Add wurlitzer stubs

[hoel-bagard]

This PR adds stubs for the wurlitzer package.

[AlexWaygood]

Mostly looks good, though I haven't looked in-depth at the stderr/stdout annotations. Usually we prefer to defined protocols for that kind of thing rather than mandating that an object has to be an instance of io.IOBase. Usually that isn't necessary and duck-typing will do.

[hoel-bagard]

Thanks for the (very fast 😮 ) review, I'll look into using protocols.

Edit: looks like some overloads will be required to get precise return types, I'm passing this PR back into draft mode for now.

[hoel-bagard]

@AlexWaygood Do you know what in the overloads is causing the CI to fail ?

[Akuli]

The error looks like a bug in stubtest, but you can probably work around it by not using overloaded context managers. Instead of @contextmanager together with -> Iterator[Foo], you could try -> AbstractContextManager[Foo] or -> _GeneratorContextManager[Foo]. You need to import from contextlib.

[AlexWaygood]

The error looks like a bug in stubtest, but you can probably work around it by not using overloaded context managers. Instead of @contextmanager together with -> Iterator[Foo], you could try -> AbstractContextManager[Foo] or -> _GeneratorContextManager[Foo]. You need to import from contextlib.

Yeah, @Akuli is right. Stubtest in general isn't great at dealing with several decorators on a single function -- see python/mypy#14950 (which is now fixed, but was only fixed by special-casing @final).

[hoel-bagard]

@AlexWaygood If you have time to review the changes.

[Akuli]

Thanks for working on this! Typing is hard :)

[hoel-bagard]

@Akuli Thanks for the improvements/fixes!

[github-actions]

According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉

[Akuli]

Thanks!