Add missing snapshots - part 1

R/BoxCox.R

@@ -160,7 +160,7 @@ bc_trans <- function(x, lambda, eps = .001) {
   if (any(x <= 0)) {
     cli::cli_warn(
       "Applying Box-Cox transformation to non-positive data in column \\
-      {names(lambda)}"
+      {.field {names(lambda)}}."
     )
   }
 

R/case_weights.R

@@ -88,10 +88,10 @@ get_case_weights <- function(info, .data, call = rlang::caller_env()) {
     if (!is.numeric(res)) {
       cli::cli_abort(
         c(
-          "x" = "{.arg {wt_col}} has a {.code case_weights} role,\\
-                 but is not numeric.",
-          "i" = "{.arg {wt_col}} is {.obj_type_friendly {wt_col}}."
-          ),
+          x = "{.field {wt_col}} has a {.code case_weights} role and should be 
+              numeric, but is {.obj_type_friendly {wt_col}}.",
+          i = "Only numeric case weights are supported in recipes."
+        ),
         call = call
       )
     }

R/colcheck.R

@@ -81,6 +81,9 @@ bake.check_cols <- function(object, new_data, ...) {
   new_cols <- names(new_data)
   missing <- setdiff(original_cols, new_cols)
   if (length(missing) > 0) {
+    # This is functionally not reachable after we added ptype checking in
+    # https://github.com/tidymodels/recipes/pull/1330
+    # but it feels too harsh to deprecate this check function.
     cli::cli_abort(c(
       x = "{cli::qty(length(missing))}The following column{?s} {?is/are} \\
       missing from {.arg new_data}:",

R/count.R

@@ -129,14 +129,6 @@ prep.step_count <- function(x, training, info = NULL, ...) {
   col_name <- recipes_eval_select(x$terms, training, info)
   check_type(training[, col_name], types = c("string", "factor", "ordered"))
 
-  if (length(col_name) > 1) {
-    cli::cli_abort(c(
-      x = "The selector should select at most a single variable.",
-      i = "The following {length(col_name)} were selected: \\
-          {.and {.var {col_name}}}."
-    ))
-  }
-
   step_count_new(
     terms = x$terms,
     role = x$role,

R/cut.R

@@ -113,6 +113,13 @@ prep.step_cut <- function(x, training, info = NULL, ...) {
   col_names <- recipes_eval_select(x$terms, training, info)
   check_type(training[, col_names], types = c("double", "integer"))
 
+  if (!is.numeric(x$breaks)) {
+    cli::cli_abort(
+      "{.arg breaks} must be a numeric vector, \\
+      not {.obj_type_friendly {breaks}}."
+    )
+  }
+
   all_breaks <- vector("list", length(col_names))
   names(all_breaks) <- col_names
   for (col_name in col_names) {
@@ -133,19 +140,6 @@ prep.step_cut <- function(x, training, info = NULL, ...) {
 }
 
 create_full_breaks <- function(var, breaks) {
-  if (!is.numeric(var)) {
-    cli::cli_abort(
-      "{.arg var} must be a numeric vector, not {.obj_type_friendly {var}}."
-    )
-  }
-
-  if (!is.numeric(breaks)) {
-    cli::cli_abort(
-      "{.arg breaks} must be a numeric vector, \\
-      not {.obj_type_friendly {breaks}}."
-    )
-  }
-
   if (min(var) < min(breaks)) {
     breaks <- c(min(var), breaks)
   }
@@ -204,12 +198,6 @@ cut_var <- function(var, breaks, include_outside_range) {
 # the levels when bake.recipe itself is called. Moreover,
 # it is cleaner to show it in this way.
 adjust_levels_min_max <- function(x) {
-  if (!is.factor(x)) {
-    cli::cli_abort(
-      "{.arg x} must be a factor, not {.obj_type_friendly {x}}.",
-      .internal = TRUE
-    )
-  }
   levs <- levels(x)
   if (length(levs) == 1) {
     return(factor(rep("[min,max]", length(x))))

