ENH: additional operator arguments for more explicit control over MultiIndex levels

[MMCMA]

Is your feature request related to a problem?

More explizit operator (add, sub, mul...) index level control when working with pd.MultiIndex on both index and column dimension

Describe the solution you'd like

Additional arguments in operator function to control both index and column level usage.

API breaking implications

Not sure

Describe alternatives you've considered

See code solution below

Additional context

Maybe there is another more elegant solution already in place and I just missed it

    import pandas as pd

    df_1 = pd.DataFrame([[1, 2], [1, 2, ], [1, 2]])
    df_1.index = pd.MultiIndex.from_arrays([[0, 1, 2], [1, 1, 1]],
                                           names=['idx_0', 'idx_1'])
    df_1.columns = pd.MultiIndex.from_arrays([[1, 2], ['a', 'b']])

    df_2 = pd.DataFrame(dict(a=[2, 3, 4], b=[1, 2, 3]))

    # uses per default idx_1 of index to align multiplication, but I need idx_0
    df_1.mul(df_2, axis=0, level=1)

   # Current solution
    df_1.index = df_1.swaplevel()
    df_1.mul(df_2, axis=0, level=1)
    df_1.index = df_1.swaplevel()

[samukweku]

I think your current solution is good : df_1.swaplevel().mul(df_2, axis=0, level=1).swaplevel(); I think the issue stems from the fact that the dataframes have different indices, and as such some alignment must occur; if you try : df_1.mul(df_2.set_index(df_1.index), axis = 0, level = 1), it works out fine. It might be an easy stuff to implement - I feel it might introduce another level of complexity (MultiIndex is already a relatively complex idea), and may not be so intuitive long run. So maybe the onus should be on the user to align the indices (either index or columns) before running the binary operation