[Python] Cumulative sum of one array in order defined by other array without doing 2 sort...

I have two numpy arrays:

import numpy as np
arr1 = np.array([1,4,2,5,6,3,2])
arr2 = np.array([1,0,2,4,3,6,5])


I would like to find the cumulative sum of arr1 according to the order defined in arr2. Expected output:

array([ 5, 4, 7, 18, 13, 23, 20])


Here's one way to calculate this:

sorting_indices = np.argsort(arr2)
result = np.cumsum(arr1[sorting_indices])[np.argsort(sorting_indices)]


However, this involves doing two sort/argsort operations (np.argsort(arr2) and then np.argsort(sorting_indices))

Is it possible to produce the same result, but doing only a single sort or argsort operation?

Continue reading...