepare visual illustrations used book, linking illustration official solution manual:Q3. Take last example shows multiple dispatch two classes use multiple inheritance. happens define method terminal classes? method dispatch save us much work ?A3. one class distance 2 terminal nodes four distance 1 two terminal nodes , introduce ambiguity.Method dispatch save us much work resolve ambiguity define five methods (one per class combination).","code":""},{"path":"s4.html","id":"s4-and-s3-exercises-15.6.3","chapter":"15 S4","heading":"15.5 S4 and S3 (Exercises 15.6.3)","text":"Q1. full setOldClass() definition look like ordered factor (.e. add slots prototype definition )?A1. can register old-style/S3 ordered class formally defined class using setOldClass().Let’s use see works expected.Q2. Define length method Person class.A2. Person class can used create objects represent multiple people, let’s say length() method returns many persons object.can define S3 method class:Alternatively, can also write S4 method:","code":"\nsetClass(\"factor\",\n  contains = \"integer\",\n  slots = c(\n    levels = \"character\"\n  ),\n  prototype = structure(\n    integer(),\n    levels = character()\n  )\n)\nsetOldClass(\"factor\", S4Class = \"factor\")\n#> Warning in rm(list = what, pos = classWhere): object\n#> '.__C__factor' not found\n\nsetClass(\"Ordered\",\n  contains = \"factor\",\n  slots = c(\n    levels  = \"character\",\n    ordered = \"logical\"\n  ),\n  prototype = structure(\n    integer(),\n    levels  = character(),\n    ordered = logical()\n  )\n)\n\nsetOldClass(\"ordered\", S4Class = \"Ordered\")\nx <- new(\"Ordered\", 1L:4L, levels = letters[1:4], ordered = TRUE)\n\nx\n#> Object of class \"Ordered\"\n#> [1] a b c d\n#> Levels: a b c d\n#> Slot \"ordered\":\n#> [1] TRUE\n\nstr(x)\n#> Formal class 'Ordered' [package \".GlobalEnv\"] with 4 slots\n#>   ..@ .Data   : int [1:4] 1 2 3 4\n#>   ..@ levels  : chr [1:4] \"a\" \"b\" \"c\" \"d\"\n#>   ..@ ordered : logi TRUE\n#>   ..@ .S3Class: chr \"factor\"\n\nclass(x)\n#> [1] \"Ordered\"\n#> attr(,\"package\")\n#> [1] \".GlobalEnv\"\nFriends <- new(\"Person\", name = c(\"Vishu\", \"Aditi\"))\nlength.Person <- function(x) length(x@name)\n\nlength(Friends)\n#> [1] 2\nsetMethod(\"length\", \"Person\", function(x) length(x@name))\n\nlength(Friends)\n#> [1] 2"},{"path":"s4.html","id":"session-information-12","chapter":"15 S4","heading":"15.6 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version    date (UTC) lib source\n#>  assertthat    0.2.1      2019-03-21 [1] RSPM\n#>  base        * 4.3.2      2023-11-01 [3] local\n#>  bookdown      0.37       2023-12-01 [1] RSPM\n#>  bslib         0.6.1      2023-11-28 [1] RSPM\n#>  cachem        1.0.8      2023-05-01 [1] RSPM\n#>  cli           3.6.2      2023-12-11 [1] RSPM\n#>  compiler      4.3.2      2023-11-01 [3] local\n#>  crayon        1.5.2      2022-09-29 [1] RSPM\n#>  datasets    * 4.3.2      2023-11-01 [3] local\n#>  digest        0.6.33     2023-07-07 [1] RSPM\n#>  downlit       0.4.3      2023-06-29 [1] RSPM\n#>  emo           0.0.0.9000 2023-11-05 [1] Github (hadley/emo@3f03b11)\n#>  evaluate      0.23       2023-11-01 [1] RSPM\n#>  fastmap       1.1.1      2023-02-24 [1] RSPM\n#>  fs            1.6.3      2023-07-20 [1] RSPM\n#>  generics      0.1.3      2022-07-05 [1] RSPM\n#>  glue          1.6.2      2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2      2023-11-01 [3] local\n#>  grDevices   * 4.3.2      2023-11-01 [3] local\n#>  highr         0.10       2022-12-22 [1] RSPM\n#>  htmltools     0.5.7      2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4      2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8      2023-12-04 [1] RSPM\n#>  knitr         1.45       2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4      2023-11-07 [1] RSPM\n#>  lubridate   * 1.9.3      2023-09-27 [1] RSPM\n#>  magrittr    * 2.0.3      2022-03-30 [1] RSPM\n#>  memoise       2.0.1      2021-11-26 [1] RSPM\n#>  methods     * 4.3.2      2023-11-01 [3] local\n#>  purrr         1.0.2      2023-08-10 [1] RSPM\n#>  R6            2.5.1      2021-08-19 [1] RSPM\n#>  rlang         1.1.2      2023-11-04 [1] RSPM\n#>  rmarkdown     2.25       2023-09-18 [1] RSPM\n#>  sass          0.4.8      2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2      2021-12-06 [1] RSPM\n#>  stats       * 4.3.2      2023-11-01 [3] local\n#>  stringi       1.8.3      2023-12-11 [1] RSPM\n#>  stringr       1.5.1      2023-11-14 [1] RSPM\n#>  timechange    0.2.0      2023-01-11 [1] RSPM\n#>  tools         4.3.2      2023-11-01 [3] local\n#>  utils       * 4.3.2      2023-11-01 [3] local\n#>  vctrs         0.6.5      2023-12-01 [1] RSPM\n#>  withr         2.5.2      2023-10-30 [1] RSPM\n#>  xfun          0.41       2023-11-01 [1] RSPM\n#>  xml2          1.3.6      2023-12-04 [1] RSPM\n#>  yaml          2.3.8      2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"trade-offs.html","id":"trade-offs","chapter":"16 Trade-offs","heading":"16 Trade-offs","text":"exercises.","code":""},{"path":"big-picture.html","id":"big-picture","chapter":"17 Big Picture","heading":"17 Big Picture","text":"exercises.","code":""},{"path":"expressions.html","id":"expressions","chapter":"18 Expressions","heading":"18 Expressions","text":"Attaching needed libraries:","code":"\nlibrary(rlang, warn.conflicts = FALSE)\nlibrary(lobstr, warn.conflicts = FALSE)"},{"path":"expressions.html","id":"abstract-syntax-trees-exercises-18.2.4","chapter":"18 Expressions","heading":"18.1 Abstract syntax trees (Exercises 18.2.4)","text":"Q1. Reconstruct code represented trees :A1. reconstructed code.can confirm drawing ASTs :Q2. Draw following trees hand check answers ast().A2. Successfully drawn hand. Checking using ast():Q3. ’s happening ASTs ? (Hint: carefully read ?\"^\".)A3. str2expression() helps make sense ASTs.non-syntactic names parsed names. Thus, backticks removed AST.mentioned docs ^:** translated parser ^rightward assignment parsed leftward assignment:Q4. special AST ?A4. mentioned section:Like objects R, functions can also possess number additional attributes(). One attribute used base R srcref, short source reference. points source code used create function. srcref used printing , unlike body(), contains code comments formatting.Therefore, last leaf AST, although specified function call, represents source reference attribute.Q5. call tree statement multiple else conditions look like? ?A5. nothing special tree. just shows nested loop structure inherent code multiple else statements.","code":"#> █─f \n#> └─█─g \n#>   └─█─h\n#> █─`+` \n#> ├─█─`+` \n#> │ ├─1 \n#> │ └─2 \n#> └─3\n#> █─`*` \n#> ├─█─`(` \n#> │ └─█─`+` \n#> │   ├─x \n#> │   └─y \n#> └─z\nf(g(h()))\n1 + 2 + 3\n(x + y) * z\nast(f(g(h())))\n#> █─f \n#> └─█─g \n#>   └─█─h\n\nast(1 + 2 + 3)\n#> █─`+` \n#> ├─█─`+` \n#> │ ├─1 \n#> │ └─2 \n#> └─3\n\nast((x + y) * z)\n#> █─`*` \n#> ├─█─`(` \n#> │ └─█─`+` \n#> │   ├─x \n#> │   └─y \n#> └─z\nf(g(h(i(1, 2, 3))))\nf(1, g(2, h(3, i())))\nf(g(1, 2), h(3, i(4, 5)))\nast(f(g(h(i(1, 2, 3)))))\n#> █─f \n#> └─█─g \n#>   └─█─h \n#>     └─█─i \n#>       ├─1 \n#>       ├─2 \n#>       └─3\n\nast(f(1, g(2, h(3, i()))))\n#> █─f \n#> ├─1 \n#> └─█─g \n#>   ├─2 \n#>   └─█─h \n#>     ├─3 \n#>     └─█─i\n\nast(f(g(1, 2), h(3, i(4, 5))))\n#> █─f \n#> ├─█─g \n#> │ ├─1 \n#> │ └─2 \n#> └─█─h \n#>   ├─3 \n#>   └─█─i \n#>     ├─4 \n#>     └─5\nast(`x` + `y`)\n#> █─`+` \n#> ├─x \n#> └─y\nast(x**y)\n#> █─`^` \n#> ├─x \n#> └─y\nast(1 -> x)\n#> █─`<-` \n#> ├─x \n#> └─1\nstr2expression(\"`x` + `y`\")\n#> expression(x + y)\nstr2expression(\"x**y\")\n#> expression(x^y)\nstr2expression(\"1 -> x\")\n#> expression(x <- 1)\nast(function(x = 1, y = 2) {})\n#> █─`function` \n#> ├─█─x = 1 \n#> │ └─y = 2 \n#> ├─█─`{` \n#> └─<inline srcref>\nast(if (FALSE) 1 else if (FALSE) 2 else if (FALSE) 3 else 4)\n#> █─`if` \n#> ├─FALSE \n#> ├─1 \n#> └─█─`if` \n#>   ├─FALSE \n#>   ├─2 \n#>   └─█─`if` \n#>     ├─FALSE \n#>     ├─3 \n#>     └─4"},{"path":"expressions.html","id":"expressions-exercises-18.3.5","chapter":"18 Expressions","heading":"18.2 Expressions (Exercises 18.3.5)","text":"Q1. two six types atomic vector can’t appear expression? ? Similarly, can’t create expression contains atomic vector length greater one?A1. six types atomic vectors, two can’t appear expression : complex raw.Complex numbers created via function call (using +), can seen AST:Similarly, raw vectors (using raw()):Contrast atomic vectors:reason, can’t create expression contains atomic vector length greater one since function call uses c() function:Q2. happens subset call object remove first element? e.g. expr(read.csv(\"foo.csv\", header = TRUE))[-1]. ?A2. captured function call like following creates call object:mentioned respective section:first element call object function position.Therefore, first element call object removed, next one moves function position, get observed output:Q3. Describe differences following call objects.A4. differences constructed call objects due different type arguments supplied first two parameters call2() function.Types arguments supplied .fn:Types arguments supplied dynamic dots:following outputs can understood using following properties:.fn argument closure, function inlined constructed function callwhen x symbol, value passed function callImportantly, constructed call objects evaluate give result:Q4. call_standardise() doesn’t work well following calls. ? makes mean() special?A4. ellipsis mean() function signature:mentioned respective section:function uses ... ’s possible standardise arguments.mean() S3 generic dots passed underlying S3 methods., output can improved using specific method. example:Q5. code make sense?A5. doesn’t make sense first position call object reserved function (function position), assigning names element just ignored R:Q6. Construct expression (x > 1) \"\" else \"b\" using multiple calls call2(). code structure reflect structure AST?A6. Using multiple calls construct required expression:construction follows prefix form expression, revealed AST:","code":"\nx_complex <- expr(1 + 1i)\ntypeof(x_complex)\n#> [1] \"language\"\n\nast(1 + 1i)\n#> █─`+` \n#> ├─1 \n#> └─1i\nx_raw <- expr(raw(2))\ntypeof(x_raw)\n#> [1] \"language\"\n\nast(raw(2))\n#> █─raw \n#> └─2\nx_int <- expr(2L)\ntypeof(x_int)\n#> [1] \"integer\"\n\nast(2L)\n#> 2L\nx_vec <- expr(c(1, 2))\ntypeof(x_vec)\n#> [1] \"language\"\n\nast(c(1, 2))\n#> █─c \n#> ├─1 \n#> └─2\nexpr(read.csv(\"foo.csv\", header = TRUE))\n#> read.csv(\"foo.csv\", header = TRUE)\n\ntypeof(expr(read.csv(\"foo.csv\", header = TRUE)))\n#> [1] \"language\"\nexpr(read.csv(\"foo.csv\", header = TRUE))[-1]\n#> \"foo.csv\"(header = TRUE)\nx <- 1:10\ncall2(median, x, na.rm = TRUE)\ncall2(expr(median), x, na.rm = TRUE)\ncall2(median, expr(x), na.rm = TRUE)\ncall2(expr(median), expr(x), na.rm = TRUE)\ntypeof(median)\n#> [1] \"closure\"\ntypeof(expr(median))\n#> [1] \"symbol\"\nx <- 1:10\ntypeof(x)\n#> [1] \"integer\"\ntypeof(expr(x))\n#> [1] \"symbol\"\nx <- 1:10\n\ncall2(median, x, na.rm = TRUE)\n#> (function (x, na.rm = FALSE, ...) \n#> UseMethod(\"median\"))(1:10, na.rm = TRUE)\n\ncall2(expr(median), x, na.rm = TRUE)\n#> median(1:10, na.rm = TRUE)\n\ncall2(median, expr(x), na.rm = TRUE)\n#> (function (x, na.rm = FALSE, ...) \n#> UseMethod(\"median\"))(x, na.rm = TRUE)\n\ncall2(expr(median), expr(x), na.rm = TRUE)\n#> median(x, na.rm = TRUE)\nx <- 1:10\n\neval(call2(median, x, na.rm = TRUE))\n#> [1] 5.5\n\neval(call2(expr(median), x, na.rm = TRUE))\n#> [1] 5.5\n\neval(call2(median, expr(x), na.rm = TRUE))\n#> [1] 5.5\n\neval(call2(expr(median), expr(x), na.rm = TRUE))\n#> [1] 5.5\ncall_standardise(quote(mean(1:10, na.rm = TRUE)))\n#> Warning: `call_standardise()` is deprecated as of rlang 0.4.11\n#> This warning is displayed once every 8 hours.