BUG: fix yaml table serialization compatibility with numpy 2 by neutrinoceros · Pull Request #16416 · astropy/astropy

neutrinoceros · 2024-05-08T16:31:09Z

Description

Combines the regression test from #16414 and cherry-picks 9cdd95d and 3be8af0 from #15065

This doesn't break any test that I could run locally against numpy 1.26, so it should be backportable, but I'm still keeping it as a draft until I see CI go green, just in case.

Fix #15792

Also resolves ~90% failures (340/380) from tests that currently rely on legacy print options as setup in

astropy/astropy/conftest.py

Line 28 in b2e38be

np.set_printoptions(legacy="1.25")

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 · 2024-05-08T16:31:39Z

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.

github-actions · 2024-05-08T16:31:41Z

👋 Thank you for your draft pull request! Do you know that you can use [ci skip] or [skip ci] in your commit messages to skip running continuous integration tests until you are ready?

pllim

Thanks!

pllim · 2024-05-08T17:33:02Z

Diff LGTM but I'll let subpackage maintainer(s) review and approve. Thanks!

astrofrog · 2024-05-08T17:36:20Z

+    def represent_float(self, data):
+        # Override to change repr(data) to str(data) since otherwise all the
+        # numpy scalars fail in not NUMPY_LT_1_20.
+        if data != data or (data == 0.0 and data == 1.0):


Just to make sure I understand, why not np.isnan?

I assume that's for performance reasons, though, that's a question for @mhvk
(math.isnan should also be faster than np.isnan for scalars)

Hmm, I vaguely recall that I just copied the snipped from how pyyaml does it, except changing repr(data) to str(data) (I clearly should have added a note with that...). I'm not quite sure why they worked this way - perhaps there are python float-like's that do not support isnan?

Ok no problem - if we can get things to work with math.isnan I think that would be preferable

For the record here's the line you apparently copied https://github.com/yaml/pyyaml/blob/48838a3c768e3d1bcab44197d800145cfd0719d6/lib/yaml/representer.py#L172
It was written 15 years ago (circa 2009), while math.isnan was introduced in Python 2.6, released in October 2008, so it's possible that the only reason they didn't use it was backward compatibility, and it just never got refactored.
Since the whole function is copied from pyyaml, it seems best to keep it as is, but I'll add a note so this knowledge doesn't get lost.

FYI @mhvk I also updated your comment. It said that numpy scalars failed on "not NUMPY_LT_1_20" which I assume was a mistake and you really meant "not NUMPY_LT_2_0"

Thanks, @neutrinoceros, that all makes sense.

mhvk

Beyond the two in-line comments, shouldn't we also get the next commit, for complex? 6b7916d4e3736eaac73ca25372c341079ae38291

mhvk · 2024-05-08T19:01:53Z

+
+
+def test_serialize_mixin_column(tmp_path):
+    # see https://github.com/astropy/astropy/issues/16414


I'm confused why our regular tests didn't catch this. We test a whole slew of different SkyCoord for all serializations (see astropy/io/tests/mixin_columns.py) - was the problem specific to frame="galactic"? If so, perhaps add an extra SkyCoord in mixin_columns? (Though note that one then needs to add quite a bit of detail further down in the file...) Though in a way also fine just as is - after all, it is the incantation that proved there was a problem.

The problem is actually not specific to frame="galactic", so I'll drop this argument to avoid confusion.
According to its docstring, astropy/io/tests/mixin_columns.py is used in ascii/tests/test_ecsv.py, fits/tests/test_connect.py, and misc/tests/test_hdf5.py, but not in yaml tests (yet). I'll see if there are any problems with using it here too.

I've wrote a new test based on test_hdf5.py::test_hdf5_mixins_qtable_to_table, which also fails on main and pass with this branch. I ditched the first test, let me know if I should keep both.

mhvk · 2024-05-08T19:08:39Z

+    def represent_float(self, data):
+        # Override to change repr(data) to str(data) since otherwise all the
+        # numpy scalars fail in not NUMPY_LT_1_20.
+        if data != data or (data == 0.0 and data == 1.0):


Hmm, I vaguely recall that I just copied the snipped from how pyyaml does it, except changing repr(data) to str(data) (I clearly should have added a note with that...). I'm not quite sure why they worked this way - perhaps there are python float-like's that do not support isnan?

neutrinoceros · 2024-05-09T08:06:20Z

shouldn't we also get the next commit, for complex?

It doesn't hurt and is clearly correct, but I should point out that it's also not tested. I'll push it now.

taldcroft · 2024-05-09T10:26:35Z


 @pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
-def test_hdf5_mixins_qtable_to_table(tmp_path):
+def code(tmp_path):


Why this change?

wow, that's a syntaxically valid mistake I must have made by trying to open vscode from another editor and typing in the wrong tab. Thanks for spotting it

taldcroft · 2024-05-09T10:29:11Z

            order: C
            shape: !!python/tuple [3]"""
        )
+


This regression test is fine but more importantly there should be YAML-level unit tests covering all the cases in the new code. I.e. serializing a simple dict structure with all the impacted data types and confirming the output and correct round-trip.

Gotcha. I'll work on it.

Actually the unit tests in question already exists, I just needed to add a pytest fixture to compensate for global settings (see #15096).
They also capture the regressions properly, so maybe we should just ditch the more complex integration test, if that's okay with you.

Excellent about the tests already being there! And yes, I'm good with dropping the integration test.

great, I rebased and greatly simplified the branch !

taldcroft · 2024-05-09T10:37:46Z

+}
+
+
+def test_fits_yaml_mixins_qtable_to_table(tmp_path):


I think this test should live in test_mixin.py instead of test_yaml.py (which should be more low-level YAML).

