pd.isna / pd.isnull raising mypy error for None

Describe the bug pd.isna / pd.isnull raising mypy error for None

To Reproduce

file.py

import pandas as pd
pd.isnull(None)

~  pdnull ❯ mypy file.py                                                                       9s 15:54:43
file.py:2: error: No overload variant matches argument type "None"
file.py:2: note: Possible overload variants:
file.py:2: note:     def isna(obj: DataFrame) -> DataFrame
file.py:2: note:     def isna(obj: Series[Any]) -> Series[bool]
file.py:2: note:     def isna(obj: Union[Index, List[Any], ExtensionArray, ndarray[Any, Any]]) -> ndarray[Any, dtype[bool_]]
file.py:2: note:     def isna(obj: Union[str, bytes, date, datetime, timedelta, bool, complex, Timestamp, Timedelta]) -> bool
file.py:2: note:     def isna(obj: Union[NaTType, NAType]) -> Literal[True]
Found 1 error in 1 file (checked 1 source file)

Please complete the following information:

• MacOS
• mypy-0.971
• pandas: 1.5.0
• python: 3.9
• pandas-stubs-1.4.4.220919

Expected behaviour: No issue since pd.isnull(None) == True and works fine

[twoertwein]

The inline annotations of isna cover any object (in the fallback overload) but None is not special-cased in the first overload to return bool: https://github.com/pandas-dev/pandas/blob/ce143a2fb0a41e78c85d5311d0e159552d651b00/pandas/core/dtypes/missing.py#L81

While we cannot enumerate all possible inputs for isna (would probably create overlapping overloads), I’m +1 on adding None.

[bashtage]

But the documentation does not actually say it can be used on any type. If we follow the documentation I think a lot of people will get type errors on perfectly legit code.

The deep question is whether it is legit code? Are there examples of usage like this in the docs, even if not in the func docs?

Actually function doc says works on any scalar object, which is pretty wide.