BUG: GroupBy.idxmax/idxmin with EA dtypes by jbrockmendel · Pull Request #38733 · pandas-dev/pandas (original) (raw)

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

those should not return -1, IMO. I would prefer to not put this "wrong" behaviour in the EAs (but keep it in Series were it already has this behaviour), as long as we didn't decide on the preferred behaviour.

Short term, could eg raise an error saying that skipna=False is not yet implemented, but by accepting skipna, it can already fix the default groupby case (see my comment at #38733 (comment))