API,BUG: Fix copyto (and ufunc) handling of scalar cast safety #27091
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This defines scalar cast safety to be "safe", "equiv", or "unsafe" (for Python int, float, complex). Although, equiv is a bit of an odd one. We define it by using the common DType with the desired one, and assuming that this cast is at least safe. With that, we can create an appropriate array for the scalar (and the rest works fine). This means that any python float can be safely cast to e.g. float32 and any Python integer to any NumPy integer even though the actual assignment may overflow (floats) or fail (integers). `can_cast` itself is not defined here. It could be, but has the disadvantage that it won't error out on assignment later for integers.
Increadibly niche and can really never matter. But did matter for the new tests (equiv)
seberg
commented
Aug 2, 2024
| Previously, NumPy checked whether the 100 fits the ``int8_arr``. | ||
|
|
||
| This aligns ``copyto``, ``full``, and ``full_like`` with the correct NumPy 2 | ||
| behavior. |
There was a problem hiding this comment.
It is intentional that I don't mention the ufunc changes. They are ridiculously niche IMO, since they require using casting= and dtype= (or a weird casting like equiv).
|
close/reopen |
|
The 32 bit windows failure looks legit, I haven't seen it before. |
|
No, it's gh-27093 EDIT: Well, it's legit, but not related to this PR :). |
|
I'm going to try to go over this today but maybe @mhvk could also look at this since he has more familiarity with the details here. |
ngoldbaum
reviewed
Aug 2, 2024
Co-authored-by: Nathan Goldbaum <nathan.goldbaum@gmail.com>
mhvk
reviewed
Aug 2, 2024
|
I guess the docs build failure is fixed in #27102? Looks good to me, feel free to hit the merge button if the docs failure is unrelated to this PR. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This defines scalar cast safety to be "safe", "equiv", or "unsafe" (for Python int, float, complex).
Although, equiv is a bit of an odd one. We define it by using the common DType with the desired one, and assuming that this cast is at least safe. With that, we can create an appropriate array for the scalar (and the rest works fine).
This means that any python float can be safely cast to e.g. float32 and any Python integer to any NumPy integer even though the actual assignment may overflow (floats) or fail (integers).
can_castitself is not defined here. It could be, but has the disadvantage that it won't error out on assignment later for integers.Closes gh-26381, gh-27075
Still needs release notes, but ready for review. Sorry it ended up as one big commit, but besides moving some helper out of ufunc.c looking at it per-file is just as well.