ValueError: Must produce aggregated value, if DataFrame contains a DataFrame as an element

[Jogala]

This minimal example assigns the column B the first value of each group:

df = pd.DataFrame(data=np.asarray([[1,1,1,2],[4,4,4,10]]).transpose(),columns = ['A','B'])
print(df)
dfagg = df.groupby('A').agg({'B' : lambda x : x.iloc[0]})
print(dfagg)

Hence

   A   B
0  1   4
1  1   4
2  1   4
3  2  10

Becomes:

A    
1   4
2  10

If I now modify the value of column B to an array of DataFrames

df1 = pd.DataFrame({'x' : [0,1], 'y' : [-1,-2]})
df2 = pd.DataFrame({'x' : [0,3], 'y' : [11,12]})

df = pd.DataFrame(data=np.asarray([[1,1,1,2],[df1,df1,df1,df2]],dtype=object).transpose(),columns = ['A','B'])
print(df)
dfagg = df.groupby('A').agg({'B' : lambda x : x.iloc[0]})
print(dfagg)

I get the error:

ValueError: Must produce aggregated value

It seems like this is a bug. There is no logic behind this behavior.
I also think that it did not raise an error in pandas like 2 years ago, when I used the code snippet the last time.
Currently using the newest version 1.2.1

[rhshadrach]

Thanks for the report. Am I correct in saying that your expected output is a Series where the values are themselves DataFrames?