BUG: Can't pass arguments to DataFrameGroupBy.agg when using list/dict-like and string aliases

[sartajsartaj]

Feature Type

• Adding new functionality to pandas

• Changing existing functionality in pandas

• Removing existing functionality in pandas

Problem Description

when i apply sum() on groupby object there is a option to select numeric_only columns like :

movies.groupby('Genre').sum(numeric_only = True) # movies here is imdb-top-1000.csv

but if I want to apply multiple aggregate function on ALL numeric_only columns, there is no option to write "numeric_only = True" anywhere according to documentation.
movies.groupby('Genre').agg( ['min', 'max', 'mean'] ) # This code gives error in 2.0.2 version of pandas

This wasn't a problem in earlier version (version 1.5) of pandas as it used to automatically select numerical columns.
Please fix this !
At least provide some option to write numeric_only = True like :
movies.groupby('Genre').agg( ['min', 'max', 'mean'] , numeric_only = True)

Feature Description

At least provide some option to write numeric_only = True like :
movies.groupby('Genre').agg( ['min', 'max', 'mean'] , numeric_only = True)

Alternative Solutions

At least provide some option to write numeric_only = True like :
movies.groupby('Genre').agg( ['min', 'max', 'mean'] , numeric_only = True)

Additional Context

No response

[samukweku]

you can use a lambda function to pass the extra arguments, instead of strings. You can also prefilter, selecting only numeric columns, before running the groupby operation

[rhshadrach]

you can use a lambda function to pass the extra arguments, instead of strings. You can also prefilter, selecting only numeric columns, before running the groupby operation

Agreed on each of these workarounds, though the former will be significantly ill-performant. We allow passing numeric_only on non-list inputs (e.g. gb.agg("sum", numeric_only=True)), and many other groupby-ops, I think we should be doing the same here.

The fix for #45658 would address this (internally we'd be using DataFrameGroupBy.sum instead of SeriesGroupBy.sum), but is currently blocked by the DataFrame case because of a transpose that happens with the results. That transpose behavior does not occur in groupby, and so we should be able to implement the fix proposed for #45658. First we'll need to split off the handling of groupby from that of DataFrame (I was already strongly leaning toward doing this).

cc @topper-123

[topper-123]

I think we should be doing the same here.

Yeah, I agree.