I have a data frame containing value_1 and value_2

df_1 = pd.DataFrame(
    {
        "id_1": [101, 202],
        "id_2": [101, 202],
        "value_1": [5.0, 10.0],
        "value_2": [10.0, 4.0],
    }
)
df_1 = df_1.set_index(["id_1", "id_2"])

that looks like this:

           value_1  value_2
id_1 id_2
101  101       5.0     10.0
202  202      10.0      4.0

I have another data frame, that contains a flag for each value, i.e. is_active_1 and is_active_2:

df_2 = pd.DataFrame(
    {
        "id_1": [101, 202],
        "id_2": [101, 202],
        "is_active_1": [True, False],
        "is_active_2": [False, False],
    }
)
df_2 = df_2.set_index(["id_1", "id_2"])

that looks like this:

           is_active_1  is_active_2
id_1 id_2
101  101          True        False
202  202         False        False

I want to multiply the value rows by *3 in df_1 depending on its flag in df_2. The end result should like this:

           value_1  value_2
id_1 id_2
101  101      15.0     10.0
202  202      10.0      4.0

i.e. the is_active_1 = True flag for (id_1, id_2) = (101, 101) causes value_1 -> 3 * 5.0 = 15.0

I have tried the following:

df_1.loc[df_2[["is_active_1", "is_active_2"]], ["value_1", "value_2"]] * 3

but ended up with a value error ValueError: Cannot index with multidimensional key.

>Solution :

By decreasing level of index alignment.

You can rename the columns to replace is_active by name:

df_1[df_2.rename(columns=lambda x: x.replace('is_active', 'value'))] *= 3

Or, you can use set_axis to avoid index alignment on the columns:

df_1[df_2.set_axis(df_1.columns, axis=1)] *= 3

Or assume the two dataframes are aligned and ignore completely the labels of df_2:

df_1[df_2.to_numpy()] *= 3

Updated df_1:

           value_1  value_2
id_1 id_2                  
101  101      15.0     10.0
202  202      10.0      4.0