How to sort a dict in numpy

I want to write a function func() like this:

import numpy as np

a = np.array([0, 1, 1, 0, 0, 2, 3, 0, 1])
b = np.array([1, 1, 4, 5, 1, 4, 1, 9, 1])

ab = np.array([[0, 1, 1, 0, 0, 2, 3, 0, 1],
               [1, 1, 4, 5, 1, 4, 1, 9, 1]])

grp = np.where(a == 0)[0]  # [0, 3, 4, 7]

def func()：
    # return the index of np.max(b[grp]) in a, here it is 7

np.argmax(b[grp]) can only return return the index of np.argmax(b[grp]) in a[grp], here it is 3

I know that dict can do this but numpy seems not supporting dict sorting and spliting like b[grp]

Any help is welcome, thanks!

>Solution :

You can use the argmax of filtered b to index the grp

$ grp[np.argmax(b[grp])]

7