\n#> mean(x = 1:10, na.rm = TRUE)\ncall_standardise(quote(mean(n = T, 1:10)))\n#> mean(x = 1:10, n = T)\ncall_standardise(quote(mean(x = 1:10, , TRUE)))\n#> mean(x = 1:10, , TRUE)\nmean\n#> function (x, ...) \n#> UseMethod(\"mean\")\n#> <bytecode: 0x562fc13895a8>\n#> <environment: namespace:base>\ncall_standardise(quote(mean.default(n = T, 1:10)))\n#> mean.default(x = 1:10, na.rm = T)\nx <- expr(foo(x = 1))\nnames(x) <- c(\"x\", \"y\")\nx <- expr(foo(x = 1))\nx\n#> foo(x = 1)\n\nnames(x) <- c(\"x\", \"y\")\nx\n#> foo(y = 1)\nx <- 5\ncall_obj1 <- call2(\">\", expr(x), 1)\ncall_obj1\n#> x > 1\n\ncall_obj2 <- call2(\"if\", cond = call_obj1, cons.expr = \"a\", alt.expr = \"b\")\ncall_obj2\n#> if (x > 1) \"a\" else \"b\"\nast(if (x > 1) \"a\" else \"b\")\n#> █─`if` \n#> ├─█─`>` \n#> │ ├─x \n#> │ └─1 \n#> ├─\"a\" \n#> └─\"b\""},{"path":"expressions.html","id":"parsing-and-grammar-exercises-18.4.4","chapter":"18 Expressions","heading":"18.3 Parsing and grammar (Exercises 18.4.4)","text":"Q1. R uses parentheses two slightly different ways illustrated two calls:Compare contrast two uses referencing AST.A1. Let’s first look AST:, can see ( used two separate ways:function right \"`(`\"part prefix syntax (f()), AST f((1)), see one \"`(`\" (first use case), f(), part function syntax (second use case).Q2. = can also used two ways. Construct simple example shows uses.A2. simple example illustrating = can also used two ways:assignmentfor named arguments function callsWe can also look AST:Q3. -2^2 yield 4 -4? ?A3. expression -2^2 evaluates -4 operator ^ higher precedence unary - operator:can also seen AST:less confusing way write :Q4. !1 + !1 return? ?A3. expression !1 + !1 evaluates FALSE.! operator higher precedence unary + operator. Thus, !1 evaluates FALSE, added 1 + FALSE, evaluates 1, logically negated !1, FALSE.can easily seen AST:Q5. x1 <- x2 <- x3 <- 0 work? Describe two reasons.A5. two reasons following works expected:<- operator right associative.Therefore, order assignment :<- operator invisibly returns assigned value.easy surmise AST:Q6. Compare ASTs x + y %+% z x ^ y %+% z. learned precedence custom infix functions?A6. Looking ASTs expressions,can say custom infix operator %+% :higher precedence + operatorlower precedence ^ operatorQ7. happens call parse_expr() string generates multiple expressions? e.g. parse_expr(\"x + 1; y + 1\")A7. produced error:expected based docs:parse_expr() returns one expression. text contains one expression (separated semicolons new lines), error issued.instead need use parse_exprs():Q8. happens attempt parse invalid expression? e.g. \"+\" \"f())\".A8. invalid expression produces error:Since underlying parse() function produces error:Q9. deparse() produces vectors input long. example, following call produces vector length two:expr_text() instead?A9. difference deparse() expr_text() latter turns (possibly multi-line) expression single string.Q10. pairwise.t.test() assumes deparse() always returns length one character vector. Can construct input violates expectation? happens?A10 Since R 4.0, possible violate expectation since new implementation produces single string matter input:New function deparse1() produces one string, wrapping deparse(), used typically deparse1(substitute(*))","code":"\nf((1))\n`(`(1 + 1)\nast(f((1)))\n#> █─f \n#> └─█─`(` \n#>   └─1\nast(`(`(1 + 1))\n#> █─`(` \n#> └─█─`+` \n#>   ├─1 \n#>   └─1\nm <- mean(x = 1)\nast({\n  m <- mean(x = 1)\n})\n#> █─`{` \n#> └─█─`<-` \n#>   ├─m \n#>   └─█─mean \n#>     └─x = 1\n-2^2\n#> [1] -4\nast(-2^2)\n#> █─`-` \n#> └─█─`^` \n#>   ├─2 \n#>   └─2\n-(2^2)\n#> [1] -4\nast(!1 + !1)\n#> █─`!` \n#> └─█─`+` \n#>   ├─1 \n#>   └─█─`!` \n#>     └─1\nx1 <- x2 <- x3 <- 0\n(x3 <- 0)\n(x2 <- x3)\n(x1 <- x2)\n(x <- 1)\n#> [1] 1\nast(x1 <- x2 <- x3 <- 0)\n#> █─`<-` \n#> ├─x1 \n#> └─█─`<-` \n#>   ├─x2 \n#>   └─█─`<-` \n#>     ├─x3 \n#>     └─0\nast(x + y %+% z)\n#> █─`+` \n#> ├─x \n#> └─█─`%+%` \n#>   ├─y \n#>   └─z\n\nast(x^y %+% z)\n#> █─`%+%` \n#> ├─█─`^` \n#> │ ├─x \n#> │ └─y \n#> └─z\nparse_expr(\"x + 1; y + 1\")\n#> Error in `parse_expr()`:\n#> ! `x` must contain exactly 1 expression, not 2.\nparse_exprs(\"x + 1; y + 1\")\n#> [[1]]\n#> x + 1\n#> \n#> [[2]]\n#> y + 1\nparse_expr(\"a +\")\n#> Error in parse(text = x, keep.source = FALSE): <text>:2:0: unexpected end of input\n#> 1: a +\n#>    ^\n\nparse_expr(\"f())\")\n#> Error in parse(text = x, keep.source = FALSE): <text>:1:4: unexpected ')'\n#> 1: f())\n#>        ^\nparse(text = \"a +\")\n#> Error in parse(text = \"a +\"): <text>:2:0: unexpected end of input\n#> 1: a +\n#>    ^\n\nparse(text = \"f())\")\n#> Error in parse(text = \"f())\"): <text>:1:4: unexpected ')'\n#> 1: f())\n#>        ^\nexpr <- expr(g(a + b + c + d + e + f + g + h + i + j + k + l +\n  m + n + o + p + q + r + s + t + u + v + w + x + y + z))\ndeparse(expr)\nexpr <- expr(g(a + b + c + d + e + f + g + h + i + j + k + l +\n  m + n + o + p + q + r + s + t + u + v + w + x + y + z))\n\ndeparse(expr)\n#> [1] \"g(a + b + c + d + e + f + g + h + i + j + k + l + m + n + o + \"\n#> [2] \"    p + q + r + s + t + u + v + w + x + y + z)\"\n\nexpr_text(expr)\n#> [1] \"g(a + b + c + d + e + f + g + h + i + j + k + l + m + n + o + \\n    p + q + r + s + t + u + v + w + x + y + z)\""},{"path":"expressions.html","id":"walking-ast-with-recursive-functions-exercises-18.5.3","chapter":"18 Expressions","heading":"18.4 Walking AST with recursive functions (Exercises 18.5.3)","text":"Q1. logical_abbr() returns TRUE T(1, 2, 3). modify logical_abbr_rec() ignores function calls use T F?A1. avoid function calls use T F, just need ignore function position call objects:Let’s try :Q2. logical_abbr() works expressions. currently fails give function. ? modify logical_abbr() make work? components function need recurse ?A2. Surprisingly, logical_abbr() currently doesn’t fail closures:see , let’s see type object produced capture user provided closure:Given closures converted call objects, surprise function works:function fails can’t find negative case. example, instead returning FALSE, produces error reasons remain (yet) elusive :Q3. Modify find_assign also detect assignment using replacement functions, .e. names(x) <- y.A3. Although simple assignment (x <- y) assignment using replacement functions (names(x) <- y) <- operator call, latter case, names(x) call object symbol:’s can detect kind assignment checking second element expression symbol language type object.Let’s try :Q4. Write function extracts calls specified function.A4. function extracts calls specified function:","code":"\nlogical_abbr_rec(expr(T(1, 2, 3)))\n#> [1] FALSE\n\nlogical_abbr_rec(expr(F(1, 2, 3)))\n#> [1] FALSE\n\nlogical_abbr_rec(expr(T))\n#> [1] TRUE\n\nlogical_abbr_rec(expr(F))\n#> [1] TRUE\nlogical_abbr(function(x = TRUE) {\n  g(x + T)\n})\nprint_enexpr <- function(.f) {\n  print(typeof(enexpr(.f)))\n  print(is.call(enexpr(.f)))\n}\n\nprint_enexpr(function(x = TRUE) {\n  g(x + T)\n})\n#> [1] \"language\"\n#> [1] TRUE\nlogical_abbr(function(x = TRUE) {\n  g(x + T)\n})\n#> [1] TRUE\nlogical_abbr(function(x = TRUE) {\n  g(x + TRUE)\n})\n#> Error: Don't know how to handle type integer\nexpr1 <- expr(names(x) <- y)\nas.list(expr1)\n#> [[1]]\n#> `<-`\n#> \n#> [[2]]\n#> names(x)\n#> \n#> [[3]]\n#> y\ntypeof(expr1[[2]])\n#> [1] \"language\"\n\nexpr2 <- expr(x <- y)\nas.list(expr2)\n#> [[1]]\n#> `<-`\n#> \n#> [[2]]\n#> x\n#> \n#> [[3]]\n#> y\ntypeof(expr2[[2]])\n#> [1] \"symbol\"\nexpr_type <- function(x) {\n  if (is_syntactic_literal(x)) {\n    \"constant\"\n  } else if (is.symbol(x)) {\n    \"symbol\"\n  } else if (is.call(x)) {\n    \"call\"\n  } else if (is.pairlist(x)) {\n    \"pairlist\"\n  } else {\n    typeof(x)\n  }\n}\n\nswitch_expr <- function(x, ...) {\n  switch(expr_type(x),\n    ...,\n    stop(\"Don't know how to handle type \", typeof(x), call. = FALSE)\n  )\n}\n\nflat_map_chr <- function(.x, .f, ...) {\n  purrr::flatten_chr(purrr::map(.x, .f, ...))\n}\n\nextract_symbol <- function(x) {\n  if (is_symbol(x[[2]])) {\n    as_string(x[[2]])\n  } else {\n    extract_symbol(as.list(x[[2]]))\n  }\n}\n\nfind_assign_call <- function(x) {\n  if (is_call(x, \"<-\") && is_symbol(x[[2]])) {\n    lhs <- as_string(x[[2]])\n    children <- as.list(x)[-1]\n  } else if (is_call(x, \"<-\") && is_call(x[[2]])) {\n    lhs <- extract_symbol(as.list(x[[2]]))\n    children <- as.list(x)[-1]\n  } else {\n    lhs <- character()\n    children <- as.list(x)\n  }\n\n  c(lhs, flat_map_chr(children, find_assign_rec))\n}\n\nfind_assign_rec <- function(x) {\n  switch_expr(x,\n    # Base cases\n    constant = ,\n    symbol = character(),\n\n    # Recursive cases\n    pairlist = flat_map_chr(x, find_assign_rec),\n    call = find_assign_call(x)\n  )\n}\n\nfind_assign <- function(x) find_assign_rec(enexpr(x))\nfind_assign(names(x))\n#> character(0)\n\nfind_assign(names(x) <- y)\n#> [1] \"x\"\n\nfind_assign(names(f(x)) <- y)\n#> [1] \"x\"\n\nfind_assign(names(x) <- y <- z <- NULL)\n#> [1] \"x\" \"y\" \"z\"\n\nfind_assign(a <- b <- c <- 1)\n#> [1] \"a\" \"b\" \"c\"\n\nfind_assign(system.time(x <- print(y <- 5)))\n#> [1] \"x\" \"y\"\nfind_function_call <- function(x, .f) {\n  if (is_call(x)) {\n    if (is_call(x, .f)) {\n      list(x)\n    } else {\n      purrr::map(as.list(x), ~ find_function_call(.x, .f)) %>%\n        purrr::compact() %>%\n        unlist(use.names = FALSE)\n    }\n  }\n}\n\n# example-1: with infix operator `:`\nfind_function_call(expr(mean(1:2)), \":\")\n#> [[1]]\n#> 1:2\n\nfind_function_call(expr(sum(mean(1:2))), \":\")\n#> [[1]]\n#> 1:2\n\nfind_function_call(expr(list(1:5, 4:6, 3:9)), \":\")\n#> [[1]]\n#> 1:5\n#> \n#> [[2]]\n#> 4:6\n#> \n#> [[3]]\n#> 3:9\n\nfind_function_call(expr(list(1:5, sum(4:6), mean(3:9))), \":\")\n#> [[1]]\n#> 1:5\n#> \n#> [[2]]\n#> 4:6\n#> \n#> [[3]]\n#> 3:9\n\n# example-2: with assignment operator `<-`\nfind_function_call(expr(names(x)), \"<-\")\n#> NULL\n\nfind_function_call(expr(names(x) <- y), \"<-\")\n#> [[1]]\n#> names(x) <- y\n\nfind_function_call(expr(names(f(x)) <- y), \"<-\")\n#> [[1]]\n#> names(f(x)) <- y\n\nfind_function_call(expr(names(x) <- y <- z <- NULL), \"<-\")\n#> [[1]]\n#> names(x) <- y <- z <- NULL\n\nfind_function_call(expr(a <- b <- c <- 1), \"<-\")\n#> [[1]]\n#> a <- b <- c <- 1\n\nfind_function_call(expr(system.time(x <- print(y <- 5))), \"<-\")\n#> [[1]]\n#> x <- print(y <- 5)"},{"path":"expressions.html","id":"session-information-13","chapter":"18 Expressions","heading":"18.5 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  crayon        1.5.2   2022-09-29 [1] RSPM\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  lobstr      * 1.1.2   2022-06-22 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  purrr         1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"quasiquotation.html","id":"quasiquotation","chapter":"19 Quasiquotation","heading":"19 Quasiquotation","text":"Attaching needed libraries:","code":"\nlibrary(rlang)\nlibrary(purrr)\nlibrary(lobstr)\nlibrary(dplyr)\nlibrary(ggplot2)"},{"path":"quasiquotation.html","id":"motivation-exercises-19.2.2","chapter":"19 Quasiquotation","heading":"19.1 Motivation (Exercises 19.2.2)","text":"Q1. function following base R code, identify arguments quoted evaluated.A1. identify arguments quoted evaluated, can use trick mentioned book:’re ever unsure whether argument quoted evaluated, try executing code outside function. doesn’t work something different, argument quoted.library(MASS)package argument library() quoted:subset(mtcars, cyl == 4)argument x evaluated, argument subset quoted.(mtcars2, sum(vs))argument data evaluated, expr argument quoted.sum(mtcars2$)argument ... evaluated.rm(mtcars2)trick using far won’t work since trying print mtcars2 always fail rm() made pass .can instead look docs ...:… objects removed, names (unquoted) character strings (quoted).Thus, argument evaluated, rather quoted.Q2. function following tidyverse code, identify arguments quoted evaluated.A2. seen answer Q1., library() quotes first argument:following code:%>% (lazily) evaluates argumentgroup_by() summarise() quote argumentsIn following code:ggplot() evaluates data argumentaes() quotes arguments","code":"\nlibrary(MASS)\n\nmtcars2 <- subset(mtcars, cyl == 4)\n\nwith(mtcars2, sum(vs))\nsum(mtcars2$am)\n\nrm(mtcars2)\nlibrary(MASS)\n\nMASS\n#> Error in eval(expr, envir, enclos): object 'MASS' not found\nmtcars2 <- subset(mtcars, cyl == 4)\n\ninvisible(mtcars)\n\ncyl == 4\n#> Error in eval(expr, envir, enclos): object 'cyl' not found\nwith(mtcars2, sum(vs))\n#> [1] 10\n\ninvisible(mtcars2)\n\nsum(vs)\n#> Error in eval(expr, envir, enclos): object 'vs' not found\nsum(mtcars2$am)\n#> [1] 8\n\nmtcars2$am\n#>  [1] 1 0 0 1 1 1 0 1 1 1 1\nrm(mtcars2)\nlibrary(dplyr)\nlibrary(ggplot2)\n\nby_cyl <- mtcars %>%\n  group_by(cyl) %>%\n  summarise(mean = mean(mpg))\n\nggplot(by_cyl, aes(cyl, mean)) +\n  geom_point()\nlibrary(dplyr)\nlibrary(ggplot2)\nby_cyl <- mtcars %>%\n  group_by(cyl) %>%\n  summarise(mean = mean(mpg))\nggplot(by_cyl, aes(cyl, mean)) +\n  geom_point()"},{"path":"quasiquotation.html","id":"quoting-exercises-19.3.6","chapter":"19 Quasiquotation","heading":"19.2 Quoting (Exercises 19.3.6)","text":"Q1. expr() implemented? Look source code.A1. Looking source code, can see expr() simple wrapper around enexpr(), captures returns user-entered expressions:example:turn, enexpr() calls native code:Q2. Compare contrast following two functions. Can predict output running ?A2. exprs() captures returns expressions specified developer instead values:hand, enexprs() captures user-entered expressions returns values:Q3. happens try use enexpr() expression (.e. enexpr(x + y)? happens enexpr() passed missing argument?A3. try use enexpr() expression, fails works symbol.enexpr() passed missing argument, returns missing argument:Q4. exprs() exprs(= ) different? Think input output.A4. key difference exprs() exprs(= ) former return unnamed list, latter return named list. former interpreted unnamed argument, latter named argument.cases, treated symbol:, argument missing latter case, since name corresponding value provided:Q5. differences exprs() alist()? Read documentation named arguments exprs() find .A5. additional differences exprs() alist().Names: inputs named, exprs() provides way name automatically using .named argument.Ignoring empty arguments: .ignore_empty argument exprs() gives much finer control empty arguments, alist() doesn’t provide way ignore arguments.Names injection: Using .unquote_names argument exprs(), can inject name argument.Q6. documentation substitute() says:Substitution takes place examining component parse tree\nfollows:bound symbol env, unchanged.promise object (.e., formal argument function)\nexpression slot promise replaces symbol.ordinary variable, value substituted, unless\nenv .GlobalEnv case symbol left unchanged.Create examples illustrate