Weak compositions #3573
Weak compositions #3573
Conversation
|
|
||
| # I have no idea what to call these getter functions | ||
| base(W::WeakCompositions) = W.n | ||
| parts(W::WeakCompositions) = W.k |
There was a problem hiding this comment.
Do we even need these at all?
There was a problem hiding this comment.
Since the fields are called .n and .k, I prefer getter functions. I could of course also give the fields more meaningful names, but then I still would not know what to call them :)
src/Combinatorics/EnumerativeCombinatorics/weak_compositions.jl
Outdated
Show resolved
Hide resolved
joschmitt
force-pushed
the
js/compositions
branch
from
253c5e4
to
a410ed7
Compare
| @test !iszero(phi[0]) | ||
| @test !iszero(phi[1]) | ||
| @test !iszero(phi[0]) | ||
| @test compose(map(dom_tot, 1), phi[0]) == compose(phi[1], map(cod_tot, 1)) | ||
|
|
||
| phi = interp(H0[3]) | ||
| dom_tot = domain(phi) | ||
| cod_tot = codomain(phi) | ||
| @test !iszero(phi[0]) | ||
| @test !iszero(phi[0]) | ||
| @test !iszero(phi[1]) |
There was a problem hiding this comment.
@HechtiDerLachs I made these adjustments to the tests: The order produced by weak_compositions is the opposite of what MultiIndicesOfDegree did and apparently this swaps the 2nd and 3rd generator of H0 in this test. I assume that this is still correct and the comment in line 195-196 suggests it's expected, but could you have a look whether this makes sense?
There was a problem hiding this comment.
I think it's fine. Thanks for the work!
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## master #3573 +/- ##
==========================================
+ Coverage 81.72% 81.90% +0.17%
==========================================
Files 573 574 +1
Lines 77490 79272 +1782
==========================================
+ Hits 63332 64925 +1593
- Misses 14158 14347 +189
|
There was a problem hiding this comment.
Once @HechtiDerLachs has approved the changes to the DoubleAndHyperComplexes, I am happy with this change. Thanks @joschmitt for working on this.
|
Hechti is on holidays. We can wait with merging until he is back as far as I'm concerned. |
|
Design question: What should a function like I would have said 'error' because asking for the number of partitions of -1 does not make sense, but apparently so far we returned 0 and this is even tested. What do you think? @ulthiel @mjrodgers (or anybody else interested in combinatorics) |
0 is almost always better in my experience. There are often applications where we want to compute a sum of such numbers and take all of the ones that “don’t make sense” as 0. It’s a pain to have to catch errors to deal with this. |
|
0 probably: number_of_partitions(-1) is the cardinality of the set of partitions of -1, this set is empty, so its cardinality is 0. |
|
Question: What is the weak in weak compositions? Isn't this just compositions? Also I have https://github.com/ulthiel/Oscar.jl/blob/ut/compositions/experimental/JuLie/compositions.jl in which I have put quite some energy (but somehow lost track of opening a PR). |
0 is an allowed summand in weak compositions. |
Sounds reasonable, thank you. |
The parts in a weak composition are allowed to be 0, in a composition they have to be positive. |
|
Excellent. |
joschmitt
force-pushed
the
js/compositions
branch
from
6066234
to
2575cfe
Compare
| @test !iszero(phi[0]) | ||
| @test !iszero(phi[1]) | ||
| @test !iszero(phi[0]) | ||
| @test compose(map(dom_tot, 1), phi[0]) == compose(phi[1], map(cod_tot, 1)) | ||
|
|
||
| phi = interp(H0[3]) | ||
| dom_tot = domain(phi) | ||
| cod_tot = codomain(phi) | ||
| @test !iszero(phi[0]) | ||
| @test !iszero(phi[0]) | ||
| @test !iszero(phi[1]) |
There was a problem hiding this comment.
I think it's fine. Thanks for the work!
|
The booktests failed. I think this is already fixed by #3604, but restarting CI here to make sure. Once it is finished, I will go ahead and merge (unless there are any objections). |
Thanks. |
As remarked in the context of #3147 we have three (known) implementations of weak compositions of
nintodparts. (To my knowledge, a weak composition is a sequence ofdnon-negative integers that sum up ton.)This pull request introduces a type
WeakCompositionanalogously toPartitionand an iterator over all weak compositions ofnintodparts. There is not much more functionality for weak compositions so far because I personally am happy if I can iterate them.This unifies the mentioned three implementations:
weak_compositionsis removed/replacedmonomials_of_degree-- the user shouldn't notice anythingMultiIndicesOfDegreeis removed. Warning:MultiIndicesOfDegree(d, n)has to be replaced byweak_compositions(n, d)!