BUG: Bug in consistency of groupby aggregation when passing a custom function (GH6715)

[jreback]

closes #6715

[jreback]

@jorisvandenbossche works but is a big hacky

[jorisvandenbossche]

Thanks! Will test it this evening.

[jorisvandenbossche]

For g.agg(P1) it does work, but g.apply(P1) not yet (or this way maybe not meant to fix that?). It just seems intuitive if you can do df.apply(P1, that you also can do g.apply(P1)

[jreback]

g.apply(P1) is correct as it is passed as a full dataframe (and not series by series). I don't see a way to avoid doing that. The paths groupby takes are pretty complicated. So not going to be able to fix. If you want to create a separeate issue maybe can address in 0.15.