Most of the hits here are from python/typeshed#6670. That typeshed commit solves a lot of issues for pyright, but it seems like it currently only causes false positives for mypy. Until mypy improves its support for ParamSpec, I'd be in favour of adding it to the typeshed changes we manually revert in our typeshed syncs. Thoughts on that?

Yeah, I'd be in favour of reverting it as well.

• The pip hit is from Fix stdlib/disutils testing typeshed#9734
• Both other hits are from Add get_origin annotations typeshed#9811

4 of the errors from the typeshed @wraps PR seem to just be due to the fact that _Wrapped doesn't define __name__:

• the pandas error
• the 2 spark errors
• the streamlit error

Looking at the implementation of update_wrapper (which is the function that returns _Wrapped), I see that by default it will set ('__module__', '__name__', '__qualname__', '__doc__', '__annotations__') on the object it returns, which maybe implies that _Wrapped should have these attributes, though this can be customized with a parameter to update_wrapper...

4 of the errors from the typeshed @wraps PR seem to just be due to the fact that _Wrapped doesn't define __name__:

• the pandas error
• the 2 spark errors
• the streamlit error

Looking at the implementation of update_wrapper (which is the function that returns _Wrapped), I see that by default it will set ('__module__', '__name__', '__qualname__', '__doc__', '__annotations__') on the object it returns, which maybe implies that _Wrapped should have these attributes, though this can be customized with a parameter to update_wrapper...

Want to try making a typeshed PR?

Sure.

You can obviously cook up theoretical situations where adding __name__ to _Wrapped would cause false negatives:

>>> class Foo:
...     def __call__(self): ...
...
>>> class Bar:
...     def __call__(self): ...
...
>>> from functools import update_wrapper
>>> wrapped = Foo()
>>> wrapper = Bar()
>>> update_wrapper(wrapper, wrapped)
<__main__.Bar object at 0x000002CB493C9C50>
>>> _.__name__
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'Bar' object has no attribute '__name__'. Did you mean: '__ne__'?

But these seem pretty rare, and not worth worrying about. Adding __name__ (and friends) to _Wrapped sounds like a promising idea to me.

FWIW though, even if we make that change, I think I'd still want the changes to functools.wraps reverted in mypy's vendored version of typeshed. There's still too many other false positives imo.

Thanks, this looks good but we should probably resync with python/typeshed#9835 since that'll be another merge conflict

python/typeshed#9815 will also cause merge conflicts for the next sync — any interest in taking a quick look at that one before I resync?

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

pip (https://github.com/pypa/pip)
+ src/pip/_internal/locations/_distutils.py:60: error: Unused "type: ignore" comment

hydra-zen (https://github.com/mit-ll-responsible-ai/hydra-zen)
+ src/hydra_zen/structured_configs/_utils.py:69: error: Name "get_origin" already defined (possibly by an import)  [no-redef]
- src/hydra_zen/structured_configs/_utils.py:69: error: All conditional function variants must have identical signatures  [misc]
- src/hydra_zen/structured_configs/_utils.py:69: note: Original:
- src/hydra_zen/structured_configs/_utils.py:69: note:     def get_origin(tp: Any) -> Optional[Any]
- src/hydra_zen/structured_configs/_utils.py:69: note: Redefinition:
- src/hydra_zen/structured_configs/_utils.py:69: note:     def get_origin(tp: Any) -> Optional[type]

pydantic (https://github.com/samuelcolvin/pydantic)
+ pydantic/_internal/_generate_schema.py:377: error: Unused "type: ignore" comment

Thanks!