Closed
Description
-
I have checked that this issue has not already been reported.
-
I have confirmed this bug exists on the latest version of pandas.
-
I have confirmed this bug exists on the master branch of pandas.
Reproducible Example
df = pd.DataFrame({'a': [(1,), (1,), (1,)], 'b': [4, 5, 6]})
gb = df.groupby(['a'])
print(gb.rolling(2).corr(other=df))
df = pd.DataFrame({'a': [(1,2,), (1,2,), (1,2,)], 'b': [4, 5, 6]})
gb = df.groupby(['a'])
print(gb.rolling(2).corr(other=df))Issue Description
The first block produces
a b
a
1 0 NaN NaN
1 NaN 1.0
2 NaN 1.0
where the first index level should be (1, ) instead. The second block raises a ValueError. It is due to usage of maybe_make_list in _apply_pairwise (related: #44056). Instead of using this function, we can pack the index keys in lists when self._grouper.nkeys is 1 and otherwise leave them as-is. I.e.
if self._grouper.nkeys == 1:
gb_pairs = (
[pair] for pair in self._grouper.indices.keys()
)
else:
gb_pairs = (
pair for pair in self._grouper.indices.keys()
)
Expected Behavior
a b
a
(1, ) 0 NaN NaN
1 NaN 1.0
2 NaN 1.0
a b
a
(1, 2) 0 NaN NaN
1 NaN 1.0
2 NaN 1.0
Installed Versions
Replace this line with the output of pd.show_versions()