BUG: groupby.<reduction>(numeric_only=) does not validate non-bool arguments

[mroeschke]

In [1]: import pandas as pd

In [2]: df = pd.DataFrame({"A": range(5), "B": range(5)})

In [3]: df.groupby(["A"]).mean(["B"])
     B
A     
0  0.0
1  1.0
2  2.0
3  3.0
4  4.0

["B"] is being passed to numeric_only which appears checked only to be truthy/falsy.

I would expect this to raise a ValueError

[Zrahay]

@mroeschke Is it fine if I work on this?

[Zrahay]

I am able to makeout the core logic for this is present in generic.py file but inside of it, how are the functional callings happening are a bit blackbox for, so, can you help me out there?

Here's what I've understood ->

When we do groupby('column'), we are ending up with DataFrameGroupBy.aggregate() inside generic.py and we are pretty much calling def aggregate function.

But after this, I am not able to comprehend what is happening when we are doing .mean() on this DataFrameGroupBy object.

I hope my question made sense, if not let me know, I'll try explaining again.

Thanks!

[Zrahay]

Hi @mroeschke ! So, here's my solution to this -

I am thinking of putting a check inside the _cython_agg_general() function which is present in the GroupBy class to verify if the numeric_only parameter is a Boolean or not and raise a ValuerError if not so. We can perform this simple check by using the isinstance function of Python. This is what should the logic be according to me.

If you think my suggestion is wrong or maybe there's a better way to implement this, please do let me know.