BUG: __get_item__ accepts more types than it's type signature says

[anilbey]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

from enum import Enum
import pandas as pd

class MyEnum(Enum):
    TAYYAR = "tayyar"
    HAYDAR = "haydar"

df = pd.DataFrame(data = [[12.2, 10], [8.8, 15], [22.1, 14]], columns=[MyEnum.TAYYAR, MyEnum.HAYDAR])
print(df[MyEnum.TAYYAR])

Issue Description

The example code above takes an Enum as column to a DataFrame and uses that enum in __get_item__.

It works fine, which creates an inconsistency with the signature of __get_item__ method.

Here is the signature of __get_item__. According to this signature, enums are not accepted as index.

def __getitem__(self, Union[str, bytes, date, datetime, timedelta, bool, int, float, complex, Timestamp, Timedelta], /) -> Series[Any]

When you use a static type checker e.g. mypy on the code example above, it fails.

mypy_enum.py:12: error: No overload variant of "__getitem__" of "DataFrame" matches argument type "MyEnum"  [call-overload]
mypy_enum.py:12: note: Possible overload variants:
mypy_enum.py:12: note:     def __getitem__(self, Union[str, bytes, date, datetime, timedelta, bool, int, float, complex, Timestamp, Timedelta], /) -> Series[Any]
mypy_enum.py:12: note:     def __getitem__(self, slice, /) -> DataFrame
mypy_enum.py:12: note:     def [ScalarT] __getitem__(self, Union[Tuple[Any, ...], Series[bool], DataFrame, List[str], List[ScalarT], Index, ndarray[Any, dtype[str_]], ndarray[Any, dtype[bool_]], Sequence[Tuple[Union[str, bytes, date, datetime, timedelta, bool, int, float, complex, Timestamp, Timedelta], ...]]], /) -> DataFrame
Found 1 error in 1 file (checked 1 source file)

More info: https://stackoverflow.com/questions/75614405/mypy-indexing-pd-dataframe-with-an-enum-raises-no-overload-variant-error

Expected Behavior

__getitem__'s signature should be updated to support the column type itself. The column type is often not limited to primitive types. In can be custom objects as well.

Installed Versions

assert '_distutils' in core.__file__, core.__file__

[twoertwein]

Moved to pandas-stubs as pandas's DataFrame.__getitem__ has actually no annotations.

[anilbey]

Does the annotation come from pandas-stubs?

[twoertwein]

Does the annotation come from pandas-stubs?

I assume so. It definitely doesn't come from pandas itself (unless you created an empty file called py.typed inside the pandas folder).

[anilbey]

I see. No I definitely haven't created that file :)

[Dr-Irv]

I think if we add Hashable to the __getitem__() declaration here, that should fix this.

pandas-stubs/pandas-stubs/core/frame.pyi

Line 519 in fbc5f20

def __getitem__(self, idx: Scalar | tuple[Hashable, ...]) -> Series: ...

PR and tests welcome

[anilbey]

Thanks @Dr-Irv I just make a PR with a test.