cases.A6. See examples illustrate -mentioned cases.bound symbol env, unchanged.Symbol x bound env, remains unchanged.promise object (.e., formal argument function)\nexpression slot promise replaces symbol.ordinary variable, value substituted, unless\nenv .GlobalEnv case symbol left unchanged.","code":"\nrlang::expr\n#> function (expr) \n#> {\n#>     enexpr(expr)\n#> }\n#> <bytecode: 0x5584f9fd1728>\n#> <environment: namespace:rlang>\nx <- expr(x <- 1)\nx\n#> x <- 1\nrlang::enexpr\n#> function (arg) \n#> {\n#>     .Call(ffi_enexpr, substitute(arg), parent.frame())\n#> }\n#> <bytecode: 0x5584f6b2cae0>\n#> <environment: namespace:rlang>\nf1 <- function(x, y) {\n  exprs(x = x, y = y)\n}\nf2 <- function(x, y) {\n  enexprs(x = x, y = y)\n}\nf1(a + b, c + d)\nf2(a + b, c + d)\nf1 <- function(x, y) {\n  exprs(x = x, y = y)\n}\n\nf1(a + b, c + d)\n#> $x\n#> x\n#> \n#> $y\n#> y\nf2 <- function(x, y) {\n  enexprs(x = x, y = y)\n}\n\nf2(a + b, c + d)\n#> $x\n#> a + b\n#> \n#> $y\n#> c + d\nenexpr(x + y)\n#> Error in `enexpr()`:\n#> ! `arg` must be a symbol\narg <- missing_arg()\n\nenexpr(arg)\n\nis_missing(enexpr(arg))\n#> [1] TRUE\nexprs(a)\n#> [[1]]\n#> a\n\nexprs(a = )\n#> $a\nmap_lgl(exprs(a), is_symbol)\n#>      \n#> TRUE\n\nmap_lgl(exprs(a = ), is_symbol)\n#>    a \n#> TRUE\nmap_lgl(exprs(a), is_missing)\n#>       \n#> FALSE\n\nmap_lgl(exprs(a = ), is_missing)\n#>    a \n#> TRUE\nalist(\"x\" = 1, TRUE, \"z\" = expr(x + y))\n#> $x\n#> [1] 1\n#> \n#> [[2]]\n#> [1] TRUE\n#> \n#> $z\n#> expr(x + y)\n\nexprs(\"x\" = 1, TRUE, \"z\" = expr(x + y), .named = TRUE)\n#> $x\n#> [1] 1\n#> \n#> $`TRUE`\n#> [1] TRUE\n#> \n#> $z\n#> expr(x + y)\nalist(\"x\" = 1, , TRUE, )\n#> $x\n#> [1] 1\n#> \n#> [[2]]\n#> \n#> \n#> [[3]]\n#> [1] TRUE\n#> \n#> [[4]]\n\nexprs(\"x\" = 1, , TRUE, , .ignore_empty = \"trailing\")\n#> $x\n#> [1] 1\n#> \n#> [[2]]\n#> \n#> \n#> [[3]]\n#> [1] TRUE\n\nexprs(\"x\" = 1, , TRUE, , .ignore_empty = \"none\")\n#> $x\n#> [1] 1\n#> \n#> [[2]]\n#> \n#> \n#> [[3]]\n#> [1] TRUE\n#> \n#> [[4]]\n\nexprs(\"x\" = 1, , TRUE, , .ignore_empty = \"all\")\n#> $x\n#> [1] 1\n#> \n#> [[2]]\n#> [1] TRUE\nalist(foo := bar)\n#> [[1]]\n#> `:=`(foo, bar)\n\nexprs(foo := bar, .unquote_names = FALSE)\n#> [[1]]\n#> `:=`(foo, bar)\n\nexprs(foo := bar, .unquote_names = TRUE)\n#> $foo\n#> bar\nsubstitute(x + y, env = list(y = 2))\n#> x + 2\nmsg <- \"old\"\ndelayedAssign(\"myVar\", msg) # creates a promise\nsubstitute(myVar)\n#> myVar\nmsg <- \"new!\"\nmyVar\n#> [1] \"new!\"\nsubstitute(x + y, env = env(x = 2, y = 1))\n#> 2 + 1\n\nx <- 2\ny <- 1\nsubstitute(x + y, env = .GlobalEnv)\n#> x + y"},{"path":"quasiquotation.html","id":"unquoting-exercises-19.4.8","chapter":"19 Quasiquotation","heading":"19.3 Unquoting (Exercises 19.4.8)","text":"Q1. Given following components:Use quasiquotation construct following calls:A1. Using quasiquotation construct specified calls:Q2. following two calls print , actually different:’s difference? one natural?A2. can see difference two expression convert lists:can seen, second element call object, b integer vector:can also noticed ASTs expressions:first call natural, since second one inlines vector directly call, something rarely done.","code":"\nxy <- expr(x + y)\nxz <- expr(x + z)\nyz <- expr(y + z)\nabc <- exprs(a, b, c)\n(x + y) / (y + z)\n-(x + z)^(y + z)\n(x + y) + (y + z) - (x + y)\natan2(x + y, y + z)\nsum(x + y, x + y, y + z)\nsum(a, b, c)\nmean(c(a, b, c), na.rm = TRUE)\nfoo(a = x + y, b = y + z)\nxy <- expr(x + y)\nxz <- expr(x + z)\nyz <- expr(y + z)\nabc <- exprs(a, b, c)\n\nexpr((!!xy) / (!!yz))\n#> (x + y)/(y + z)\n\nexpr(-(!!xz)^(!!yz))\n#> -(x + z)^(y + z)\n\nexpr(((!!xy)) + (!!yz) - (!!xy))\n#> (x + y) + (y + z) - (x + y)\n\ncall2(\"atan2\", expr(!!xy), expr(!!yz))\n#> atan2(x + y, y + z)\n\ncall2(\"sum\", expr(!!xy), expr(!!xy), expr(!!yz))\n#> sum(x + y, x + y, y + z)\n\ncall2(\"sum\", !!!abc)\n#> sum(a, b, c)\n\nexpr(mean(c(!!!abc), na.rm = TRUE))\n#> mean(c(a, b, c), na.rm = TRUE)\n\ncall2(\"foo\", a = expr(!!xy), b = expr(!!yz))\n#> foo(a = x + y, b = y + z)\n(a <- expr(mean(1:10)))\n#> mean(1:10)\n(b <- expr(mean(!!(1:10))))\n#> mean(1:10)\nidentical(a, b)\n#> [1] FALSE\nas.list(expr(mean(1:10)))\n#> [[1]]\n#> mean\n#> \n#> [[2]]\n#> 1:10\n\nas.list(expr(mean(!!(1:10))))\n#> [[1]]\n#> mean\n#> \n#> [[2]]\n#>  [1]  1  2  3  4  5  6  7  8  9 10\nwaldo::compare(a, b)\n#> `old[[2]]` is a call\n#> `new[[2]]` is an integer vector (1, 2, 3, 4, 5, ...)\nast(expr(mean(1:10)))\n#> █─expr \n#> └─█─mean \n#>   └─█─`:` \n#>     ├─1 \n#>     └─10\n\nast(expr(mean(!!(1:10))))\n#> █─expr \n#> └─█─mean \n#>   └─<inline integer>"},{"path":"quasiquotation.html","id":"dot-dot-dot-exercises-19.6.5","chapter":"19 Quasiquotation","heading":"19.4 ... (dot-dot-dot) (Exercises 19.6.5)","text":"Q1. One way implement exec() shown . Describe works. key ideas?A1. keys ideas underlie implementation exec() function following:constructs call using function f argument ..., evaluates call environment .env.constructs call using function f argument ..., evaluates call environment .env.uses dynamic dots via list2(), means can splice arguments using !!!, can inject names using :=, trailing commas problem.uses dynamic dots via list2(), means can splice arguments using !!!, can inject names using :=, trailing commas problem.example:Q2. Carefully read source code interaction(), expand.grid(), par(). Compare contrast techniques use switching dots list behaviour.A2. Source code reveals following comparison table:functions capture dots list.Using dots, functions check:list entered argument checking number argumentsif count 1, checking argument listQ3. Explain problem definition set_attr()A3. set_attr() function signature parameter called x, additionally uses dynamic dots pass multiple arguments specify additional attributes x., shown example, creates problem attribute named x. Naming arguments won’t help either:can avoid issues renaming parameter:","code":"\nexec <- function(f, ..., .env = caller_env()) {\n  args <- list2(...)\n  do.call(f, args, envir = .env)\n}\nvec <- c(1:5, NA)\nargs_list <- list(trim = 0, na.rm = TRUE)\n\nexec(mean, vec, !!!args_list, , .env = caller_env())\n#> [1] 3\n\nrm(\"exec\")\nset_attr <- function(x, ...) {\n  attr <- rlang::list2(...)\n  attributes(x) <- attr\n  x\n}\nset_attr(1:10, x = 10)\n#> Error in attributes(x) <- attr: attributes must be named\nset_attr <- function(x, ...) {\n  attr <- rlang::list2(...)\n  attributes(x) <- attr\n  x\n}\nset_attr(x = 1:10, x = 10)\n#> Error in set_attr(x = 1:10, x = 10): formal argument \"x\" matched by multiple actual arguments\nset_attr <- function(.x, ...) {\n  attr <- rlang::list2(...)\n  attributes(.x) <- attr\n  .x\n}\n\nset_attr(.x = 1:10, x = 10)\n#>  [1]  1  2  3  4  5  6  7  8  9 10\n#> attr(,\"x\")\n#> [1] 10"},{"path":"quasiquotation.html","id":"case-studies-exercises-19.7.5","chapter":"19 Quasiquotation","heading":"19.5 Case studies (Exercises 19.7.5)","text":"Q1. linear-model example, replace expr() reduce(summands, ~ expr(!!.x + !!.y)) call2(): reduce(summands, call2, \"+\"). Compare contrast two approaches. think easier read?A1. can rewrite linear() function chapter using call2() follows:personally find version call2() much readable since !! syntax bit esoteric.Q2. Re-implement Box-Cox transform defined using unquoting new_function():A2. Re-implementation Box-Cox transform using unquoting new_function():Let’s try see produces output :Q3. Re-implement simple compose() defined using quasiquotation new_function():A3. Following re-implementation compose() using quasiquotation new_function():Checking new version behaves way original version:","code":"\nlinear <- function(var, val) {\n  var <- ensym(var)\n  coef_name <- map(seq_along(val[-1]), ~ expr((!!var)[[!!.x]]))\n\n  summands <- map2(val[-1], coef_name, ~ expr((!!.x * !!.y)))\n  summands <- c(val[[1]], summands)\n\n  reduce(summands, ~ call2(\"+\", .x, .y))\n}\n\nlinear(x, c(10, 5, -4))\n#> 10 + (5 * x[[1L]]) + (-4 * x[[2L]])\nbc <- function(lambda) {\n  if (lambda == 0) {\n    function(x) log(x)\n  } else {\n    function(x) (x^lambda - 1) / lambda\n  }\n}\nbc_new <- function(lambda) {\n  lambda <- enexpr(lambda)\n\n  if (!!lambda == 0) {\n    new_function(\n      exprs(x = ),\n      expr(log(x))\n    )\n  } else {\n    new_function(\n      exprs(x = ),\n      expr((x^(!!lambda) - 1) / (!!lambda))\n    )\n  }\n}\nbc(0)(1)\n#> [1] 0\nbc_new(0)(1)\n#> [1] 0\n\nbc(2)(2)\n#> [1] 1.5\nbc_new(2)(2)\n#> [1] 1.5\ncompose <- function(f, g) {\n  function(...) f(g(...))\n}\ncompose_new <- function(f, g) {\n  f <- enexpr(f) # or ensym(f)\n  g <- enexpr(g) # or ensym(g)\n\n  new_function(\n    exprs(... = ),\n    expr((!!f)((!!g)(...)))\n  )\n}\nnot_null <- compose(`!`, is.null)\nnot_null(4)\n#> [1] TRUE\n\nnot_null2 <- compose_new(`!`, is.null)\nnot_null2(4)\n#> [1] TRUE"},{"path":"quasiquotation.html","id":"session-information-14","chapter":"19 Quasiquotation","heading":"19.6 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  colorspace    2.1-0   2023-01-23 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  crayon        1.5.2   2022-09-29 [1] RSPM\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  diffobj       0.3.5   2021-10-05 [1] RSPM\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr       * 1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  farver        2.1.1   2022-07-06 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  ggplot2     * 3.4.4   2023-10-12 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  grid          4.3.2   2023-11-01 [3] local\n#>  gtable        0.3.4   2023-08-21 [1] RSPM\n#>  highr         0.10    2022-12-22 [1] RSPM\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  labeling      0.4.3   2023-08-29 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  lobstr      * 1.1.2   2022-06-22 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  MASS        * 7.3-60  2023-05-04 [3] CRAN (R 4.3.2)\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  munsell       0.5.0   2018-06-12 [1] RSPM\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  purrr       * 1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rematch2      2.1.2   2020-05-01 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  scales        1.3.0   2023-11-28 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  waldo         0.5.2   2023-11-02 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"evaluation.html","id":"evaluation","chapter":"20 Evaluation","heading":"20 Evaluation","text":"Attaching needed libraries:","code":"\nlibrary(rlang)"},{"path":"evaluation.html","id":"evaluation-basics-exercises-20.2.4","chapter":"20 Evaluation","heading":"20.1 Evaluation basics (Exercises 20.2.4)","text":"Q1. Carefully read documentation source(). environment use default? supply local = TRUE? provide custom environment?A1. parameter local source() decides environment parsed expressions evaluated.default local = FALSE, corresponds user’s workspace (global environment, .e.).local = TRUE, environment source() called used.specify custom environment, sys.source() function can used, provides envir parameter.Q2. Predict results following lines code:A2. Correctly predicted 😉Q3. Fill function bodies re-implement get() using sym() eval(), assign() using sym(), expr(), eval(). Don’t worry multiple ways choosing environment get() assign() support; assume user supplies explicitly.A3. required re-implementations:get()assign()Q4. Modify source2() returns result every expression, just last one. Can eliminate loop?A4. can use purrr::map() iterate every expression return result every expression:Q5. can make base::local() slightly easier understand spreading multiple lines:Explain local() works words. (Hint: might want print(call) help understand substitute() , read documentation remind environment new.env() inherit .)A5. order figure function works, let’s add suggested print(call):docs substitute() mention:Substituting quoting often cause confusion argument expression(…). result call expression constructor function needs evaluated eval give actual expression object.Thus, get actual expression object, quoted expression needs evaluated using eval():Finally, generated call evaluated caller environment. final function call looks like following:Note bindings x y found inner environment, bindings functions eval(), quote(), etc. found outer environment.","code":"\nwithr::with_tempdir(\n  code = {\n    f <- tempfile()\n    writeLines(\"rlang::env_print()\", f)\n    foo <- function() source(f, local = FALSE)\n    foo()\n  }\n)\n#> <environment: global>\n#> Parent: <environment: package:rlang>\n#> Bindings:\n#> • .Random.seed: <int>\n#> • foo: <fn>\n#> • f: <chr>\nwithr::with_tempdir(\n  code = {\n    f <- tempfile()\n    writeLines(\"rlang::env_print()\", f)\n    foo <- function() source(f, local = TRUE)\n    foo()\n  }\n)\n#> <environment: 0x564fa1bed6a8>\n#> Parent: <environment: global>\neval(expr(eval(expr(eval(expr(2 + 2))))))\neval(eval(expr(eval(expr(eval(expr(2 + 2)))))))\nexpr(eval(expr(eval(expr(eval(expr(2 + 2)))))))\neval(expr(eval(expr(eval(expr(2 + 2))))))\n#> [1] 4\n\neval(eval(expr(eval(expr(eval(expr(2 + 2)))))))\n#> [1] 4\n\nexpr(eval(expr(eval(expr(eval(expr(2 + 2)))))))\n#> eval(expr(eval(expr(eval(expr(2 + 2))))))\n# name is a string\nget2 <- function(name, env) {}\nassign2 <- function(name, value, env) {}\nget2 <- function(name, env = caller_env()) {\n  name <- sym(name)\n  eval(name, env)\n}\n\nx <- 2\n\nget2(\"x\")\n#> [1] 2\nget(\"x\")\n#> [1] 2\n\ny <- 1:4\nassign(\"y[1]\", 2)\n\nget2(\"y[1]\")\n#> [1] 2\nget(\"y[1]\")\n#> [1] 2\nassign2 <- function(name, value, env = caller_env()) {\n  name <- sym(name)\n  eval(expr(!!name <- !!value), env)\n}\n\nassign(\"y1\", 4)\ny1\n#> [1] 4\n\nassign2(\"y2\", 4)\ny2\n#> [1] 4\nsource2 <- function(path, env = caller_env()) {\n  file <- paste(readLines(path, warn = FALSE), collapse = \"\\n\")\n  exprs <- parse_exprs(file)\n  purrr::map(exprs, ~ eval(.x, env))\n}\n\nwithr::with_tempdir(\n  code = {\n    f <- tempfile(fileext = \".R\")\n    writeLines(\"1 + 1; 2 + 4\", f)\n    source2(f)\n  }\n)\n#> [[1]]\n#> [1] 2\n#> \n#> [[2]]\n#> [1] 6\nlocal3 <- function(expr, envir = new.env()) {\n  call <- substitute(eval(quote(expr), envir))\n  eval(call, envir = parent.frame())\n}\nlocal3 <- function(expr, envir = new.env()) {\n  call <- substitute(eval(quote(expr), envir))\n  print(call)\n\n  eval(call, envir = parent.frame())\n}\n\nlocal3({\n  x <- 10\n  y <- 200\n  x + y\n})\n#> eval(quote({\n#>     x <- 10\n#>     y <- 200\n#>     x + y\n#> }), new.env())\n#> [1] 210\nis_expression(eval(quote({\n  x <- 10\n  y <- 200\n  x + y\n}), new.env()))\n#> [1] TRUE\n# outer environment\neval(\n  # inner environment\n  eval(quote({\n    x <- 10\n    y <- 200\n    x + y\n  }), new.env()),\n  envir = parent.frame()\n)"},{"path":"evaluation.html","id":"quosures-exercises-20.3.6","chapter":"20 Evaluation","heading":"20.2 Quosures (Exercises 20.3.6)","text":"Q1. Predict following quosures return evaluated.A1. Correctly predicted 😉Q2. Write enenv() function captures environment associated argument. (Hint: require two function calls.)A2. can make use get_env() helper get environment associated argument:","code":"\nq1 <- new_quosure(expr(x), env(x = 1))\nq1\n#> <quosure>\n#> expr: ^x\n#> env:  0x564fa1a3e9a0\nq2 <- new_quosure(expr(x + !!q1), env(x = 10))\nq2\n#> <quosure>\n#> expr: ^x + (^x)\n#> env:  0x564fa20915d8\nq3 <- new_quosure(expr(x + !!q2), env(x = 100))\nq3\n#> <quosure>\n#> expr: ^x + (^x + (^x))\n#> env:  0x564fa23500b0\nq1 <- new_quosure(expr(x), env(x = 1))\neval_tidy(q1)\n#> [1] 1\n\nq2 <- new_quosure(expr(x + !!q1), env(x = 10))\neval_tidy(q2)\n#> [1] 11\n\nq3 <- new_quosure(expr(x + !!q2), env(x = 100))\neval_tidy(q3)\n#> [1] 111\nenenv <- function(x) {\n  x <- enquo(x)\n  get_env(x)\n}\n\nenenv(x)\n#> <environment: R_GlobalEnv>\n\nfoo <- function(x) enenv(x)\nfoo()\n#> <environment: 0x564fa32bc708>"},{"path":"evaluation.html","id":"data-masks-exercises-20.4.6","chapter":"20 Evaluation","heading":"20.3 Data masks (Exercises 20.4.6)","text":"Q1. use loop transform2() instead map()? Consider transform2(df, x = x * 2, x = x * 2).A1. see map() appropriate function, let’s create version function map() see happens.use () loop, iteration, updating x column current expression evaluation. , repeatedly modifying column works.use map() instead, trying evaluate expressions time; .e., column attempted modify using multiple expressions.Q2. ’s alternative implementation subset2():Compare contrast subset3() subset2(). advantages disadvantages?A2. Let’s first juxtapose functions outputs can compare better.Disadvantages subset3() subset2()filtering conditions specified rows don’t evaluate logical, function doesn’t fail informatively. Indeed, silently returns incorrect result.Advantages subset3() subset2()might argue function shorter advantage, much subjective preference.Q3. following function implements basics dplyr::arrange(). Annotate line comment explaining . Can explain !!.na.last strictly correct, omitting !! unlikely cause problems?A3. Annotated version function:see doesn’t matter whether whether unquote .na.last argument , let’s look smaller example:can seen:without unquoting, .na.last found function environmentwith unquoting, .na.last included order call object ","code":"\ntransform2 <- function(.data, ...) {\n  dots <- enquos(...)\n\n  for (i in seq_along(dots)) {\n    name <- names(dots)[[i]]\n    dot <- dots[[i]]\n\n    .data[[name]] <- eval_tidy(dot, .data)\n  }\n\n  .data\n}\n\ntransform3 <- function(.data, ...) {\n  dots <- enquos(...)\n\n  purrr::map(dots, function(x, .data = .data) {\n    name <- names(x)\n    dot <- x\n\n    .data[[name]] <- eval_tidy(dot, .data)\n\n    .data\n  })\n}\ndf <- data.frame(x = 1:3)\ntransform2(df, x = x * 2, x = x * 2)\n#>    x\n#> 1  4\n#> 2  8\n#> 3 12\ndf <- data.frame(x = 1:3)\ntransform3(df, x = x * 2, x = x * 2)\n#> Error in `purrr::map()`:\n#> ℹ In index: 1.\n#> ℹ With name: x.\n#> Caused by error:\n#> ! promise already under evaluation: recursive default argument reference or earlier problems?\nsubset3 <- function(data, rows) {\n  rows <- enquo(rows)\n  eval_tidy(expr(data[!!rows, , drop = FALSE]), data = data)\n}\ndf <- data.frame(x = 1:3)\nsubset3(df, x == 1)\nsubset2 <- function(data, rows) {\n  rows <- enquo(rows)\n  rows_val <- eval_tidy(rows, data)\n  stopifnot(is.logical(rows_val))\n\n  data[rows_val, , drop = FALSE]\n}\n\ndf <- data.frame(x = 1:3)\nsubset2(df, x == 1)\n#>   x\n#> 1 1\nsubset3 <- function(data, rows) {\n  rows <- enquo(rows)\n  eval_tidy(expr(data[!!rows, , drop = FALSE]), data = data)\n}\n\nsubset3(df, x == 1)\n#>   x\n#> 1 1\nrm(\"x\")\nexists(\"x\")\n#> [1] FALSE\n\nsubset2(df, x + 1)\n#> Error in subset2(df, x + 1): is.logical(rows_val) is not TRUE\n\nsubset3(df, x + 1)\n#>     x\n#> 2   2\n#> 3   3\n#> NA NA\narrange2 <- function(.df, ..., .na.last = TRUE) {\n  args <- enquos(...)\n  order_call <- expr(order(!!!args, na.last = !!.na.last))\n  ord <- eval_tidy(order_call, .df)\n  stopifnot(length(ord) == nrow(.df))\n  .df[ord, , drop = FALSE]\n}\narrange2 <- function(.df, ..., .na.last = TRUE) {\n  # capture user-supplied expressions (and corresponding environments) as quosures\n  args <- enquos(...)\n\n  # create a call object by splicing a list of quosures\n  order_call <- expr(order(!!!args, na.last = !!.na.last))\n\n  # and evaluate the constructed call in the data frame\n  ord <- eval_tidy(order_call, .df)\n\n  # sanity check\n  stopifnot(length(ord) == nrow(.df))\n\n  .df[ord, , drop = FALSE]\n}\nx <- TRUE\neval(expr(c(x = !!x)))\n#>    x \n#> TRUE\neval(expr(c(x = x)))\n#>    x \n#> TRUE"},{"path":"evaluation.html","id":"using-tidy-evaluation-exercises-20.5.4","chapter":"20 Evaluation","heading":"20.4 Using tidy evaluation (Exercises 20.5.4)","text":"Q1. ’ve included alternative implementation threshold_var() . makes different approach used ? makes harder?A1. First, let’s compare two definitions function make sure produce output:key difference subsetting operator used:old version uses non-quoting [[ operator. Thus, var argument first needs converted string.new version uses quoting $ operator. Thus, var argument first quoted unquoted (using !!).","code":"\nthreshold_var <- function(df, var, val) {\n  var <- ensym(var)\n  subset2(df, `$`(.data, !!var) >= !!val)\n}\nthreshold_var_old <- function(df, var, val) {\n  var <- as_string(ensym(var))\n  subset2(df, .data[[var]] >= !!val)\n}\n\nthreshold_var_new <- threshold_var\n\ndf <- data.frame(x = 1:10)\n\nidentical(\n  threshold_var(df, x, 8),\n  threshold_var(df, x, 8)\n)\n#> [1] TRUE"},{"path":"evaluation.html","id":"base-evaluation-exercises-20.6.3","chapter":"20 Evaluation","heading":"20.5 Base evaluation (Exercises 20.6.3)","text":"Q1. function fail?A1. doesn’t work lm_call call evaluated caller_env(), finds binding base::data() function, data execution environment.make work, need unquote data expression:Q2. model building, typically response data relatively constant rapidly experiment different predictors. Write small wrapper allows reduce duplication code .A2. small wrapper allows enter predictors:function flexible enough also allow changing data dependent variable:Q3. Another way write resample_lm() include resample expression (data[sample(nrow(data), replace = TRUE), , drop = FALSE]) data argument. Implement approach. advantages? disadvantages?A3. variant resample_lm(), providing resampled data argument.makes use R’s lazy evaluation function arguments. , resample_data argument evaluated needed function.","code":"\nlm3a <- function(formula, data) {\n  formula <- enexpr(formula)\n  lm_call <- expr(lm(!!formula, data = data))\n  eval(lm_call, caller_env())\n}\n\nlm3a(mpg ~ disp, mtcars)$call\n#> Error in as.data.frame.default(data, optional = TRUE):\n#> cannot coerce class ‘\"function\"’ to a data.frame\nlm3a <- function(formula, data) {\n  formula <- enexpr(formula)\n  lm_call <- expr(lm(!!formula, data = !!data))\n  eval(lm_call, caller_env())\n}\n\nis_call(lm3a(mpg ~ disp, mtcars)$call)\n#> [1] TRUE\nlm(mpg ~ disp, data = mtcars)\nlm(mpg ~ I(1 / disp), data = mtcars)\nlm(mpg ~ disp * cyl, data = mtcars)\nlm_custom <- function(data = mtcars, x, y = mpg) {\n  x <- enexpr(x)\n  y <- enexpr(y)\n  data <- enexpr(data)\n\n  lm_call <- expr(lm(formula = !!y ~ !!x, data = !!data))\n\n  eval(lm_call, caller_env())\n}\n\nidentical(\n  lm_custom(x = disp),\n  lm(mpg ~ disp, data = mtcars)\n)\n#> [1] TRUE\n\nidentical(\n  lm_custom(x = I(1 / disp)),\n  lm(mpg ~ I(1 / disp), data = mtcars)\n)\n#> [1] TRUE\n\nidentical(\n  lm_custom(x = disp * cyl),\n  lm(mpg ~ disp * cyl, data = mtcars)\n)\n#> [1] TRUE\nlm_custom(data = iris, x = Sepal.Length, y = Petal.Width)\n#> \n#> Call:\n#> lm(formula = Petal.Width ~ Sepal.Length, data = iris)\n#> \n#> Coefficients:\n#>  (Intercept)  Sepal.Length  \n#>      -3.2002        0.7529\nresample_lm3 <- function(formula,\n                         data,\n                         resample_data = data[sample(nrow(data), replace = TRUE), , drop = FALSE],\n                         env = current_env()) {\n  formula <- enexpr(formula)\n  lm_call <- expr(lm(!!formula, data = resample_data))\n  expr_print(lm_call)\n  eval(lm_call, env)\n}\n\ndf <- data.frame(x = 1:10, y = 5 + 3 * (1:10) + round(rnorm(10), 2))\nresample_lm3(y ~ x, data = df)\n#> lm(y ~ x, data = resample_data)\n#> \n#> Call:\n#> lm(formula = y ~ x, data = resample_data)\n#> \n#> Coefficients:\n#> (Intercept)            x  \n#>       2.654        3.420"},{"path":"evaluation.html","id":"session-information-15","chapter":"20 Evaluation","heading":"20.6 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  purrr         1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"translation.html","id":"translation","chapter":"21 Translation","heading":"21 Translation","text":"Needed libraries:","code":"\nlibrary(rlang)\nlibrary(purrr)"},{"path":"translation.html","id":"html-exercises-21.2.6","chapter":"21 Translation","heading":"21.1 HTML (Exercises 21.2.6)","text":"Q1. escaping rules <script> tags different contain JavaScript, HTML. Instead escaping angle brackets ampersands, need escape <\/script> tag isn’t closed early. example, script(\"'<\/script>'\"), shouldn’t generate :ButAdapt escape() follow rules new argument script set TRUE.A1. Let’s first start boilerplate code included book:also need tweak boilerplate pass additional parameter escape():Q2. use ... functions big downsides. ’s input validation little information documentation autocomplete used function. Create new function , given named list tags attribute names (like ), creates tag functions named arguments.tags get class id attributes.Q3. Reason following code calls with_html() referencing objects environment. work fail? ? Run code verify predictions.A3. work , first need copy-paste relevant code book:Note with_html() uses eval_tidy(), therefore code argument evaluated first html_tags named list, acts data mask, object found data mask, searches caller environment.reason, first example code work:following code, however, going work already address element data mask, p() take function address() input, escape() doesn’t know deal objects function type:Q4. Currently HTML doesn’t look terribly pretty, ’s hard see structure. adapt tag() indenting formatting? (may need research block inline tags.)A4. Let’s first look currently looks like:can improve follow Google HTML/CSS Style Guide., need create new function indent code conditionally:can update body() function use new helper:new formatting looks much better:","code":"  <script>'<\/script>'<\/script>  <script>'<\\/script>'<\/script>\nescape <- function(x, ...) UseMethod(\"escape\")\n\nescape.character <- function(x, script = FALSE) {\n  if (script) {\n    x <- gsub(\"<\/script>\", \"<\\\\/script>\", x, fixed = TRUE)\n  } else {\n    x <- gsub(\"&\", \"&amp;\", x)\n    x <- gsub(\"<\", \"&lt;\", x)\n    x <- gsub(\">\", \"&gt;\", x)\n  }\n\n  html(x)\n}\n\nescape.advr_html <- function(x, ...) x\nhtml <- function(x) structure(x, class = \"advr_html\")\n\nprint.advr_html <- function(x, ...) {\n  out <- paste0(\"<HTML> \", x)\n  cat(paste(strwrap(out), collapse = \"\\n\"), \"\\n\", sep = \"\")\n}\n\ndots_partition <- function(...) {\n  dots <- list2(...)\n\n  if (is.null(names(dots))) {\n    is_named <- rep(FALSE, length(dots))\n  } else {\n    is_named <- names(dots) != \"\"\n  }\n\n  list(\n    named = dots[is_named],\n    unnamed = dots[!is_named]\n  )\n}\n\ntag <- function(tag, script = FALSE) {\n  force(script)\n  new_function(\n    exprs(... = ),\n    expr({\n      dots <- dots_partition(...)