将2D值数组分配给Pandas多索引数据框

Question

I have outputs from calculations that is best stored in a Pandas MultiIndex format. For concrete purposes, let us consider the form below (though the actual structure is dictated programmatically)

                        X         Y         Z       
DATE                                                      
2018-01-01 A           NaN       NaN       NaN      
           B           NaN       NaN       NaN      
           C           NaN       NaN       NaN      
2018-01-02 A           NaN       NaN       NaN       
           B           NaN       NaN       NaN      
           C           NaN       NaN       NaN       

I want to assign the numpy array outputs to a particular time slice. Say I have

output = np.array([[1,2,3],[2,2,1],[4,2,3]])

so the desired output is

                        X         Y         Z       
DATE                                                      
2018-01-01 A           NaN       NaN       NaN      
           B           NaN       NaN       NaN      
           C           NaN       NaN       NaN      
2018-01-02 A             1         2         3       
           B             2         2         1   
           C             4         2         3   

I have tried pandas.IndexSlice where j is the j-th time slice.

df.loc[pd.IndexSlice[j,:], :] = output

but that doesn't work. I have also tried by replacing loc by iloc but to no avail. In non-MultiIndex dataframes, I can assign a list to a particular column in a DataFrame without having to assign each element individually. Is there a way to do it for a matrix into a MultiIndex dataframe?

Answer 1

your code works just fine.

Demo:

In [70]: df.loc[pd.IndexSlice['2018-01-02', :], :] = output

In [71]: df
Out[71]:
                 X    Y    Z
DATE       I2
2018-01-01 A   NaN  NaN  NaN
           B   NaN  NaN  NaN
           C   NaN  NaN  NaN
2018-01-02 A   1.0  2.0  3.0
           B   2.0  2.0  1.0
           C   4.0  2.0  3.0

PS i tested both options when the DATE index column is of string and when it's is of datetime dtype - in both cases the code above is working properly.