groupby with as_index=False, inconsistency between max and idxmax · Issue #8582 · pandas-dev/pandas (original) (raw)
>>> df = pd.DataFrame(np.random.randint(0, 100, (50, 2)), columns=['jim', 'joe'])
>>> ts = pd.Series(np.random.randint(0, 3, 50))
>>> df.groupby(ts, as_index=False).idxmax()
jim joe
0 25 36
1 22 5
2 3 4
>>> df.groupby(ts, as_index=False).max()
NaN jim joe
0 0 93 94
1 1 94 89
2 2 85 93
max
adds the grouper to the result frame, whereas idxmax
doesn't.
The API documentation says:
as_index=False is effectively “SQL-style” grouped output
so i would assume, it should not inject a new column with NaN
name.