Codestin Search App

mhvk · 2024-02-28T03:09:52Z

This pull request builds on #16120 to add support for axes for gufuncs with Masked inputs or outputs. It also refactors __array_ufunc__ a little, ensuring that if Masked output is passed in, its mask always get used for possible in-place mask calculations.

By checking this box, the PR author has requested that maintainers do NOT use the "Squash and Merge" button. Maintainers should respect this when possible; however, the final decision is at the discretion of the maintainer that merges the PR.

github-actions · 2024-02-28T03:10:17Z

Thank you for your contribution to Astropy! 🌌 This checklist is meant to remind the package maintainers who will review this pull request of some common things to look for.

Do the proposed changes actually accomplish desired goals?
Do the proposed changes follow the Astropy coding guidelines?
Are tests added/updated as required? If so, do they follow the Astropy testing guidelines?
Are docs added/updated as required? If so, do they follow the Astropy documentation guidelines?
Is rebase and/or squash necessary? If so, please provide the author with appropriate instructions. Also see instructions for rebase and squash.
Did the CI pass? If no, are the failures related? If you need to run daily and weekly cron jobs as part of the PR, please apply the "Extra CI" label. Codestyle issues can be fixed by the bot.
Is a change log needed? If yes, did the change log check pass? If no, add the "no-changelog-entry-needed" label. If this is a manual backport, use the "skip-changelog-checks" label unless special changelog handling is necessary.
Is this a big PR that makes a "What's new?" entry worthwhile and if so, is (1) a "what's new" entry included in this PR and (2) the "whatsnew-needed" label applied?
At the time of adding the milestone, if the milestone set requires a backport to release branch(es), apply the appropriate "backport-X.Y.x" label(s) before merge.

neutrinoceros · 2024-02-28T12:14:35Z

I think we should fix #16123 before merging this

mhvk · 2024-02-28T13:39:57Z

I think we should fix #16123 before merging this

I think it is somewhat orthogonal to what goes wrong in #16123, so it would seem fine to deal with this PR independently (#16123 will require changes to the __call__ path, which this PR barely touches, so it won't be much trouble to rebase/backport).

neutrinoceros · 2024-02-28T13:42:32Z

Ok then. I can review this one after #16120 goes in !

pllim · 2024-02-28T16:52:20Z

#16120 is merged.

…ritten in-place

neutrinoceros

Looks good, mostly minor questions (and nits) !

neutrinoceros · 2024-02-28T17:22:53Z

+        mxm2 = np.matmul(m1, m2, axes=[(0, 2), (-2, -1), (0, 1)])
+        exp2 = np.matmul(m1.unmasked, m2.unmasked, axes=[(0, 2), (-2, -1), (0, 1)])
+        mask2 = (
+            np.matmul(
+                (~m1.mask).astype(int),
+                (~m2.mask).astype(int),
+                axes=[(0, 2), (-2, -1), (0, 1)],


Just so that it's crystal clear that these are the same thing

Suggested change

mxm2 = np.matmul(m1, m2, axes=[(0, 2), (-2, -1), (0, 1)])

exp2 = np.matmul(m1.unmasked, m2.unmasked, axes=[(0, 2), (-2, -1), (0, 1)])

mask2 = (

np.matmul(

(~m1.mask).astype(int),

(~m2.mask).astype(int),

axes=[(0, 2), (-2, -1), (0, 1)],

axes = [(0, 2), (-2, -1), (0, 1)]

mxm2 = np.matmul(m1, m2, axes=axes)

exp2 = np.matmul(m1.unmasked, m2.unmasked, axes=axes)

mask2 = (

np.matmul(

(~m1.mask).astype(int),

(~m2.mask).astype(int),

axes=axes,

neutrinoceros · 2024-02-28T17:26:47Z

+                (~m2.mask).astype(int),
+                axes=[(0, 2), (-2, -1), (0, 1)],
+            )
+            != 3


I think this would be the correct generalisation ? I'm not convinced that my suggestion actually makes the code clearer, but it took me a minute to check what's going here so maybe a comment would help ?

Suggested change

!= 3

!= len(axes)

It is meant to be the size of the dimension that is operated on; I've replaced with m1.shape[axes[0][1]] (which will at least hint at what this is).