\n      attribs <- html_attributes(dots$named)\n      children <- map_chr(.x = dots$unnamed, .f = ~ escape(.x, !!script))\n\n      html(paste0(\n        !!paste0(\"<\", tag), attribs, \">\",\n        paste(children, collapse = \"\"),\n        !!paste0(\"<\/\", tag, \">\")\n      ))\n    }),\n    caller_env()\n  )\n}\n\nvoid_tag <- function(tag) {\n  new_function(\n    exprs(... = ),\n    expr({\n      dots <- dots_partition(...)\n      if (length(dots$unnamed) > 0) {\n        abort(!!paste0(\"<\", tag, \"> must not have unnamed arguments\"))\n      }\n      attribs <- html_attributes(dots$named)\n\n      html(paste0(!!paste0(\"<\", tag), attribs, \" />\"))\n    }),\n    caller_env()\n  )\n}\n\np <- tag(\"p\")\nscript <- tag(\"script\", script = TRUE)\nscript(\"'<\/script>'\")\n#> <HTML> <script>'<\\/script>'<\/script>\nlist(\n  a = c(\"href\"),\n  img = c(\"src\", \"width\", \"height\")\n)\ngreeting <- \"Hello!\"\nwith_html(p(greeting))\np <- function() \"p\"\naddress <- \"123 anywhere street\"\nwith_html(p(address))\ntags <- c(\n  \"a\", \"abbr\", \"address\", \"article\", \"aside\", \"audio\",\n  \"b\", \"bdi\", \"bdo\", \"blockquote\", \"body\", \"button\", \"canvas\",\n  \"caption\", \"cite\", \"code\", \"colgroup\", \"data\", \"datalist\",\n  \"dd\", \"del\", \"details\", \"dfn\", \"div\", \"dl\", \"dt\", \"em\",\n  \"eventsource\", \"fieldset\", \"figcaption\", \"figure\", \"footer\",\n  \"form\", \"h1\", \"h2\", \"h3\", \"h4\", \"h5\", \"h6\", \"head\", \"header\",\n  \"hgroup\", \"html\", \"i\", \"iframe\", \"ins\", \"kbd\", \"label\",\n  \"legend\", \"li\", \"mark\", \"map\", \"menu\", \"meter\", \"nav\",\n  \"noscript\", \"object\", \"ol\", \"optgroup\", \"option\", \"output\",\n  \"p\", \"pre\", \"progress\", \"q\", \"ruby\", \"rp\", \"rt\", \"s\", \"samp\",\n  \"script\", \"section\", \"select\", \"small\", \"span\", \"strong\",\n  \"style\", \"sub\", \"summary\", \"sup\", \"table\", \"tbody\", \"td\",\n  \"textarea\", \"tfoot\", \"th\", \"thead\", \"time\", \"title\", \"tr\",\n  \"u\", \"ul\", \"var\", \"video\"\n)\n\nvoid_tags <- c(\n  \"area\", \"base\", \"br\", \"col\", \"command\", \"embed\",\n  \"hr\", \"img\", \"input\", \"keygen\", \"link\", \"meta\", \"param\",\n  \"source\", \"track\", \"wbr\"\n)\n\nhtml_tags <- c(\n  tags %>% set_names() %>% map(tag),\n  void_tags %>% set_names() %>% map(void_tag)\n)\n\nwith_html <- function(code) {\n  code <- enquo(code)\n  eval_tidy(code, html_tags)\n}\ngreeting <- \"Hello!\"\nwith_html(p(greeting))\n#> <HTML> <p>Hello!<\/p>\n\"address\" %in% names(html_tags)\n#> [1] TRUE\n\np <- function() \"p\"\naddress <- \"123 anywhere street\"\nwith_html(p(address))\n#> Error in `map_chr()`:\n#> ℹ In index: 1.\n#> Caused by error in `UseMethod()`:\n#> ! no applicable method for 'escape' applied to an object of class \"function\"\nwith_html(\n  body(\n    h1(\"A heading\", id = \"first\"),\n    p(\"Some text &\", b(\"some bold text.\")),\n    img(src = \"myimg.png\", width = 100, height = 100)\n  )\n)\n#> <HTML> <body><h1 id='first'>A heading<\/h1><p>Some\n#> text &amp;<b>some bold text.<\/b><\/p><img\n#> src='myimg.png' width='100' height='100' /><\/body>\nprint.advr_html <- function(x, ...) {\n  cat(paste(\"<HTML>\", x, sep = \"\\n\"))\n}\n\nindent <- function(x) {\n  paste0(\"  \", gsub(\"\\n\", \"\\n  \", x))\n}\n\nformat_code <- function(children, indent = FALSE) {\n  if (indent) {\n    paste0(\"\\n\", paste0(indent(children), collapse = \"\\n\"), \"\\n\")\n  } else {\n    paste(children, collapse = \"\")\n  }\n}\nhtml_tags$body <- function(...) {\n  dots <- dots_partition(...)\n  attribs <- html_attributes(dots$named)\n  children <- map_chr(dots$unnamed, escape)\n\n  html(paste0(\n    \"<body\", attribs, \">\",\n    format_code(children, indent = TRUE),\n    \"<\/body>\"\n  ))\n}\nwith_html(\n  body(\n    h1(\"A heading\", id = \"first\"),\n    p(\"Some text &\", b(\"some bold text.\")),\n    img(src = \"myimg.png\", width = 100, height = 100)\n  )\n)\n#> <HTML>\n#> <body>\n#>   <h1 id='first'>A heading<\/h1>\n#>   <p>Some text &amp;<b>some bold text.<\/b><\/p>\n#>   <img src='myimg.png' width='100' height='100' />\n#> <\/body>"},{"path":"translation.html","id":"latex-exercises-21.3.8","chapter":"21 Translation","heading":"21.2 LaTeX (Exercises 21.3.8)","text":"didn’t manage solve exercises, ’d recommend checking solutions official solutions manual.Q1. Add escaping. special symbols escaped adding backslash front \\, $, %. Just HTML, ’ll need make sure don’t end double-escaping. ’ll need create small S3 class use function operators. also allow embed arbitrary LaTeX needed.Q2. Complete DSL support functions plotmath supports.","code":""},{"path":"translation.html","id":"session-information-16","chapter":"21 Translation","heading":"21.3 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  purrr       * 1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang       * 1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"debugging.html","id":"debugging","chapter":"22 Debugging","heading":"22 Debugging","text":"exercises.","code":""},{"path":"measuring-performance.html","id":"measuring-performance","chapter":"23 Measuring performance","heading":"23 Measuring performance","text":"Attaching needed libraries:","code":"\nlibrary(profvis, warn.conflicts = FALSE)\nlibrary(dplyr, warn.conflicts = FALSE)"},{"path":"measuring-performance.html","id":"profiling-exercises-23.2.4","chapter":"23 Measuring performance","heading":"23.1 Profiling (Exercises 23.2.4)","text":"Q1. Profile following function torture = TRUE. surprising? Read source code rm() figure ’s going .A1. Let’s source functions mentioned exercises.First, try without torture = TRUE: returns meaningful results.mentioned docs, setting torture = TRUETriggers garbage collection every torture memory allocation call.process somehow never seems finish crashes RStudio session stops!question says documentation rm() may provide clues:still couldn’t figure . recommend checking official answer.","code":"\nf <- function(n = 1e5) {\n  x <- rep(1, n)\n  rm(x)\n}\nsource(\"profiling-exercises.R\")\nprofvis(f())\n#> Error in parse_rprof_lines(lines, expr_source): No parsing data available. Maybe your function was too fast?\nprofvis(f(), torture = TRUE)\nrm\n#> function (..., list = character(), pos = -1, envir = as.environment(pos), \n#>     inherits = FALSE) \n#> {\n#>     if (...length()) {\n#>         dots <- match.call(expand.dots = FALSE)$...\n#>         if (!all(vapply(dots, function(x) is.symbol(x) || is.character(x), \n#>             NA, USE.NAMES = FALSE))) \n#>             stop(\"... must contain names or character strings\")\n#>         list <- .Primitive(\"c\")(list, vapply(dots, as.character, \n#>             \"\"))\n#>     }\n#>     .Internal(remove(list, envir, inherits))\n#> }\n#> <bytecode: 0x56191ab34df0>\n#> <environment: namespace:base>"},{"path":"measuring-performance.html","id":"microbenchmarking-exercises-23.3.3","chapter":"23 Measuring performance","heading":"23.2 Microbenchmarking (Exercises 23.3.3)","text":"Q1. Instead using bench::mark(), use built-function system.time(). system.time() much less precise, ’ll need repeat operation many times loop, divide find average time operation, code .estimates system.time() compare bench::mark()? different?A1. Let’s benchmark first using two approaches:Now can compare results alternatives:comparison reveals two approaches yield quite similar results. Slight differences exact values possibly due differences precision timers used internally functions.Q2. two ways compute square root vector. think fastest? slowest? Use microbenchmarking test answers.A2. Microbenchmarking ways compute square root vector mentioned chapter.specialized primitive function sqrt() (written C) fastest way compute square root.","code":"\nn <- 1e6\nsystem.time(for (i in 1:n) sqrt(x)) / n\nsystem.time(for (i in 1:n) x^0.5) / n\nn <- 1e6\nx <- runif(100)\n\n# bench -------------------\n\nbench_df <- bench::mark(\n  sqrt(x),\n  x^0.5,\n  iterations = n,\n  time_unit = \"us\"\n)\n\nt_bench_df <- bench_df %>%\n  select(expression, time) %>%\n  rowwise() %>%\n  mutate(bench_mean = mean(unlist(time))) %>%\n  ungroup() %>%\n  select(-time)\n\n# system.time -------------------\n\n# garbage collection performed immediately before the timing\nt1_systime_gc <- system.time(for (i in 1:n) sqrt(x), gcFirst = TRUE) / n\nt2_systime_gc <- system.time(for (i in 1:n) x^0.5, gcFirst = TRUE) / n\n\n# garbage collection not performed immediately before the timing\nt1_systime_nogc <- system.time(for (i in 1:n) sqrt(x), gcFirst = FALSE) / n\nt2_systime_nogc <- system.time(for (i in 1:n) x^0.5, gcFirst = FALSE) / n\n\nt_systime_df <- tibble(\n  \"expression\" = bench_df$expression,\n  \"systime_with_gc\" = c(t1_systime_gc[\"elapsed\"], t2_systime_gc[\"elapsed\"]),\n  \"systime_with_nogc\" = c(t1_systime_nogc[\"elapsed\"], t2_systime_nogc[\"elapsed\"])\n) %>%\n  mutate(\n    systime_with_gc = systime_with_gc * 1e6, # in microseconds\n    systime_with_nogc = systime_with_nogc * 1e6 # in microseconds\n  )\n# note that system time columns report time in microseconds\nfull_join(t_bench_df, t_systime_df, by = \"expression\")\n#> # A tibble: 2 × 4\n#>   expression bench_mean systime_with_gc systime_with_nogc\n#>   <bch:expr>   <bch:tm>           <dbl>             <dbl>\n#> 1 sqrt(x)       677.2ns           0.629             0.428\n#> 2 x^0.5          2.23µs           2.02              2.00\nx^(1 / 2)\nexp(log(x) / 2)\nx <- runif(1000)\n\nbench::mark(\n  sqrt(x),\n  x^0.5,\n  x^(1 / 2),\n  exp(log(x) / 2),\n  iterations = 1000\n) %>%\n  select(expression, median) %>%\n  arrange(median)\n#> # A tibble: 4 × 2\n#>   expression      median\n#>   <bch:expr>    <bch:tm>\n#> 1 sqrt(x)         3.02µs\n#> 2 exp(log(x)/2)  12.66µs\n#> 3 x^0.5          18.96µs\n#> 4 x^(1/2)        19.01µs"},{"path":"measuring-performance.html","id":"session-information-17","chapter":"23 Measuring performance","heading":"23.3 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bench         1.1.3   2023-05-04 [1] RSPM\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  dplyr       * 1.1.4   2023-11-17 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  generics      0.1.3   2022-07-05 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  htmlwidgets   1.6.4   2023-12-06 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  profmem       0.6.0   2020-12-13 [1] RSPM\n#>  profvis     * 0.3.8   2023-05-02 [1] RSPM\n#>  purrr         1.0.2   2023-08-10 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  stringi       1.8.3   2023-12-11 [1] RSPM\n#>  stringr       1.5.1   2023-11-14 [1] RSPM\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tidyselect    1.2.0   2022-10-10 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"},{"path":"improving-performance.html","id":"improving-performance","chapter":"24 Improving performance","heading":"24 Improving performance","text":"Attaching needed libraries:","code":"\nlibrary(ggplot2)\nlibrary(dplyr)\nlibrary(purrr)"},{"path":"improving-performance.html","id":"exercises-24.3.1","chapter":"24 Improving performance","heading":"24.1 Exercises 24.3.1","text":"Q1. faster alternatives lm()? specifically designed work larger datasets?A1. Faster alternatives lm() can found visiting CRAN Task View: High-Performance Parallel Computing R page.available options:speedglm::speedlm() (large datasets)speedglm::speedlm() (large datasets)biglm::biglm() (specifically designed data large fit memory)biglm::biglm() (specifically designed data large fit memory)RcppEigen::fastLm() (using Eigen linear algebra library)RcppEigen::fastLm() (using Eigen linear algebra library)High performances can obtained packages especially R linked optimized BLAS, ATLAS. can check information using sessionInfo():Comparing performance different alternatives:results might change depending size dataset, performance benefits accruing bigger dataset.experiment different algorithms find one fits needs dataset best.Q2. package implements version match() ’s faster repeated look ups? much faster ?A2. package (respective function) fastmatch::fmatch()7.documentation function notes:slightly faster built-version uses specialized code, addition retains hash table within table object can re-used, dramatically reducing look-time especially large table.small vector, fmatch() slightly faster, order magnitude., larger vector, fmatch() orders magnitude faster! ⚡can also look hash table:Additionally, fastmatch provides equivalent familiar infix operator:Q3. List four functions (just base R) convert string date time object. strengths weaknesses?A3. four functions convert string date time object:base::.POSIXct()base::.POSIXlt()lubridate::ymd_hms()fasttime::fastPOSIXct()can also compare performance:many packages implement way convert string date time object. , see CRAN Task View: Time Series AnalysisQ4. packages provide ability compute rolling mean?A4. packages respective functions provide way compute rolling mean:RcppRoll::roll_mean()data.table::frollmean()roll::roll_mean()zoo::rollmean()slider::slide_dbl()Q5. alternatives optim()?A5. optim() function provides general-purpose optimization. noted docs:General-purpose optimization based Nelder–Mead, quasi-Newton conjugate-gradient algorithms. includes option box-constrained optimization simulated annealing.many alternatives exact one want choose depend type optimization like .available options can seen CRAN Task View: Optimization Mathematical Programming.","code":"\nsessInfo <- sessionInfo()\nsessInfo$matprod\n#> [1] \"default\"\nsessInfo$LAPACK\n#> [1] \"/usr/lib/x86_64-linux-gnu/openblas-pthread/libopenblasp-r0.3.20.so\"\nlibrary(gapminder)\n\n# having a look at the data\nglimpse(gapminder)\n#> Rows: 1,704\n#> Columns: 6\n#> $ country   <fct> \"Afghanistan\", \"Afghanistan\", \"Afghanist…\n#> $ continent <fct> Asia, Asia, Asia, Asia, Asia, Asia, Asia…\n#> $ year      <int> 1952, 1957, 1962, 1967, 1972, 1977, 1982…\n#> $ lifeExp   <dbl> 28.801, 30.332, 31.997, 34.020, 36.088, …\n#> $ pop       <int> 8425333, 9240934, 10267083, 11537966, 13…\n#> $ gdpPercap <dbl> 779.4453, 820.8530, 853.1007, 836.1971, …\n\nbench::mark(\n  \"lm\"       = stats::lm(lifeExp ~ continent * gdpPercap, gapminder),\n  \"speedglm\" = speedglm::speedlm(lifeExp ~ continent * gdpPercap, gapminder),\n  \"biglm\"    = biglm::biglm(lifeExp ~ continent * gdpPercap, gapminder),\n  \"fastLm\"   = RcppEigen::fastLm(lifeExp ~ continent * gdpPercap, gapminder),\n  check      = FALSE,\n  iterations = 1000\n)[1:5]\n#> # A tibble: 4 × 5\n#>   expression      min   median `itr/sec` mem_alloc\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 lm          881.5µs  916.5µs     1045.    