双索引后更改数组元素

Question

I have 2 logical arrays of unequal sizes (let's call bigger one b and smaller one c) and one which has numbers in it (array a). Size of the larger logical array is the same size as the array of numbers, while the size of smaller is equal to the sum of all truths in a bigger array. I want to change the element at index i only if b_i is jth truth and c_j is true.
Sorry for a confusing explanation, but hopefully it will be clearer in code.

In other words we would only check c if b was true, and index at which we check c is given by which truth in b it was.

I can create a tmp variable, which is of size c, and then apply c to it, but it makes my code really messy, which I would prefer to avoid.

import numpy as np
a = np.arange(6)
b = [True, False, True, False, True, False]
c = [True, False, True]
print(a[b][c])
a[b][c] = 2
print(a[b][c])

# Messy workaround
tmp = a[b]
tmp[c] = 2
a[b] = tmp
print(a[b][c])

First print outputs [0, 4] , which is correct. However, I expected second output to be an array of [2, 2 ], while it's [0, 4] again (similarly to third output, the workaround).

Answer 1

Mask-the-mask approach

We can mask the first mask with itself while assigning into it the second one, so that only the True ones in the first one are changed according to the second mask. This sets up the first mask for boolean-indexing into the values array, like so -

b[b] = c # mask first mask with itself and assign second mask
a[b] = 2 # assign new value(s)

Note that it works with arrays.

Sample run -

In [48]: a = np.arange(6)
    ...: b = np.array([True, False, True, False, True, False])
    ...: c = np.array([True, False, True])

In [49]: a
Out[49]: array([0, 1, 2, 3, 4, 5])

In [50]: b
Out[50]: array([ True, False,  True, False,  True, False])

# Mask the first mask with itself and assign the second one into it
In [51]: b[b] = c

# Verify the new mask
In [52]: b
Out[52]: array([ True, False, False, False,  True, False])

# Index values array with it and assign new value(s)
In [53]: a[b] = 2

# Verify edited values array
In [54]: a
Out[54]: array([2, 1, 2, 3, 2, 5])

Alternative with Indices

An alternative would be using the indices of the True values in the first mask -

In [59]: a = np.arange(6)
    ...: b = np.array([True, False, True, False, True, False])
    ...: c = np.array([True, False, True])

In [60]: idx = np.flatnonzero(b)

In [61]: a[idx[c]] = 2

In [62]: a
Out[62]: array([2, 1, 2, 3, 2, 5])

I would go with the first one, for its memory-efficiency as boolean arrays have much lower memory footprint and moreover in this case, it's an in-situ edit into the already existing boolean array.