`is_*_dtype` functions don't accept extension dtypes

[gsakkis]

The _ArrayOrDtype alias doesn't include ExtensionDtype, which I believe all the is_*_dtype functions accept.

[Dr-Irv]

Thanks for the report. Simple fix is to add ExtensionDtype here:

pandas-stubs/pandas-stubs/core/dtypes/common.pyi

Line 32 in db3b883

ArrayLike, npt.DTypeLike, pd.Series, pd.DataFrame, pd.Index

Should add tests for various types like str, "str", ExtensionDtype, etc. to the tests in https://github.com/pandas-dev/pandas-stubs/blob/main/tests/test_api_types.py

[ramvikrams]

Should add tests for various types like str, "str", ExtensionDtype, etc. to the tests in https://github.com/pandas-dev/pandas-stubs/blob/main/tests/test_api_types.py

I think we had test for ExtensionDtype in this file.

[Dr-Irv]

I think we had test for ExtensionDtype in this file.

Yes, but not for the is_*_dtype functions.

[ramvikrams]

Yes, but not for the is_*_dtype functions.

Ya, i'll add them.

[ramvikrams]

Sir, do you have a short example for ExtensionDtype i am unable to get one

[Dr-Irv]

Sir, do you have a short example for ExtensionDtype i am unable to get one

You just want to make sure that things like is_bool_dtype(ExtensionDtype) work (and similar for the other is_*_dtype() functions.

Also want to add tests for native types and their string representations.

[ramvikrams]

ExtensionDtype is a custom data-type but many is_*_dtype() functions in pytest are saying that they cannot interpret it as data-type

[Dr-Irv]

ExtensionDtype is a custom data-type but many is_*_dtype() functions in pytest are saying that they cannot interpret it as data-type

Write an example that fails and show it, and I can look into it

[ramvikrams]

    check(assert_type(api.is_object_dtype(ExtensionDtype), bool), bool)

this returns

FAILED tests/test_api_types.py::test_is_object_dtype - TypeError: Cannot interpret '<property object at 0x7fc56b581fd0>' as a data...

[Dr-Irv]

I think that's a bug in is_object_dtype() that I'll have to report. So leave out that test for now there.

[ramvikrams]

Actually 11 of the are failing, which are :

is_complex_dtype
is_datetime64_any_dtype
is_datetime64_dtype
is_float_dtype
is_int64_dtype
is_integer_dtype
is_numeric_dtype
is_object_dtype
is_signed_integer_dtype
is_timedelta64_dtype
is_unsigned_integer_dtype

[Dr-Irv]

Thanks. I created pandas-dev/pandas#50923 to report this.

Can you still include them in the tests, but comment them out. E.g.

# check(assert_type(api.is_object_dtype(ExtensionDtype), bool), bool)   pandas GH 50923

Then whenever that gets fixed, we can put back the tests in the future.