R/date.R

@@ -111,8 +111,7 @@ step_date <-
         offenders <- features[!features %in% feat]
 
         cli::cli_abort(c(
-          x = "Possible values of {.arg features} should include:",
-          "*" = "{.or {.val {feat}}}.",
+          x = "Possible values of {.arg features} are {.or {.val {feat}}}.",
           i = "Invalid values were: {.val {offenders}}."
         ))
       }

R/dummy.R

@@ -271,13 +271,9 @@ bake.step_dummy <- function(object, new_data, ...) {
     # the original (see the note above)
     is_ordered <- attr(levels, "dataClasses") == "ordered"
 
-    if (is.null(levels_values)) {
-      cli::cli_abort("Factor level values not recorded in {.var col_name}.")
-    }
-
     if (length(levels_values) == 1) {
       cli::cli_abort(
-        "Only one factor level in {.var col_name}: {levels_values}."
+        "Only one factor level in {.var {col_name}}: {.val {levels_values}}."
       )
     }
 

R/dummy_multi_choice.R

@@ -142,8 +142,6 @@ prep.step_dummy_multi_choice <- function(x, training, info = NULL, ...) {
   col_names <- recipes_eval_select(x$terms, training, info)
   check_type(training[, col_names], types = c("nominal", "logical"))
 
-  multi_dummy_check_type(training[, col_names])
-
   levels <- purrr::map(training[, col_names], levels)
   levels <- vctrs::list_unchop(levels, ptype = character(), name_spec = rlang::zap())
   levels <- levels[!is.na(levels)]
@@ -165,23 +163,6 @@ prep.step_dummy_multi_choice <- function(x, training, info = NULL, ...) {
   )
 }
 
-multi_dummy_check_type <- function(dat, call = rlang::caller_env()) {
-  is_good <- function(x) {
-    is.factor(x) | is.character(x) | all(is.na(x))
-  }
-
-  all_good <- vapply(dat, is_good, logical(1))
-  if (!all(all_good)) {
-    offenders <- names(dat)[!all_good]
-    cli::cli_abort(c(
-      "x" = "All columns selected for the step should be \\
-            factor, character, or NA. The following were not:",
-      "*" = "{.var {offenders}}."
-    ), call = call)
-  }
-  invisible(all_good)
-}
-
 #' @export
 bake.step_dummy_multi_choice <- function(object, new_data, ...) {
   col_names <- object$input

R/impute_knn.R

@@ -118,7 +118,7 @@ step_impute_knn <-
     if (length(options) > 0) {
       if (any(!(opt_nms %in% c("eps", "nthread")))) {
         cli::cli_abort(
-          "Valid values for {.arg options} are {.val eps}, and {.val nthread}."
+          "Valid values for {.arg options} are {.val eps} and {.val nthread}."
         )
       }
       if (all(opt_nms != "nthread")) {

R/impute_mean.R

@@ -113,14 +113,8 @@ trim <- function(x, trim) {
   # Adapted from mean.default
   x <- sort(x, na.last = TRUE)
   na_ind <- is.na(x)
-  if (!is.numeric(trim) || length(trim) != 1L) {
-    cli::cli_abort("{.arg trim} must be numeric of length one.")
-  }
   n <- length(x[!na_ind])
   if (trim > 0 && n) {
-    if (is.complex(x)) {
-      cli::cli_abort("Trimmed means are not defined for complex data.")
-    }
     if (trim >= 0.5)
       return(stats::median(x[!na_ind], na.rm = FALSE))
     lo <- floor(n * trim) + 1

R/interact.R

@@ -162,7 +162,8 @@ prep.step_interact <- function(x, training, info = NULL, ...) {
     )
     if (!is_formula(tmp_terms)) {
       cli::cli_abort(
-        "{.arg terms} must be a formula. Not {.obj_type_friendly {term}}."
+        "{.arg terms} must be a formula, not {.obj_type_friendly {term}}.",
+        .internal = TRUE
       )
     }
 
