DOC: Dataframe.skew invalid function signature for "skipna"

[bf]

Location of the documentation

https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.skew.html

Documentation problem

In the function signature it says skipna=None even though the parameter documentation underneath says skipna bool, default True.

Suggested fix for documentation

The skipna=None in function signature should be changed to skipna=True.

[MarcoGorelli]

Thanks @bf - are you interested in opening a pull request?

[bf]

@MarcoGorelli I'd love to create a PR, but I have trouble finding the proper source file.

Could you help me find the root cause?

I have found

pandas/pandas/core/generic.py

Line 10197 in dec736f

cls.skew = _make_stat_function(

which seems to indicate that the definition of _make_stat_function() in

pandas/pandas/core/generic.py

Line 11139 in dec736f

def _make_stat_function(

erroneously sets skipna=None at

pandas/pandas/core/generic.py

Line 11161 in dec736f

self, axis=None, skipna=None, level=None, numeric_only=None, **kwargs

But as _make_stat_function() is used all over the place in that file, we'd need to check every single function if the default skipna=True is really correct.

Do you agree with this process?

[MarcoGorelli]

Thanks for taking a look

Have just had a look at the source code for _make_stat_function, and it says

        if skipna is None:
            skipna = True

TBH I don't understand why the default value is set to None, and then it's changed to True in every case.

@simonjayhawkins would it be OK to change stat_func's signature to have

skipna=True

and then remove the lines

        if skipna is None:
            skipna = True

so that in the documentation, the default value of True (from _num_doc) would match the default shown in the signature? I'm sorry if there's something obvious I'm missing here

[simonjayhawkins]

@bf Thanks for the report.

the latest docs are at https://pandas.pydata.org/docs/dev/reference/api/pandas.DataFrame.skew.html

The type annotations in the function signature are now removed to avoid this type of conflict with the paramater descriptions, see #33312

[bf]

You are right, the issue exists also for other doc pages which use _make_stat_function, e.g. https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.std.html

I imagine the if skipna is None: skipna = True is most likely done to enable some static type checking for the underlying functions, or to remove some weird testing error, as it makes sure all functions only receive a proper boolean parameter.

[bf]

@simonjayhawkins seems like we posted at the same time. Thanks for the fix and for your great work.

[simonjayhawkins]

I'll reopen, since even without the type annotations, the default is still shown as None

[MarcoGorelli]

@simonjayhawkins isn't this still present even without the type annotations?

EDIT

sorry, commenting almost at the same again :)