[fnmatch] calls os.path.normcase(arg)

[hunterhogan]

fnmatch.fnmatch, fnmatch.filter, and fnmatch.filterfalse each pass their parameters, name and pat, to os.path.casenorm before operating on the parameters.

typeshed/stdlib/posixpath.pyi

Lines 125 to 134 in cc36aca

	if sys.version_info >= (3, 15):
	@overload
	def normcase(s: PathLike[AnyStr], /) -> AnyStr: ...
	@overload
	def normcase(s: AnyOrLiteralStr, /) -> AnyOrLiteralStr: ...
	else:
	@overload
	def normcase(s: PathLike[AnyStr]) -> AnyStr: ...
	@overload
	def normcase(s: AnyOrLiteralStr) -> AnyOrLiteralStr: ...

Therefore, the three fnmatch functions' parameters are AnyOrLiteralStr | PathLike[AnyStr] instead of AnyStr.

If any parameter or parameter's element in filter or filterfalse is PathLike[AnyStr], then the return is limited to list[AnyStr], otherwise it is the superset list[AnyOrLiteralStr].

[hunterhogan]

Diff from mypy_primer, showing the effect of this PR on open source code:

setuptools (https://github.com/pypa/setuptools)
+ setuptools/glob.py:106: error: Unused "type: ignore[return-value]" comment  [unused-ignore]

Link to code above.

I feel like this suggests that my proposed type annotations are a step in the right direction because they seem to be more aligned with fnmatch usage "in the wild." Is that a valid interpretation, or am I reading too much into this single hit out of eleventy jillion lines of code?

[github-actions]

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

[srittau]

Thanks!

[hunterhogan]

Thank you for your guidance and patience!