Add type parameters to generics in isna and notna

[JanEricNitschke]

• Closes pd.notna does not work with pyright strict mode. #944
• Tests added: Please use assert_type() to assert the type of any return value

This is to fix issues with pyright strict mode which complains about partially unknown types.

I verified that this test fails in strict mode without these changes and passes with them.

There are still some points i am unsure about:

• Whether to use Any or an (unbounded?) TypeVar (or maybe i should use Incomplete?)
• I had to split the ArrayLike annotation into its components to add the type parameters to the ndarray. While i did also change it in the pandas-stubs/_typing.pyi, it was still importing them from the actual pandas.

[JanEricNitschke]

Closes #944

[JanEricNitschke]

I checked that the Any is probably the way to go unless we already know some bounds that need to be respected here.

Only open question would then be the handling of the ArrayLike. The best approach would be if we could modify (which i did) and pull in the version from the stubs. But i dont know how, do i explicitely import from pandas-stubs? But it seems to work with importing from pandas in the previous pr.

[Dr-Irv]

Only open question would then be the handling of the ArrayLike. The best approach would be if we could modify (which i did) and pull in the version from the stubs. But i dont know how, do i explicitely import from pandas-stubs? But it seems to work with importing from pandas in the previous pr.

In the stubs, you can do from pandas._typing import ArrayLike . The stubs do imports from themselves instead of from the pandas source.

[JanEricNitschke]

Just double checked and you are right. I was getting some weird messages from vscode there. I guess it was an issue with that.

I ran pyright over just that one test and it did pick up the ArrayLike from the stubs _typing.

[Dr-Irv]

thanks @JanEricNitschke