Remove obsolete mypy bug workaround in abspath()

[srittau]

No description provided.

[AlexWaygood]

Can we also get rid of the similar overloads in re.pyi? Or is that a different issue?

typeshed/stdlib/re.pyi

Lines 236 to 246 in c709686

	# Type-wise the compile() overloads are unnecessary, they could also be modeled using
	# unions in the parameter types. However mypy has a bug regarding TypeVar
	# constraints (https://github.com/python/mypy/issues/11880),
	# which limits us here because AnyStr is a constrained TypeVar.
	# pattern arguments do *not* accept arbitrary buffers such as bytearray,
	# because the pattern must be hashable.
	@overload
	def compile(pattern: AnyStr, flags: _FlagsType = 0) -> Pattern[AnyStr]: ...
	@overload
	def compile(pattern: Pattern[AnyStr], flags: _FlagsType = 0) -> Pattern[AnyStr]: ...

[AlexWaygood]

Okay it looks like weird things happen if you try to pass an object typed as Pattern[Any] for the re.pyi overloads (https://github.com/unionai-oss/pandera/blob/7420fc5de7b8b1b8bac31139ea95577ebf0a427c/pandera/api/checks.py#L432-L439), so maybe we should leave those for another PR and consider them separately. Sorry for the back-and-forth :(

[AlexWaygood]

LGTM. From the mypy_primer result when we tried to do similarly with the re.pyi overloads, I guess you might run into false positives if you tried to pass an object typed as PathLike[Any] into one of these functions. But I'd guess that's pretty rare. By far the most common PathLike object around is pathlib.Path, which is PathLike[str] rather than PathLike[Any].

[github-actions]

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