1.27MB\n#> 2 speedglm      1.6ms   1.66ms      601.   75.01MB\n#> 3 biglm       772.1µs 809.23µs     1174.   589.1KB\n#> 4 fastLm        993µs   1.02ms      978.    4.48MB\nlibrary(fastmatch, warn.conflicts = FALSE)\n\nsmall_vec <- c(\"a\", \"b\", \"x\", \"m\", \"n\", \"y\")\n\nlength(small_vec)\n#> [1] 6\n\nbench::mark(\n  \"base\" = match(c(\"x\", \"y\"), small_vec),\n  \"fastmatch\" = fmatch(c(\"x\", \"y\"), small_vec)\n)[1:5]\n#> # A tibble: 2 × 5\n#>   expression      min   median `itr/sec` mem_alloc\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 base         1.18µs   1.24µs   756897.     2.8KB\n#> 2 fastmatch    1.02µs   1.06µs   904225.    2.66KB\nlarge_vec <- c(rep(c(\"a\", \"b\"), 1e4), \"x\", rep(c(\"m\", \"n\"), 1e6), \"y\")\n\nlength(large_vec)\n#> [1] 2020002\n\nbench::mark(\n  \"base\" = match(c(\"x\", \"y\"), large_vec),\n  \"fastmatch\" = fmatch(c(\"x\", \"y\"), large_vec)\n)[1:5]\n#> # A tibble: 2 × 5\n#>   expression      min   median `itr/sec` mem_alloc\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 base         20.6ms  21.95ms      43.0    31.4MB\n#> 2 fastmatch   992.1ns   1.04µs  897837.         0B\nfmatch.hash(c(\"x\", \"y\"), small_vec)\n#> [1] \"a\" \"b\" \"x\" \"m\" \"n\" \"y\"\n#> attr(,\".match.hash\")\n#> <hash table>\nlibrary(fastmatch)\n\nsmall_vec <- c(\"a\", \"b\", \"x\", \"m\", \"n\", \"y\")\n\nc(\"x\", \"y\") %in% small_vec\n#> [1] TRUE TRUE\n\nc(\"x\", \"y\") %fin% small_vec\n#> [1] TRUE TRUE\nbase::as.POSIXct(\"2022-05-05 09:23:22\")\n#> [1] \"2022-05-05 09:23:22 UTC\"\nbase::as.POSIXlt(\"2022-05-05 09:23:22\")\n#> [1] \"2022-05-05 09:23:22 UTC\"\nlubridate::ymd_hms(\"2022-05-05-09-23-22\")\n#> [1] \"2022-05-05 09:23:22 UTC\"\nfasttime::fastPOSIXct(\"2022-05-05 09:23:22\")\n#> [1] \"2022-05-05 09:23:22 UTC\"\nbench::mark(\n  \"as.POSIXct\" = base::as.POSIXct(\"2022-05-05 09:23:22\"),\n  \"as.POSIXlt\" = base::as.POSIXlt(\"2022-05-05 09:23:22\"),\n  \"ymd_hms\" = lubridate::ymd_hms(\"2022-05-05-09-23-22\"),\n  \"fastPOSIXct\" = fasttime::fastPOSIXct(\"2022-05-05 09:23:22\"),\n  check = FALSE,\n  iterations = 1000\n)\n#> # A tibble: 4 × 6\n#>   expression       min   median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1 as.POSIXct   29.74µs   31.4µs    31531.        0B     0   \n#> 2 as.POSIXlt   20.59µs  21.79µs    45434.        0B     0   \n#> 3 ymd_hms       2.25ms   2.33ms      425.    21.5KB     3.43\n#> 4 fastPOSIXct   1.16µs   1.22µs   766780.        0B     0"},{"path":"improving-performance.html","id":"exercises-24.4.3","chapter":"24 Improving performance","heading":"24.2 Exercises 24.4.3","text":"Q1. ’s difference rowSums() .rowSums()?A1. documentation functions state:versions initial dot name (.colSums() etc) ‘bare-bones’ versions use programming: apply numeric (like) matrices name result.Looking source code,rowSums() function number checks validate arguments acceptable.rowSums() directly proceeds computation using internal code built R interpreterBut comparable performance:Q2. Make faster version chisq.test() computes chi-square test statistic input two numeric vectors missing values. can try simplifying chisq.test() coding mathematical definition.A2. function supposed accept two numeric vectors without missing values, can make chisq.test() less work removing code corresponding following :checks data frame matrix inputsgoodness--fit testsimulating p-valueschecking missing valuesThis leaves us much simpler, bare bones implementation:, indeed, custom function performs slightly better8 base equivalent:Q3. Can make faster version table() case input two integer vectors missing values? Can use speed chi-square test?A3. order make leaner version table(), can take similar approach trim unnecessary input checks light new API accepting just two vectors without missing values. can remove following components code:extracting data objects entered ... argumentdealing missing valuesother input validation checksIn addition removal, can also use fastmatch::fmatch() instead match():custom function indeed performs slightly better:can also use function custom chi-squared test function see performance improves :, indeed, new version custom function performs even better previously :","code":"\nrowSums\n#> function (x, na.rm = FALSE, dims = 1L) \n#> {\n#>     if (is.data.frame(x)) \n#>         x <- as.matrix(x)\n#>     if (!is.array(x) || length(dn <- dim(x)) < 2L) \n#>         stop(\"'x' must be an array of at least two dimensions\")\n#>     if (dims < 1L || dims > length(dn) - 1L) \n#>         stop(\"invalid 'dims'\")\n#>     p <- prod(dn[-(id <- seq_len(dims))])\n#>     dn <- dn[id]\n#>     z <- if (is.complex(x)) \n#>         .Internal(rowSums(Re(x), prod(dn), p, na.rm)) + (0+1i) * \n#>             .Internal(rowSums(Im(x), prod(dn), p, na.rm))\n#>     else .Internal(rowSums(x, prod(dn), p, na.rm))\n#>     if (length(dn) > 1L) {\n#>         dim(z) <- dn\n#>         dimnames(z) <- dimnames(x)[id]\n#>     }\n#>     else names(z) <- dimnames(x)[[1L]]\n#>     z\n#> }\n#> <bytecode: 0x563353356400>\n#> <environment: namespace:base>\n.rowSums\n#> function (x, m, n, na.rm = FALSE) \n#> .Internal(rowSums(x, m, n, na.rm))\n#> <bytecode: 0x56335cd3f7a8>\n#> <environment: namespace:base>\nx <- cbind(x1 = 3, x2 = c(4:1e4, 2:1e5))\n\nbench::mark(\n  \"rowSums\" = rowSums(x),\n  \".rowSums\" = .rowSums(x, dim(x)[[1]], dim(x)[[2]])\n)[1:5]\n#> # A tibble: 2 × 5\n#>   expression      min   median `itr/sec` mem_alloc\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 rowSums       821µs   1.29ms      859.     859KB\n#> 2 .rowSums      817µs   1.28ms      833.     859KB\nmy_chisq_test <- function(x, y) {\n  x <- table(x, y)\n  n <- sum(x)\n\n  nr <- as.integer(nrow(x))\n  nc <- as.integer(ncol(x))\n\n  sr <- rowSums(x)\n  sc <- colSums(x)\n  E <- outer(sr, sc, \"*\") / n\n  v <- function(r, c, n) c * r * (n - r) * (n - c) / n^3\n  V <- outer(sr, sc, v, n)\n  dimnames(E) <- dimnames(x)\n\n  STATISTIC <- sum((abs(x - E))^2 / E)\n  PARAMETER <- (nr - 1L) * (nc - 1L)\n  PVAL <- pchisq(STATISTIC, PARAMETER, lower.tail = FALSE)\n\n  names(STATISTIC) <- \"X-squared\"\n  names(PARAMETER) <- \"df\"\n\n  structure(\n    list(\n      statistic = STATISTIC,\n      parameter = PARAMETER,\n      p.value = PVAL,\n      method = \"Pearson's Chi-squared test\",\n      observed = x,\n      expected = E,\n      residuals = (x - E) / sqrt(E),\n      stdres = (x - E) / sqrt(V)\n    ),\n    class = \"htest\"\n  )\n}\nm <- c(rep(\"a\", 1000), rep(\"b\", 9000))\nn <- c(rep(c(\"x\", \"y\"), 5000))\n\nbench::mark(\n  \"base\" = chisq.test(m, n)$statistic[[1]],\n  \"custom\" = my_chisq_test(m, n)$statistic[[1]]\n)[1:5]\n#> # A tibble: 2 × 5\n#>   expression      min   median `itr/sec` mem_alloc\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 base          906µs    926µs     1071.    1.57MB\n#> 2 custom        706µs    722µs     1371.    1.13MB\nmy_table <- function(x, y) {\n  x_sorted <- sort(unique(x))\n  y_sorted <- sort(unique(y))\n\n  x_length <- length(x_sorted)\n  y_length <- length(y_sorted)\n\n  bin <-\n    fastmatch::fmatch(x, x_sorted) +\n    x_length * fastmatch::fmatch(y, y_sorted) -\n    x_length\n\n  y <- tabulate(bin, x_length * y_length)\n\n  y <- array(\n    y,\n    dim = c(x_length, y_length),\n    dimnames = list(x = x_sorted, y = y_sorted)\n  )\n\n  class(y) <- \"table\"\n  y\n}\nx <- c(rep(\"a\", 1000), rep(\"b\", 9000))\ny <- c(rep(c(\"x\", \"y\"), 5000))\n\n# `check = FALSE` because the custom function has an additional attribute:\n# \".match.hash\"\nbench::mark(\n  \"base\" = table(x, y),\n  \"custom\" = my_table(x, y),\n  check = FALSE\n)[1:5]\n#> # A tibble: 2 × 5\n#>   expression      min   median `itr/sec` mem_alloc\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 base          638µs    653µs     1515.     960KB\n#> 2 custom        354µs    359µs     2730.     488KB\nmy_chisq_test2 <- function(x, y) {\n  x <- my_table(x, y)\n  n <- sum(x)\n\n  nr <- as.integer(nrow(x))\n  nc <- as.integer(ncol(x))\n\n  sr <- rowSums(x)\n  sc <- colSums(x)\n  E <- outer(sr, sc, \"*\") / n\n  v <- function(r, c, n) c * r * (n - r) * (n - c) / n^3\n  V <- outer(sr, sc, v, n)\n  dimnames(E) <- dimnames(x)\n\n  STATISTIC <- sum((abs(x - E))^2 / E)\n  PARAMETER <- (nr - 1L) * (nc - 1L)\n  PVAL <- pchisq(STATISTIC, PARAMETER, lower.tail = FALSE)\n\n  names(STATISTIC) <- \"X-squared\"\n  names(PARAMETER) <- \"df\"\n\n  structure(\n    list(\n      statistic = STATISTIC,\n      parameter = PARAMETER,\n      p.value = PVAL,\n      method = \"Pearson's Chi-squared test\",\n      observed = x,\n      expected = E,\n      residuals = (x - E) / sqrt(E),\n      stdres = (x - E) / sqrt(V)\n    ),\n    class = \"htest\"\n  )\n}\nm <- c(rep(\"a\", 1000), rep(\"b\", 9000))\nn <- c(rep(c(\"x\", \"y\"), 5000))\n\nbench::mark(\n  \"base\" = chisq.test(m, n)$statistic[[1]],\n  \"custom\" = my_chisq_test2(m, n)$statistic[[1]]\n)[1:5]\n#> # A tibble: 2 × 5\n#>   expression      min   median `itr/sec` mem_alloc\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 base          912µs    940µs     1057.    1.28MB\n#> 2 custom        420µs    429µs     2296.  594.45KB"},{"path":"improving-performance.html","id":"exercises-24.5.1","chapter":"24 Improving performance","heading":"24.3 Exercises 24.5.1","text":"Q1. density functions, e.g., dnorm(), common interface. arguments vectorised ? rnorm(10, mean = 10:1) ?A1. density function family following interface:Reading documentation reveals following parameters vectorized:\nx, q, p, mean, sd.means something like following work:, functions don’t multiple vectorized parameters, won’t. example,following function call generates 10 random numbers (since n = 10) 10 different distributions means supplied vector 10:1.Q2. Compare speed apply(x, 1, sum) rowSums(x) varying sizes x.A2. can write custom function vary number rows matrix extract data frame comparing performance two functions.Plotting data reveals rowSums(x) O(1) behavior, O(n) behavior.Q3. can use crossprod() compute weighted sum? much faster naive sum(x * w)?A3. functions provide way compute weighted sum:benchmarking performance reveals latter significantly faster former!","code":"\ndnorm(x, mean = 0, sd = 1, log = FALSE)\npnorm(q, mean = 0, sd = 1, lower.tail = TRUE, log.p = FALSE)\nqnorm(p, mean = 0, sd = 1, lower.tail = TRUE, log.p = FALSE)\nrnorm(n, mean = 0, sd = 1)\nrnorm(c(1, 2, 3), mean = c(0, -1, 5))\n#> [1] 1.124335 0.930398 3.844935\npnorm(c(1, 2, 3), mean = c(0, -1, 5), log.p = c(FALSE, TRUE, TRUE))\n#> [1] 0.84134475 0.99865010 0.02275013\nrnorm(n = 10, mean = 10:1)\n#>  [1]  8.2421770  9.3920474  7.1362118  7.5789906  5.2551688\n#>  [6]  6.0143714  4.6147891  1.1096247  2.8759129 -0.6756857\nbenc_perform <- function(nRow, nCol = 100) {\n  x <- matrix(data = rnorm(nRow * nCol), nrow = nRow, ncol = nCol)\n\n  bench::mark(\n    rowSums(x),\n    apply(x, 1, sum)\n  )[1:5]\n}\n\nnRowList <- list(10, 100, 500, 1000, 5000, 10000, 50000, 100000)\n\nnames(nRowList) <- as.character(nRowList)\n\nbenchDF <- map_dfr(\n  .x = nRowList,\n  .f = ~ benc_perform(.x),\n  .id = \"nRows\"\n) %>%\n  mutate(nRows = as.numeric(nRows))\nggplot(\n  benchDF,\n  aes(\n    x = as.numeric(nRows),\n    y = median,\n    group = as.character(expression),\n    color = as.character(expression)\n  )\n) +\n  geom_point() +\n  geom_line() +\n  labs(\n    x = \"Number of Rows\",\n    y = \"Median Execution Time\",\n    colour = \"Function used\"\n  )\nx <- c(1:6, 2, 3)\nw <- rnorm(length(x))\n\ncrossprod(x, w)[[1]]\n#> [1] 15.94691\nsum(x * w)[[1]]\n#> [1] 15.94691\nbench::mark(\n  crossprod(x, w)[[1]],\n  sum(x * w)[[1]],\n  iterations = 1e6\n)[1:5]\n#> # A tibble: 2 × 5\n#>   expression                min   median `itr/sec` mem_alloc\n#>   <bch:expr>           <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 crossprod(x, w)[[1]]    901ns    961ns   966191.        0B\n#> 2 sum(x * w)[[1]]         461ns    511ns  1759060.        