@@ -209,7 +210,7 @@ prep.step_interact <- function(x, training, info = NULL, ...) {
     vars <-
       unique(unlist(lapply(make_new_formula(int_terms), all.vars)))
     var_check <- info[info$variable %in% vars, ]
-    if (any(var_check$type == "nominal")) {
+    if (any(vapply(var_check$type, function(x) "nominal" %in% x, logical(1)))) {
       cli::cli_warn(
         "Categorical variables used in {.fn step_interact} should probably be \\
         avoided; This can lead to differences in dummy variable values that \\
@@ -401,7 +402,10 @@ find_selectors <- function(f) {
     list()
   } else {
     # User supplied incorrect input
-    cli::cli_abort("Don't know how to handle type {typeof(f)}.")
+    cli::cli_abort(
+      "Don't know how to handle type {.code {typeof(f)}}.", 
+      .internal = TRUE
+    )
   }
 }
 
@@ -418,7 +422,10 @@ replace_selectors <- function(x, elem, value) {
     map_pairlist(x, replace_selectors, elem, value)
   } else {
     # User supplied incorrect input
-    cli::cli_abort("Don't know how to handle type {typeof(f)}.")
+    cli::cli_abort(
+      "Don't know how to handle type {.code {typeof(f)}}.", 
+      .internal = TRUE
+    )
   }
 }
 

R/kpca_rbf.R

@@ -134,7 +134,7 @@ prep.step_kpca_rbf <- function(x, training, info = NULL, ...) {
       cli::cli_abort(c(
         x = "Failed with error:",
         i = as.character(kprc)
-      ))
+      ), .internal = TRUE)
     }
   } else {
     kprc <- NULL

R/nnmf_sparse.R

@@ -170,9 +170,10 @@ prep.step_nnmf_sparse <- function(x, training, info = NULL, ...) {
     } else {
       na_w <- sum(is.na(nnm$w))
       if (na_w > 0) {
-        cli::cli_abort(
-          "The NNMF loadings are missing. The penalty may have been to high."
-        )
+        cli::cli_abort(c(
+          x = "The NNMF loadings are missing.",
+          i = "The penalty may have been too high or missing values are present in data."
+        ))
       } else {
         nnm <- list(x_vars = col_names, w = nnm$w)
         rownames(nnm$w) <- col_names

R/pls.R

@@ -344,7 +344,7 @@ prep.step_pls <- function(x, training, info = NULL, ...) {
     res <- try(rlang::eval_tidy(cl), silent = TRUE)
     if (inherits(res, "try-error")) {
       cli::cli_warn(
-        "{.fn step_pls failed with error: {as.character(res))}.",
+        "{.fn step_pls} failed with error: {as.character(res)}.",
       )
       res <- list(x_vars = x_names, y_vars = y_names)
     } else {

R/profile.R

@@ -122,10 +122,10 @@ step_profile <- function(recipe,
       i = "See {.help [?step_profile](recipes::step_profile)} for information."
     ))
   }
-  if (all(sort(names(grid)) == c("len", "ptcl"))) {
+  if (!identical(sort(names(grid)), c("len", "pctl"))) {
     cli::cli_abort(c(
       x = "`grid` should have two named elements {.field len} and \\
-          {.field ptcl}, not {sort(names(grid))}.",
+          {.field pctl}, not {.val {sort(names(grid))}}.",
       i = "See {.help [?step_profile](recipes::step_profile)} for information."
     ))
   }

R/relevel.R

@@ -84,19 +84,12 @@ step_relevel_new <-
 #' @export
 prep.step_relevel <- function(x, training, info = NULL, ...) {
   col_names <- recipes_eval_select(x$terms, training, info)
-  check_type(training[, col_names], types = c("string", "factor", "ordered"))
+  check_type(training[, col_names], types = c("string", "factor"))
 
