gh-148468: Accept string-like help proxies in argparse

[RaghunandanKumar]

Summary

• restore support for string-like proxy objects in colorized argparse descriptions, epilogs, and argument help text
• normalize formatter text at the string-formatting boundary instead of assuming concrete str objects
• add regression coverage for proxied descriptions, epilogs, and interpolated help text

Issue

• Cannot use lazy translation proxies in argparse help #148468

Testing

• ./python.exe -m test test_argparse
• runtime repro with a minimal LazyStr proxy and parser.format_help()

[savannahostrowski]

Thanks for the PR! As mentioned in my comment on the issue, the str() coercion can go at the re.sub call site rather than at the top of _expand_help. Also, please scope the change down. The _format_action changes and the broader test expansion aren't necessary for this fix.

[bedevere-app]

A Python core developer has requested some changes be made to your pull request before we can consider merging it. If you could please address their requests along with any other requests in other reviews from core developers that would be appreciated.

Once you have made the requested changes, please leave a comment on this pull request containing the phrase I have made the requested changes; please review again. I will then notify any core developers who have left a review that you're ready for them to take another look at this pull request.

[RaghunandanKumar]

Addressed review feedback:

• moved str() coercion to re.sub() call site in _expand_help()
• removed _format_action changes
• trimmed regression coverage down to focused interpolated help proxy test

Testing:

• ./python.exe -m test test_argparse

I have made the requested changes; please review again

[bedevere-app]

Thanks for making the requested changes!

@savannahostrowski: please review the changes made to this pull request.

[RaghunandanKumar]

Inspected the Win32 failure. test_argparse passed there; the red job is unrelated: test_profiling.test_sampling_profiler.test_generator_not_under_consumer_arithmetic in Windows / Build and test (Win32, switch-case).

This matches gh-150429, with follow-up fix in #150433. I pushed an empty commit to retrigger CI on the current PR head.

[savannahostrowski]

Unfortunately, this will still break with a proxy help string without a format specifier. The interpolation branch and the no-% branch in _expand_help reach re.sub by different routes, and only one of them is handled here. Let's add a test for this so we can catch it too.

Additionally, the summary and NEWS also overstate the scope of the change. This isn't fixing descriptions and epilogs and we don't need to since they weren't part of this regression (#142960 only changed the argument-help path). Please scope the wording down to what's actually fixed.

Please don't push empty commits to retrigger CI. It just creates additional noise. A comment stating that you don't think the failure is related is just fine.

[bedevere-app]

A Python core developer has requested some changes be made to your pull request before we can consider merging it. If you could please address their requests along with any other requests in other reviews from core developers that would be appreciated.

Once you have made the requested changes, please leave a comment on this pull request containing the phrase I have made the requested changes; please review again. I will then notify any core developers who have left a review that you're ready for them to take another look at this pull request.

[RaghunandanKumar]

Addressed latest review feedback:

• fixed string-like proxy handling for argument help without a % format specifier
• kept scope limited to argument help only
• narrowed NEWS wording to match actual fix scope

Testing:

• ./python.exe -m test -v test_argparse -m test_argument_help_without_interpolation_accepts_string_like_proxy
• ./python.exe -m test -v test_argparse -m test_argument_help_interpolation_accepts_string_like_proxy
• ./python.exe -m test test_argparse

I have made the requested changes; please review again

[bedevere-app]

Thanks for making the requested changes!

@savannahostrowski: please review the changes made to this pull request.

[savannahostrowski]

Alright, something about this issue had been nagging at me this afternoon, so I went back through the commit history again. I didn't realize this before but there's some asymmetry here between proxies with and without format specifiers. While proxies with a format specifier didn't raise before #142960, proxies without did, even with color off. As such, I think we should keep this PR minimally scoped to just resolving the immediate regression with proxies with format specifiers, and leave the discussion re. consistency for another issue/PR.

With that in mind, your commit dc8b7d7 was actually correct. The whole fix should just be the single str() coercion on the re.sub subject in _expand_help.

However, for next steps, I'd like to slow down a bit here. Instead of another revision, it would be helpful to hear your thoughts on the fix first. I would like to make sure that the changes landed are ones that work but are also ones you understand. We have a policy for this sort of thing in our devguide; please give our AI policy a read.