BUG: DataFrameGroupBy.nunique fails with categorical, multiple groupings, and as_index=False

[rhshadrach]

df = DataFrame({"a1": [0, 0, 1], "a2": [2, 3, 3], "b": [4, 5, 6]})
df = df.astype({"a1": "category", "a2": "category"})
gb = df.groupby(by=["a1", "a2"], as_index=False, observed=False)
gb.nunique()

raises ValueError: Length of values (3) does not match length of index (4)

[szczekulskij]

This looks a bit hard, but I'll give it a try. If I won't be able to solve this within a week, I'll re-assign

[szczekulskij]

take

[szczekulskij]

Commenting out lines 1911-1913 solves issue. File: pandas/core/groupby/generic.py

        # if not self.as_index:
        #     res_df.index = default_index(len(res_df))
        #     res_df = self._insert_inaxis_grouper(res_df)

Gonna work on fixing it now.
Initial solution idea w. temp as_index didn't work : /

[szczekulskij]

Weirdly, It's the same function causing the issue as in #52397 that I'm working on

The name of function: self._insert_inaxis_grouper - in both issues we need to skip call to this function(or fix the way function works)

[szczekulskij]

Is it okay if I force as_index = True using with com.temp_setattr(self, "as_index", True): ?

With as_index=True forced, the output would always look like this:

#        b
# a1 a2   
# 0  2   1
#    3   1
# 1  2   0
#    3   1

I assume we want the behaviour to differ depending on as_index, but wanted to double check

@rhshadrach

[rhshadrach]

I assume we want the behaviour to differ depending on as_index, but wanted to double check

Correct. Running the OP example without categorical data produces:

   a1  a2  b
0   0   2  1
1   0   3  1
2   1   3  1

[szczekulskij]

Hey, I'm sorry but I'll need to drop this issue. Pandas is a bit too much for me right now, I'll come back to this in the future