0B"},{"path":"rewriting-r-code-in-c.html","id":"rewriting-r-code-in-c","chapter":"25 Rewriting R code in C++","heading":"25 Rewriting R code in C++","text":"","code":"\nlibrary(Rcpp, warn.conflicts = FALSE)"},{"path":"rewriting-r-code-in-c.html","id":"getting-started-with-c-exercises-25.2.6","chapter":"25 Rewriting R code in C++","heading":"25.1 Getting started with C++ (Exercises 25.2.6)","text":"Q1. basics C++ hand, ’s now great time practice reading writing simple C++ functions. following functions, read code figure corresponding base R function . might understand every part code yet, able figure basics function .A1.f1() mean():f2() cumsum():f3() ():f4() Position():f5() pmin():Q2. practice function writing skills, convert following functions C++. now, assume inputs missing values.().().cumprod(), cummin(), cummax().cumprod(), cummin(), cummax().diff(). Start assuming lag 1, generalise lag n.diff(). Start assuming lag 1, generalise lag n.range().range().var(). Read approaches can take Wikipedia. Whenever implementing numerical algorithm, ’s always good check already known problem.var(). Read approaches can take Wikipedia. Whenever implementing numerical algorithm, ’s always good check already known problem.A2. performance benefits going observed function primitive since already tuned max R performance. , expect performance gain diff() var().()cumprod()cumminC()cummaxC()diff()range()var()","code":"#include <Rcpp.h>\nusing namespace Rcpp;\n\n// [[Rcpp::export]]\ndouble f1(NumericVector x) {\n  int n = x.size();\n  double y = 0;\n\n  for(int i = 0; i < n; ++i) {\n    y += x[i] / n;\n  }\n  return y;\n}\n\n// [[Rcpp::export]]\nNumericVector f2(NumericVector x) {\n  int n = x.size();\n  NumericVector out(n);\n\n  out[0] = x[0];\n  for(int i = 1; i < n; ++i) {\n    out[i] = out[i - 1] + x[i];\n  }\n  return out;\n}\n\n// [[Rcpp::export]]\nbool f3(LogicalVector x) {\n  int n = x.size();\n\n  for(int i = 0; i < n; ++i) {\n    if (x[i]) return true;\n  }\n  return false;\n}\n\n// [[Rcpp::export]]\nint f4(Function pred, List x) {\n  int n = x.size();\n\n  for(int i = 0; i < n; ++i) {\n    LogicalVector res = pred(x[i]);\n    if (res[0]) return i + 1;\n  }\n  return 0;\n}\n\n// [[Rcpp::export]]\nNumericVector f5(NumericVector x, NumericVector y) {\n  int n = std::max(x.size(), y.size());\n  NumericVector x1 = rep_len(x, n);\n  NumericVector y1 = rep_len(y, n);\n\n  NumericVector out(n);\n\n  for (int i = 0; i < n; ++i) {\n    out[i] = std::min(x1[i], y1[i]);\n  }\n\n  return out;\n}\nx <- c(1, 2, 3, 4, 5, 6)\n\nf1(x)\n#> [1] 3.5\nmean(x)\n#> [1] 3.5\nx <- c(1, 3, 5, 6)\n\nf2(x)\n#> [1]  1  4  9 15\ncumsum(x)\n#> [1]  1  4  9 15\nx1 <- c(TRUE, FALSE, FALSE, TRUE)\nx2 <- c(FALSE, FALSE)\n\nf3(x1)\n#> [1] TRUE\nany(x1)\n#> [1] TRUE\n\nf3(x2)\n#> [1] FALSE\nany(x2)\n#> [1] FALSE\nx <- list(\"a\", TRUE, \"m\", 2)\n\nf4(is.numeric, x)\n#> [1] 4\nPosition(is.numeric, x)\n#> [1] 4\nv1 <- c(1, 3, 4, 5, 6, 7)\nv2 <- c(1, 2, 7, 2, 8, 1)\n\nf5(v1, v2)\n#> [1] 1 2 4 2 6 1\npmin(v1, v2)\n#> [1] 1 2 4 2 6 1\nis.primitive(all)\n#> [1] TRUE\nis.primitive(cumprod)\n#> [1] TRUE\nis.primitive(diff)\n#> [1] FALSE\nis.primitive(range)\n#> [1] TRUE\nis.primitive(var)\n#> [1] FALSE#include <vector>\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nbool allC(std::vector<bool> x)\n{\n    for (const auto& xElement : x)\n    {\n        if (!xElement) return false;\n    }\n\n    return true;\n}\nv1 <- rep(TRUE, 10)\nv2 <- c(rep(TRUE, 5), rep(FALSE, 5))\n\nall(v1)\n#> [1] TRUE\nallC(v1)\n#> [1] TRUE\n\nall(v2)\n#> [1] FALSE\nallC(v2)\n#> [1] FALSE\n\n# performance benefits?\nbench::mark(\n  all(c(rep(TRUE, 1000), rep(FALSE, 1000))),\n  allC(c(rep(TRUE, 1000), rep(FALSE, 1000))),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression                                      min\n#>   <bch:expr>                                 <bch:tm>\n#> 1 all(c(rep(TRUE, 1000), rep(FALSE, 1000)))    6.13µs\n#> 2 allC(c(rep(TRUE, 1000), rep(FALSE, 1000)))   8.73µs\n#>     median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1   6.49µs   152211.    15.8KB        0\n#> 2   9.15µs   106881.    18.3KB        0#include <vector>\n\n// [[Rcpp::export]]\nstd::vector<double> cumprodC(const std::vector<double> &x)\n{\n    std::vector<double> out{x};\n\n    for (std::size_t i = 1; i < x.size(); i++)\n    {\n        out[i] = out[i - 1] * x[i];\n    }\n\n    return out;\n}\nv1 <- c(10, 4, 6, 8)\n\ncumprod(v1)\n#> [1]   10   40  240 1920\ncumprodC(v1)\n#> [1]   10   40  240 1920\n\n# performance benefits?\nbench::mark(\n  cumprod(v1),\n  cumprodC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression        min   median `itr/sec` mem_alloc\n#>   <bch:expr>   <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 cumprod(v1)   89.99ns 101.05ns  8156808.        0B\n#> 2 cumprodC(v1)   1.45µs   1.55µs   612672.    6.62KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0#include <vector>\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::vector<double> cumminC(const std::vector<double> &x)\n{\n    std::vector<double> out{x};\n\n    for (std::size_t i = 1; i < x.size(); i++)\n    {\n        out[i] = (out[i] < out[i - 1]) ? out[i] : out[i - 1];\n    }\n\n    return out;\n}\nv1 <- c(3:1, 2:0, 4:2)\n\ncummin(v1)\n#> [1] 3 2 1 1 1 0 0 0 0\ncumminC(v1)\n#> [1] 3 2 1 1 1 0 0 0 0\n\n# performance benefits?\nbench::mark(\n  cummin(v1),\n  cumminC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression       min   median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1 cummin(v1)  120.02ns 131.08ns  5672695.        0B        0\n#> 2 cumminC(v1)   1.56µs   1.87µs   507394.    6.62KB        0#include <vector>\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::vector<double> cummaxC(const std::vector<double> &x)\n{\n    std::vector<double> out{x};\n\n    for (std::size_t i = 1; i < x.size(); i++)\n    {\n        out[i] = (out[i] > out[i - 1]) ? out[i] : out[i - 1];\n    }\n    \n    return out;\n}\nv1 <- c(3:1, 2:0, 4:2)\n\ncummax(v1)\n#> [1] 3 3 3 3 3 3 4 4 4\ncummaxC(v1)\n#> [1] 3 3 3 3 3 3 4 4 4\n\n# performance benefits?\nbench::mark(\n  cummax(v1),\n  cummaxC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression       min   median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1 cummax(v1)  110.01ns 121.07ns  6496158.        0B        0\n#> 2 cummaxC(v1)   1.58µs   1.76µs   545922.    6.62KB        0#include <vector>\n#include <functional>\n#include <algorithm>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::vector<double> diffC(const std::vector<double> &x, int lag)\n{\n    std::vector<double> vec_start;\n    std::vector<double> vec_lagged;\n    std::vector<double> vec_diff;\n\n    for (std::size_t i = lag; i < x.size(); i++)\n    {\n        vec_lagged.push_back(x[i]);\n    }\n\n    for (std::size_t i = 0; i < (x.size() - lag); i++)\n    {\n        vec_start.push_back(x[i]);\n    }\n\n    std::transform(\n        vec_lagged.begin(), vec_lagged.end(),\n        vec_start.begin(), std::back_inserter(vec_diff),\n        std::minus<double>());\n\n    return vec_diff;\n}\nv1 <- c(1, 2, 4, 8, 13)\nv2 <- c(1, 2, NA, 8, 13)\n\ndiff(v1, 2)\n#> [1] 3 6 9\ndiffC(v1, 2)\n#> [1] 3 6 9\n\ndiff(v2, 2)\n#> [1] NA  6 NA\ndiffC(v2, 2)\n#> [1] NA  6 NA\n\n# performance benefits?\nbench::mark(\n  diff(v1, 2),\n  diffC(v1, 2),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression        min   median `itr/sec` mem_alloc\n#>   <bch:expr>   <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 diff(v1, 2)    3.88µs   4.46µs   217285.        0B\n#> 2 diffC(v1, 2)   1.89µs   2.13µs   452394.    2.49KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0#include <iostream>\n#include <vector>\n#include <algorithm>\nusing namespace std;\n\n// [[Rcpp::export]]\nstd::vector<double> rangeC(std::vector<double> x)\n{\n    std::vector<double> rangeVec{0.0, 0.0};\n\n    rangeVec.at(0) = *std::min_element(x.begin(), x.end());\n    rangeVec.at(1) = *std::max_element(x.begin(), x.end());\n\n    return rangeVec;\n}\nv1 <- c(10, 4, 6, 8)\n\nrange(v1)\n#> [1]  4 10\nrangeC(v1)\n#> [1]  4 10\n\n# performance benefits?\nbench::mark(\n  range(v1),\n  rangeC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression      min   median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1 range(v1)    1.68µs   1.94µs   481017.        0B        0\n#> 2 rangeC(v1)   1.47µs   1.68µs   569929.    6.62KB        0#include <vector>\n#include <cmath>\n#include <numeric>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\ndouble variance(std::vector<double> x)\n{\n    double sumSquared{0};\n\n    double mean = std::accumulate(x.begin(), x.end(), 0.0) / x.size();\n\n    for (const auto& xElement : x)\n    {\n        sumSquared += pow(xElement - mean, 2.0);\n    }\n\n    return sumSquared / (x.size() - 1);\n}\nv1 <- c(1, 4, 7, 8)\n\nvar(v1)\n#> [1] 10\nvariance(v1)\n#> [1] 10\n\n# performance benefits?\nbench::mark(\n  var(v1),\n  variance(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression        min   median `itr/sec` mem_alloc\n#>   <bch:expr>   <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 var(v1)        7.54µs   8.51µs   111092.        0B\n#> 2 variance(v1)   1.49µs   1.66µs   567349.    6.62KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0"},{"path":"rewriting-r-code-in-c.html","id":"missing-values-exercises-25.4.5","chapter":"25 Rewriting R code in C++","heading":"25.2 Missing values (Exercises 25.4.5)","text":"Q1. Rewrite functions Exercise 25.2.6 deal missing values. na.rm true, ignore missing values. na.rm false, return missing value input contains missing values. good functions practice min(), max(), range(), mean(), var().A1. create version range() deals missing values. principle applies others:Q2. Rewrite cumsum() diff() can handle missing values. Note functions slightly complicated behaviour.A2. cumsum() docs say:NA value x causes corresponding following elements return value NA, integer overflow cumsum (warning).Similarly, diff() docs say:NA’s propagate.Therefore, functions don’t allow removing missing values NAs propagate.seen examples , diffC() already behaves way.Similarly, cumsumC() propagates NAs well.","code":"#include <iostream>\n#include <vector>\n#include <algorithm>\n#include <math.h>\n#include <Rcpp.h>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::vector<double> rangeC_NA(std::vector<double> x, bool removeNA = true)\n{\n    std::vector<double> rangeVec{0.0, 0.0};\n\n    bool naPresent = std::any_of(\n        x.begin(),\n        x.end(),\n        [](double d)\n        { return isnan(d); });\n\n    if (naPresent)\n    {\n        if (removeNA)\n        {\n            std::remove(x.begin(), x.end(), NAN);\n        }\n        else\n        {\n            rangeVec.at(0) = NA_REAL; // NAN;\n            rangeVec.at(1) = NA_REAL; // NAN;\n\n            return rangeVec;\n        }\n    }\n\n    rangeVec.at(0) = *std::min_element(x.begin(), x.end());\n    rangeVec.at(1) = *std::max_element(x.begin(), x.end());\n\n    return rangeVec;\n}\nv1 <- c(10, 4, NA, 6, 8)\n\nrange(v1, na.rm = FALSE)\n#> [1] NA NA\nrangeC_NA(v1, FALSE)\n#> [1] NA NA\n\nrange(v1, na.rm = TRUE)\n#> [1]  4 10\nrangeC_NA(v1, TRUE)\n#> [1]  4 10#include <Rcpp.h>\nusing namespace Rcpp;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nNumericVector cumsumC(NumericVector x) {\n  int n = x.size();\n  NumericVector out(n);\n  \n  out[0] = x[0];\n  for(int i = 1; i < n; ++i) {\n    out[i] = out[i - 1] + x[i];\n  }\n  \n  return out;\n}\nv1 <- c(1, 2, 3, 4)\nv2 <- c(1, 2, NA, 4)\n\ncumsum(v1)\n#> [1]  1  3  6 10\ncumsumC(v1)\n#> [1]  1  3  6 10\n\ncumsum(v2)\n#> [1]  1  3 NA NA\ncumsumC(v2)\n#> [1]  1  3 NA NA"},{"path":"rewriting-r-code-in-c.html","id":"standard-template-library-exercises-25.5.7","chapter":"25 Rewriting R code in C++","heading":"25.3 Standard Template Library (Exercises 25.5.7)","text":"Q1. practice using STL algorithms data structures, implement following using R functions C++, using hints provided:A1.median.default() using partial_sort.%% using unordered_set find() count() methods.unique() using unordered_set (challenge: one line!).Note functions comparable. far can see, way get output R version function using unordered_set data structure.can make comparable version using set data structure:min() using std::min(), max() using std::max()..min() using min_element, .max() using max_element.setdiff(), union(), intersect() integers using sorted ranges set_union, set_intersection set_difference.Note following C++ implementations given functions strictly equivalent R versions. far can see, way identical satisfying specifications mentioned question.union()intersect()setdiff()","code":"#include <iostream>\n#include <vector>\n#include <algorithm>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\ndouble medianC(std::vector<double> &x)\n{\n    int middleIndex = static_cast<int>(x.size() / 2);\n\n    std::partial_sort(x.begin(), x.begin() + middleIndex, x.end());\n\n    // for even number of observations\n    if (x.size() % 2 == 0)\n    {\n        return (x[middleIndex - 1] + x[middleIndex]) / 2;\n    }\n\n    return x[middleIndex];\n}\nv1 <- c(1, 3, 3, 6, 7, 8, 9)\nv2 <- c(1, 2, 3, 4, 5, 6, 8, 9)\n\nmedian.default(v1)\n#> [1] 6\nmedianC(v1)\n#> [1] 6\n\nmedian.default(v2)\n#> [1] 4.5\nmedianC(v2)\n#> [1] 4.5\n\n# performance benefits?