EDIT: also added a comment. In the test, I didn't want to re-use the nan logic of the code itself, since then the result is somewhat circular.

neutrinoceros · 2024-02-28T17:31:53Z

+            if out is not None:
+                np.copyto(out, False, where=where)


while this seems correct, I'm slightly confused that out and the return value may differ here (if out is an actual array). I can't help but notice that it doesn't follow the convention from numpy's API, though I'm honestly not sure what to suggest instead.

You're right, this is inconsistent. Changed...

neutrinoceros · 2024-02-28T17:53:05Z

                keepdims = kwargs.get("keepdims", False)
                in_masks = []
-                for sig, mask in zip(in_sig, masks):
+                for sig, mask, axis in zip(in_sig, masks, axes[: ufunc.nin]):


Here too, I'm wondering about arguments being ignored silently (specifically, what happens if axes has more elements than ufunc.nin). Also, wouldn't zip already drop extraneous elements even if we didn't slice here ?

axes will have been checked already to have the right length.

Well, that's not completely true: it will have at least nin elements, but nout can be omitted if they're all (). I'll adjust!

neutrinoceros · 2024-02-28T17:57:58Z

+                    # Here, some masks may need expansion, so we forego in-place.
+                    mask = self._combine_masks(in_masks, copy=False)
+                    result_masks = []
+                    for os, omask, axis in zip(out_sig, out_masks, axes[ufunc.nin :]):


same question as earlier with zip

This one was indeed right. Now added test...

mhvk · 2024-02-28T19:56:55Z

@neutrinoceros - thanks for the careful look! I think I addressed your comments (and also tried to add some extra tests to cover the somewhat rarer branches of the code).

neutrinoceros

All good now, thanks for walking me through this !

pllim

Again, I am approving by proxy to unblock merge. I'll let mhvk decide when to merge.

Thanks!

mhvk · 2024-02-28T22:30:14Z

Thanks, @pllim and @neutrinoceros!

pllim · 2024-02-28T22:40:51Z

You did the hard work. Thank YOU!

mhvk added utils Enhancement utils.masked labels Feb 28, 2024

mhvk added this to the v6.1.0 milestone Feb 28, 2024

neutrinoceros mentioned this pull request Feb 28, 2024

BUG: MaskedNDArray.__array_ufunc__ breaks for erfa.atioq #16123

Closed

mhvk added 3 commits February 28, 2024 12:06

MAINT: calculate unmasked result before calculating output mask

f46cd79

BUG: fix erfa rxp and rxr, and introduce axes support

7d9f901

MAINT: refactor __array_ufunc__ a little, ensuring output masks get w…

651e9c0

…ritten in-place

mhvk force-pushed the masked-gufunc-axes-support branch from d875da2 to 651e9c0 Compare February 28, 2024 17:06

neutrinoceros reviewed Feb 28, 2024

View reviewed changes

MAINT,BUG: Address comments, fix small bugs, add tests

8710e38

mhvk force-pushed the masked-gufunc-axes-support branch from 54ee9ca to 8710e38 Compare February 28, 2024 19:57

neutrinoceros approved these changes Feb 28, 2024

View reviewed changes

pllim approved these changes Feb 28, 2024

View reviewed changes

mhvk merged commit ac2427c into astropy:main Feb 28, 2024

mhvk deleted the masked-gufunc-axes-support branch February 28, 2024 22:30

Uh oh!

Uh oh!

Conversation

mhvk commented Feb 28, 2024

Uh oh!

github-actions Bot commented Feb 28, 2024

Uh oh!

neutrinoceros commented Feb 28, 2024

Uh oh!

mhvk commented Feb 28, 2024

Uh oh!

neutrinoceros commented Feb 28, 2024

Uh oh!

pllim commented Feb 28, 2024

Uh oh!

neutrinoceros left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mhvk Feb 28, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

neutrinoceros Feb 28, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

mhvk commented Feb 28, 2024

Uh oh!

neutrinoceros left a comment

Choose a reason for hiding this comment

Uh oh!

pllim left a comment

Choose a reason for hiding this comment

Uh oh!

mhvk commented Feb 28, 2024

Uh oh!

pllim commented Feb 28, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

mhvk Feb 28, 2024 •

edited

Loading

neutrinoceros Feb 28, 2024 •

edited

Loading