Putting it here is consistent with how the "inspiration" test lives in test_hdf5.py. Moving that one would be out of scope here but maybe we can move both in a follow up PR ?

…numpy 2

mhvk

Ah, thanks, now I understand why the tests didn't have a problem before - it was the use of legacy printing. With this, it looks all great!

…ility with numpy 2

mhvk · 2024-05-09T15:09:37Z

Sorry, @taldcroft, I merged perhaps too quickly! A problem with trying to do a few things while really not at work; should just have approved and left it. Though hopefully you were OK with the latest push too...

taldcroft · 2024-05-09T16:40:16Z

@mhvk @neutrinoceros - Sometimes codecov seems to give false reports, but here I don't see any coverage of the inf, nan, or 1e17 cases. Those would be good, so a follow-up PR would be a good thing.

neutrinoceros · 2024-05-09T16:42:37Z

added to my todo list for tomorrow !

…416-on-v6.1.x Backport PR #16416 on branch v6.1.x (BUG: fix yaml table serialization compatibility with numpy 2)

pllim · 2024-05-09T18:13:14Z

Thanks, all! I opened a follow-up issue at #16429

github-actions Bot added the io.misc label May 8, 2024

neutrinoceros force-pushed the io.misc/bug/yaml_serialization_numpy2 branch from 08c5a4f to 4444cf3 Compare May 8, 2024 16:31

neutrinoceros mentioned this pull request May 8, 2024

Adjust to numpy dev scalar format #15065

Merged

5 tasks done

neutrinoceros marked this pull request as ready for review May 8, 2024 16:52

neutrinoceros requested review from WilliamJamieson and matteobachetti as code owners May 8, 2024 16:52

pllim reviewed May 8, 2024

View reviewed changes

Comment thread astropy/io/misc/tests/test_yaml.py Resolved

Comment thread astropy/io/misc/tests/test_yaml.py Resolved

Comment thread docs/changes/io.misc/16416.bugfix.rst Resolved

pllim requested review from mhvk and taldcroft May 8, 2024 17:23

pllim added this to the v6.1.1 milestone May 8, 2024

pllim added Bug numpy-dev backport-v6.1.x labels May 8, 2024

astrofrog reviewed May 8, 2024

View reviewed changes

mhvk reviewed May 8, 2024

View reviewed changes

neutrinoceros force-pushed the io.misc/bug/yaml_serialization_numpy2 branch 2 times, most recently from 9a7c15d to 0f855e8 Compare May 9, 2024 07:45

taldcroft reviewed May 9, 2024

View reviewed changes

neutrinoceros force-pushed the io.misc/bug/yaml_serialization_numpy2 branch 2 times, most recently from 2d8894a to 928098f Compare May 9, 2024 12:56

neutrinoceros and others added 4 commits May 9, 2024 16:02

TST: make sure YAML unit tests correctly capture regressions against …

ae4c0b5

…numpy 2

Adjust yaml representer for numpy float repr change

fe67c38

Also fix yaml for complex

fda54df

update comment

43a42ae

neutrinoceros force-pushed the io.misc/bug/yaml_serialization_numpy2 branch from 928098f to 43a42ae Compare May 9, 2024 14:02

neutrinoceros mentioned this pull request May 9, 2024

TST: 400+ doctest failures because numpy 2.0.dev changed string representation #15095

Closed

mhvk approved these changes May 9, 2024

View reviewed changes

mhvk merged commit 539115c into astropy:main May 9, 2024

meeseeksmachine pushed a commit to meeseeksmachine/astropy that referenced this pull request May 9, 2024

Backport PR astropy#16416: BUG: fix yaml table serialization compatib…

f84b766

…ility with numpy 2

meeseeksmachine mentioned this pull request May 9, 2024

Backport PR #16416 on branch v6.1.x (BUG: fix yaml table serialization compatibility with numpy 2) #16422

Merged

neutrinoceros mentioned this pull request May 9, 2024

numpy 2 compat: tests failing with default print options #16423

Closed

neutrinoceros deleted the io.misc/bug/yaml_serialization_numpy2 branch May 9, 2024 15:15

pllim added a commit that referenced this pull request May 9, 2024

Merge pull request #16422 from meeseeksmachine/auto-backport-of-pr-16…

3f2e0fa

…416-on-v6.1.x Backport PR #16416 on branch v6.1.x (BUG: fix yaml table serialization compatibility with numpy 2)

pllim mentioned this pull request May 9, 2024

TST: Add test coverage to PR 16416 (BUG: fix yaml table serialization compatibility with numpy 2) #16429

Closed

neutrinoceros mentioned this pull request May 13, 2024

TST: improve test coverage for io.misc.yaml.AstropyDumper #16444

Merged

1 task

echoix mentioned this pull request Jan 7, 2025

CI(Ubuntu): Install latest version of Python (3.13) for full config OSGeo/grass#4372

Open



		def test_serialize_mixin_column(tmp_path):
		# see https://github.com/astropy/astropy/issues/16414

Conversation

neutrinoceros commented May 8, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Uh oh!

github-actions Bot commented May 8, 2024

Uh oh!

github-actions Bot commented May 8, 2024

Uh oh!

pllim left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

pllim commented May 8, 2024

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!

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 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!

Choose a reason for hiding this comment

Uh oh!

neutrinoceros commented May 9, 2024

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!

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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mhvk left a comment

Choose a reason for hiding this comment

Uh oh!

mhvk commented May 9, 2024

Uh oh!

taldcroft commented May 9, 2024

Uh oh!

neutrinoceros commented May 9, 2024

Uh oh!

pllim commented May 9, 2024

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

neutrinoceros commented May 8, 2024 •

edited

Loading