\nbench::mark(\n  median.default(v2),\n  medianC(v2),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression              min   median `itr/sec` mem_alloc\n#>   <bch:expr>         <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 median.default(v2)  22.17µs  24.29µs    40480.        0B\n#> 2 medianC(v2)          1.49µs   1.73µs   546971.    2.49KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0#include <vector>\n#include <unordered_set>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::vector<bool> matchC(const std::vector<double> &x, const std::vector<double> &table)\n{\n    std::unordered_set<double> tableUnique(table.begin(), table.end());\n    std::vector<bool> out;\n\n    for (const auto &xElem : x)\n    {\n        out.push_back(tableUnique.find(xElem) != tableUnique.end() ? true : false);\n    }\n\n    return out;\n}\nx1 <- c(3, 4, 8)\nx2 <- c(1, 2, 3, 3, 4, 4, 5, 6)\n\nx1 %in% x2\n#> [1]  TRUE  TRUE FALSE\nmatchC(x1, x2)\n#> [1]  TRUE  TRUE FALSE\n\n# performance benefits?\nbench::mark(\n  x1 %in% x2,\n  matchC(x1, x2),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression          min   median `itr/sec` mem_alloc\n#>   <bch:expr>     <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 x1 %in% x2     881.03ns   1.11µs   870819.        0B\n#> 2 matchC(x1, x2)   2.15µs   2.41µs   399261.    6.62KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0#include <unordered_set>\n#include <vector>\n#include <iostream>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::unordered_set<double> uniqueC(const std::vector<double> &x)\n{\n    std::unordered_set<double> xSet(x.begin(), x.end());\n\n    return xSet;\n}\nv1 <- c(1, 3, 3, 6, 7, 8, 9)\n\nunique(v1)\n#> [1] 1 3 6 7 8 9\nuniqueC(v1)\n#> [1] 9 8 7 6 3 1#include <set>\n#include <vector>\n#include <iostream>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::set<double> uniqueC2(const std::vector<double> &x)\n{\n    std::set<double> xSet(x.begin(), x.end());\n\n    return xSet;\n}\nv1 <- c(1, 3, 3, 6, 7, 8, 9)\n\nunique(v1)\n#> [1] 1 3 6 7 8 9\nuniqueC2(v1)\n#> [1] 1 3 6 7 8 9\n\n# performance benefits?\nbench::mark(\n  unique(v1),\n  uniqueC2(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression        min   median `itr/sec` mem_alloc\n#>   <bch:expr>   <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 unique(v1)      2.1µs   2.41µs   355209.        0B\n#> 2 uniqueC2(v1)   1.63µs   1.85µs   502571.    6.62KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0#include <iostream>\n#include <vector>\n#include <algorithm>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nconst double minC(const std::vector<double> &x)\n{\n     return *std::min_element(x.begin(), x.end());\n}\n\n// [[Rcpp::export]]\nconst double maxC(std::vector<double> x)\n{\n     return *std::max_element(x.begin(), x.end());\n}\nv1 <- c(3, 3, 6, 1, 9, 7, 8)\n\nmin(v1)\n#> [1] 1\nminC(v1)\n#> [1] 1\n\n# performance benefits?\nbench::mark(\n  min(v1),\n  minC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression      min   median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1 min(v1)    230.04ns 250.93ns  3402682.        0B        0\n#> 2 minC(v1)     1.42µs   1.56µs   606873.    6.62KB        0\n\nmax(v1)\n#> [1] 9\nmaxC(v1)\n#> [1] 9\n\n# performance benefits?\nbench::mark(\n  max(v1),\n  maxC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression      min   median `itr/sec` mem_alloc `gc/sec`\n#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>\n#> 1 max(v1)    211.06ns  241.1ns  3455199.        0B        0\n#> 2 maxC(v1)     1.41µs   1.66µs   571011.    6.62KB        0#include <vector>\n#include <algorithm>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nint which_maxC(std::vector<double> &x)\n{\n    int maxIndex = std::distance(x.begin(), std::max_element(x.begin(), x.end()));\n  \n    // R is 1-index based, while C++ is 0-index based\n    return maxIndex + 1;\n}\n\n// [[Rcpp::export]]\nint which_minC(std::vector<double> &x)\n{\n    int minIndex = std::distance(x.begin(), std::min_element(x.begin(), x.end()));\n  \n    // R is 1-index based, while C++ is 0-index based\n    return minIndex + 1;\n}\nv1 <- c(3, 3, 6, 1, 9, 7, 8)\n\nwhich.min(v1)\n#> [1] 4\nwhich_minC(v1)\n#> [1] 4\n\n# performance benefits?\nbench::mark(\n  which.min(v1),\n  which_minC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression          min   median `itr/sec` mem_alloc\n#>   <bch:expr>     <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 which.min(v1)  401.05ns  421.1ns  2181122.        0B\n#> 2 which_minC(v1)   1.44µs    1.6µs   591193.    6.62KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0\n\nwhich.max(v1)\n#> [1] 5\nwhich_maxC(v1)\n#> [1] 5\n\n# performance benefits?\nbench::mark(\n  which.max(v1),\n  which_maxC(v1),\n  iterations = 100\n)\n#> # A tibble: 2 × 6\n#>   expression          min   median `itr/sec` mem_alloc\n#>   <bch:expr>     <bch:tm> <bch:tm>     <dbl> <bch:byt>\n#> 1 which.max(v1)  410.01ns 490.92ns  1870056.        0B\n#> 2 which_maxC(v1)   1.41µs   1.57µs   585441.    6.62KB\n#>   `gc/sec`\n#>      <dbl>\n#> 1        0\n#> 2        0#include <algorithm>\n#include <iostream>\n#include <vector>\n#include <set>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::set<int> unionC(std::vector<int> &v1, std::vector<int> &v2)\n{\n    std::sort(v1.begin(), v1.end());\n    std::sort(v2.begin(), v2.end());\n\n    std::vector<int> union_vec(v1.size() + v2.size());\n    auto it = std::set_union(v1.begin(), v1.end(), v2.begin(), v2.end(), union_vec.begin());\n\n    union_vec.resize(it - union_vec.begin());\n    std::set<int> union_set(union_vec.begin(), union_vec.end());\n\n    return union_set;\n}\nv1 <- c(1, 4, 5, 5, 5, 6, 2)\nv2 <- c(4, 1, 6, 8)\n\nunion(v1, v2)\n#> [1] 1 4 5 6 2 8\nunionC(v1, v2)\n#> [1] 1 2 4 5 6 8#include <algorithm>\n#include <iostream>\n#include <vector>\n#include <set>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::set<int> intersectC(std::vector<int> &v1, std::vector<int> &v2)\n{\n    std::sort(v1.begin(), v1.end());\n    std::sort(v2.begin(), v2.end());\n\n    std::vector<int> union_vec(v1.size() + v2.size());\n    auto it = std::set_intersection(v1.begin(), v1.end(), v2.begin(), v2.end(), union_vec.begin());\n\n    union_vec.resize(it - union_vec.begin());\n    std::set<int> union_set(union_vec.begin(), union_vec.end());\n\n    return union_set;\n}\nv1 <- c(1, 4, 5, 5, 5, 6, 2)\nv2 <- c(4, 1, 6, 8)\n\nintersect(v1, v2)\n#> [1] 1 4 6\nintersectC(v1, v2)\n#> [1] 1 4 6#include <algorithm>\n#include <iostream>\n#include <vector>\n#include <set>\nusing namespace std;\n// [[Rcpp::plugins(cpp11)]]\n\n// [[Rcpp::export]]\nstd::set<int> setdiffC(std::vector<int> &v1, std::vector<int> &v2)\n{\n    std::sort(v1.begin(), v1.end());\n    std::sort(v2.begin(), v2.end());\n\n    std::vector<int> union_vec(v1.size() + v2.size());\n    auto it = std::set_difference(v1.begin(), v1.end(), v2.begin(), v2.end(), union_vec.begin());\n\n    union_vec.resize(it - union_vec.begin());\n    std::set<int> union_set(union_vec.begin(), union_vec.end());\n\n    return union_set;\n}\nv1 <- c(1, 4, 5, 5, 5, 6, 2)\nv2 <- c(4, 1, 6, 8)\n\nsetdiff(v1, v2)\n#> [1] 5 2\nsetdiffC(v1, v2)\n#> [1] 2 5"},{"path":"rewriting-r-code-in-c.html","id":"session-information-18","chapter":"25 Rewriting R code in C++","heading":"25.4 Session information","text":"","code":"\nsessioninfo::session_info(include_base = TRUE)\n#> ─ Session info ───────────────────────────────────────────\n#>  setting  value\n#>  version  R version 4.3.2 (2023-10-31)\n#>  os       Ubuntu 22.04.3 LTS\n#>  system   x86_64, linux-gnu\n#>  ui       X11\n#>  language (EN)\n#>  collate  C.UTF-8\n#>  ctype    C.UTF-8\n#>  tz       UTC\n#>  date     2024-01-07\n#>  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)\n#> \n#> ─ Packages ───────────────────────────────────────────────\n#>  package     * version date (UTC) lib source\n#>  base        * 4.3.2   2023-11-01 [3] local\n#>  bench         1.1.3   2023-05-04 [1] RSPM\n#>  bookdown      0.37    2023-12-01 [1] RSPM\n#>  bslib         0.6.1   2023-11-28 [1] RSPM\n#>  cachem        1.0.8   2023-05-01 [1] RSPM\n#>  cli           3.6.2   2023-12-11 [1] RSPM\n#>  compiler      4.3.2   2023-11-01 [3] local\n#>  datasets    * 4.3.2   2023-11-01 [3] local\n#>  digest        0.6.33  2023-07-07 [1] RSPM\n#>  downlit       0.4.3   2023-06-29 [1] RSPM\n#>  evaluate      0.23    2023-11-01 [1] RSPM\n#>  fansi         1.0.6   2023-12-08 [1] RSPM\n#>  fastmap       1.1.1   2023-02-24 [1] RSPM\n#>  fs            1.6.3   2023-07-20 [1] RSPM\n#>  glue          1.6.2   2022-02-24 [1] RSPM\n#>  graphics    * 4.3.2   2023-11-01 [3] local\n#>  grDevices   * 4.3.2   2023-11-01 [3] local\n#>  htmltools     0.5.7   2023-11-03 [1] RSPM\n#>  jquerylib     0.1.4   2021-04-26 [1] RSPM\n#>  jsonlite      1.8.8   2023-12-04 [1] RSPM\n#>  knitr         1.45    2023-10-30 [1] RSPM\n#>  lifecycle     1.0.4   2023-11-07 [1] RSPM\n#>  magrittr    * 2.0.3   2022-03-30 [1] RSPM\n#>  memoise       2.0.1   2021-11-26 [1] RSPM\n#>  methods     * 4.3.2   2023-11-01 [3] local\n#>  pillar        1.9.0   2023-03-22 [1] RSPM\n#>  pkgconfig     2.0.3   2019-09-22 [1] RSPM\n#>  profmem       0.6.0   2020-12-13 [1] RSPM\n#>  R6            2.5.1   2021-08-19 [1] RSPM\n#>  Rcpp        * 1.0.11  2023-07-06 [1] RSPM\n#>  rlang         1.1.2   2023-11-04 [1] RSPM\n#>  rmarkdown     2.25    2023-09-18 [1] RSPM\n#>  sass          0.4.8   2023-12-06 [1] RSPM\n#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM\n#>  stats       * 4.3.2   2023-11-01 [3] local\n#>  tibble        3.2.1   2023-03-20 [1] RSPM\n#>  tools         4.3.2   2023-11-01 [3] local\n#>  utf8          1.2.4   2023-10-22 [1] RSPM\n#>  utils       * 4.3.2   2023-11-01 [3] local\n#>  vctrs         0.6.5   2023-12-01 [1] RSPM\n#>  withr         2.5.2   2023-10-30 [1] RSPM\n#>  xfun          0.41    2023-11-01 [1] RSPM\n#>  xml2          1.3.6   2023-12-04 [1] RSPM\n#>  yaml          2.3.8   2023-12-11 [1] RSPM\n#> \n#>  [1] /home/runner/work/_temp/Library\n#>  [2] /opt/R/4.3.2/lib/R/site-library\n#>  [3] /opt/R/4.3.2/lib/R/library\n#> \n#> ──────────────────────────────────────────────────────────"}]
diff --git a/subsetting.html b/subsetting.html
index e083325d..66daf2fe 100644
--- a/subsetting.html
+++ b/subsetting.html
@@ -417,7 +417,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -500,7 +500,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/trade-offs.html b/trade-offs.html
index 23ce0da9..2f94da85 100644
--- a/trade-offs.html
+++ b/trade-offs.html
@@ -111,7 +111,7 @@ <h1>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/translation.html b/translation.html
index f6664080..b3179d9e 100644
--- a/translation.html
+++ b/translation.html
@@ -338,7 +338,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -419,7 +419,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">
diff --git a/vectors.html b/vectors.html
index 77119931..c36718dc 100644
--- a/vectors.html
+++ b/vectors.html
@@ -244,7 +244,7 @@ <h2>
 <span><span class="co">#&gt;     names(object) &lt;- nm</span></span>
 <span><span class="co">#&gt;     object</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x5635a7f8a880&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x55c0caac1880&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:stats&gt;</span></span></code></pre></div>
 <p>Given this function signature, we can see why, when no first argument is given, the result is still a named vector.</p>
 <div class="sourceCode" id="cb46"><pre class="downlit sourceCode r">
@@ -268,7 +268,7 @@ <h2>
 <span><span class="co">#&gt;         dimnames(obj) &lt;- NULL</span></span>
 <span><span class="co">#&gt;     obj</span></span>
 <span><span class="co">#&gt; }</span></span>
-<span><span class="co">#&gt; &lt;bytecode: 0x5635a7395578&gt;</span></span>
+<span><span class="co">#&gt; &lt;bytecode: 0x55c0c9ecc578&gt;</span></span>
 <span><span class="co">#&gt; &lt;environment: namespace:base&gt;</span></span></code></pre></div>
 <p><code><a href="https://rdrr.io/r/base/unname.html">unname()</a></code> removes existing names (or dimnames) by setting them to <code>NULL</code>.</p>
 <div class="sourceCode" id="cb48"><pre class="downlit sourceCode r">
@@ -804,7 +804,7 @@ <h2>
 <span><span class="co">#&gt;  collate  C.UTF-8</span></span>
 <span><span class="co">#&gt;  ctype    C.UTF-8</span></span>
 <span><span class="co">#&gt;  tz       UTC</span></span>
-<span><span class="co">#&gt;  date     2023-12-31</span></span>
+<span><span class="co">#&gt;  date     2024-01-07</span></span>
 <span><span class="co">#&gt;  pandoc   3.1.8 @ /usr/bin/ (via rmarkdown)</span></span>
 <span><span class="co">#&gt; </span></span>
 <span><span class="co">#&gt; ─ Packages ───────────────────────────────────────────────</span></span>
@@ -885,7 +885,7 @@ <h2>
 <footer class="bg-primary text-light mt-5"><div class="container"><div class="row">
 
   <div class="col-12 col-md-6 mt-3">
-    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2023-12-31.</p>
+    <p>"<strong>Advanced R Exercises</strong>" was written by <a class="text-light" href="https://sites.google.com/site/indrajeetspatilmorality/">Indrajeet Patil</a>. It was last built on 2024-01-07.</p>
   </div>
 
   <div class="col-12 col-md-6 mt-3">