Fix routing.get_name() not to assume all routines have __name__

[mgorny]

Summary

Fix routing.get_name() to use the __name__ attribute only if it is actually present, rather than assuming that all routine and class types have it, and use the fallback to class name otherwise. This is necessary for functools.partial() that doesn't have a __name__ attribute, but is recognized as a routine starting with Python 3.13.0b3. Since for older versions of Python, it would have used the fallback anyway, this doesn't really change the behavior there.

Fixes #2638

Checklist

• I understand that this PR may be closed in case there was no previous discussion. (This doesn't apply to typos!)
• I've added a test for each change that was introduced, and I tried as much as possible to make a single atomic change. ← this is covered by existing tests already
• I've updated the documentation accordingly. ← I don't think any changes are necessary

[Kludex]

Can we link the CPython PR here?

I don't understand why change this behavior.

[Kludex]

Oh well... Seems okay anyway. If it has a __name__, use it, otherwise use __class__.__name__.

We should add some tests anyway.

[mgorny]

Can we link the CPython PR here?

Do you mean the commit that changed partial() behavior?

[mgorny]

Can we link the CPython PR here?

Do you mean the commit that changed partial() behavior?

The change in question is python/cpython@49e5740. Actually, I'm not sure if inspect.isroutine() change was intentional or an accidental side-effect, I'll file a bug there too, just in case.

We should add some tests anyway.

What kind of tests do you have in mind?

[Kludex]

There's actually a test called test_route_name that tests this code.

All good here. Thanks @mgorny :)

[Kludex]

But... To be honest... If it was unintended on CPython's side, I'd prefer to wait for them to revert the unintended changes.

[musicinmybrain]

Thanks for this PR. It works nicely as a downstream patch in Fedora Rawhide, which allows me to keep building and updating Starlette. I’ll monitor python/cpython#122087 and drop the downstream patch if this ends up being handled in CPython 3.13.0rc1.

Looks like CPython is going to fix the issue.

[mgorny]

@Kludex, upstream confirmed that:

Yes, not all method descriptors (and therefore not all routines) have the __name__ attribute. This always was so, you always should have a fallback.

So I think we should merge this anyway.

[Kludex]

@Kludex, upstream confirmed that:

Yes, not all method descriptors (and therefore not all routines) have the __name__ attribute. This always was so, you always should have a fallback.

So I think we should merge this anyway.

Well... But they are fixing it for functools.partial, right?

[mgorny]

Well... But they are fixing it for functools.partial, right?

Yes, they're adding a workaround for Python 3.13. However, it will fail again in 3.14.

[Kludex]

Well... But they are fixing it for functools.partial, right?

Yes, they're adding a workaround for Python 3.13. However, it will fail again in 3.14.

Well... Then let's go. 😅

Thanks @mgorny 🙏

[mgorny]

Thanks a lot!