   # Get existing levels and their factor type (i.e. ordered)
   objects <- lapply(training[, col_names], get_existing_values)
   # Check to make sure that no ordered levels are provided
   order_check <- map_lgl(objects, attr, "is_ordered")
-  if (any(order_check)) {
-    offenders <- names(order_check)[order_check]
-    cli::cli_abort(
-      "Columns contain ordered factors (which cannot be releveled) \\
-      {.val {x$ref_level}}: {offenders}."
-    )
-  }
 
   # Check to make sure that the reference level exists in the factor
   ref_check <- map_lgl(objects, function(x, y) !y %in% x,

tests/testthat/_snaps/BoxCox.md

@@ -33,6 +33,22 @@
       -- Operations 
       * Box-Cox transformation on: <none> | Trained
 
+---
+
+    Code
+      recipe(~ mpg + disp, data = mtcars) %>% step_BoxCox(mpg, disp) %>% prep() %>%
+        bake(new_data = tibble(mpg = -1, disp = -1))
+    Condition
+      Warning:
+      Applying Box-Cox transformation to non-positive data in column mpg.
+      Warning:
+      Applying Box-Cox transformation to non-positive data in column disp.
+    Output
+      # A tibble: 1 x 2
+          mpg  disp
+        <dbl> <dbl>
+      1   NaN   NaN
+
 # bake method errors when needed non-standard role columns are missing
 
     Code

tests/testthat/_snaps/case-weight-functions.md

@@ -15,3 +15,13 @@
       ! There should only be a single column with the role `case_weights`.
       i In these data, there are 2 columns: `var1` and `var2`.
 
+# get_case_weights() catches non-numeric case weights
+
+    Code
+      recipe(~., data = mtcars) %>% step_normalize(all_predictors()) %>% prep()
+    Condition
+      Error in `step_normalize()`:
+      Caused by error in `prep()`:
+      x vs has a `case_weights` role and should be numeric, but is a string.
+      i Only numeric case weights are supported in recipes.
+

tests/testthat/_snaps/corr.md

@@ -52,6 +52,14 @@
       -- Operations 
       * Correlation filter on: V6, V1, V3 | Trained, ignored weights
 
+# corr_filter() warns on many NA values
+
+    Code
+      tmp <- recipe(~., data = mtcars) %>% step_corr(all_predictors()) %>% prep()
+    Condition
+      Warning:
+      Too many correlations are `NA`; skipping correlation filter.
+
 # empty printing
 
     Code

tests/testthat/_snaps/count.md

@@ -27,6 +27,16 @@
       ! Name collision occurred. The following variable names already exist:
       * `Sepal.Width`
 
+# checks for grepl arguments
+
+    Code
+      recipe(~., data = mtcars) %>% step_count(options = list(not_real_option = TRUE))
+    Condition
+      Error in `step_count()`:
+      x The following elements of `options` are not allowed:
+      * "not_real_option".
+      i Valid options are: "ignore.case", "perl", "fixed", and "useBytes".
+
 # bake method errors when needed non-standard role columns are missing
 
     Code

tests/testthat/_snaps/cut.md

@@ -34,6 +34,29 @@
       Warning:
       This will create a factor with one value only.
 
+# breaks argument are type checked
+
+    Code
+      recipe(~., data = mtcars) %>% step_cut(disp, hp, breaks = TRUE) %>% prep()
+    Condition
+      Error in `step_cut()`:
+      Caused by error:
+      ! Could not evaluate cli `{}` expression: `breaks`.
+      Caused by error:
+      ! object 'breaks' not found
+
+---
+
+    Code
+      recipe(~., data = mtcars) %>% step_cut(disp, hp, breaks = c("100", "200")) %>%
+        prep()
+    Condition
+      Error in `step_cut()`:
+      Caused by error:
+      ! Could not evaluate cli `{}` expression: `breaks`.
+      Caused by error:
+      ! object 'breaks' not found
+
 # bake method errors when needed non-standard role columns are missing
 
     Code