allow args and kwargs in groupby.apply

[Dr-Irv]

• Closes GroupBy.apply: error: No overloads for "apply" match the provided arguments (reportCallIssue) #1266(
• Tests added: test_groupby.py:test_dataframe_apply_kwargs()

[schorlton-bugseq]

Thank you!

[twoertwein]

Two small optional changes (happy to merge without them):

• include_groups is not deprecated, only include_groups=True, I would be fine keeping the test and adding include_groups: Literal[False] in the signature
• it might be nice to have a test where a kwargs/args is incompatible with the func

[Dr-Irv]

Two small optional changes (happy to merge without them):

• include_groups is not deprecated, only include_groups=True, I would be fine keeping the test and adding include_groups: Literal[False] in the signature

Yes, but there are 2 issues with that.

1. The argument will become irrelevant since only a value of False is accepted. So we shouldn't encourage its use.
2. If we use ParamSpec, we can't have a keyword argument between *args and **kwargs in the typing stubs. See Is `typing.Protocol` the only way to concatenate the keyword-only arguments in `ParamSpec`? python/typing#1905

• it might be nice to have a test where a kwargs/args is incompatible with the func

Yes, although I'm not sure how to write one! I just wanted to handle the use case reported.

[twoertwein]

1. The argument will become irrelevant since only a value of False is accepted. So we shouldn't encourage its use.
2. If we use ParamSpec, we can't have a keyword argument between *args and **kwargs in the typing stubs. See Is typing.Protocol the only way to concatenate the keyword-only arguments in ParamSpec? python/typing#1905

Thank you for the explanations!

Yes, although I'm not sure how to write one! I just wanted to handle the use case reported.

Something like

df.groupby("group", group_keys=False)[["group", "value"]].apply( add_constant_to_mean, constant="5" # typing: ignore )

should hopefully be flagged as wrong.

[Dr-Irv]

Something like

df.groupby("group", group_keys=False)[["group", "value"]].apply( add_constant_to_mean, constant="5" # typing: ignore )

should hopefully be flagged as wrong.

Thanks for the suggestion. I've incorporated into the last commit

[twoertwein]

Thanks @Dr-Irv !

[DeflateAwning]

This PR is the bane of my existence. Can you please add back the overload for include_groups: Literal[False]

[Dr-Irv]

This PR is the bane of my existence. Can you please add back the overload for include_groups: Literal[False]

No, you should modify your code to avoid using include_groups