How to select on masked indices from non-masked array?

Question

I just realized that my masked array doesn't work as indices for selection.

When I do mathematical operations, such as max() it works, but not with selections.

import numpy as np

array = np.arange(3,8)
indices = np.ma.masked_array(np.arange(5),np.random.randint(0,2,5))

print('array data: %s' % array)
print('indices: %s' % indices)
print('-')
print('max index: %s - ok' % indices.max())
print('selecting on indices %s - not ok' % array[indices])

Am I missing something? Why wouldn't the above work?

Please provide a concrete example. It's not clear why you aren't using normal masking syntax with negation. — ddejohn, Feb 17 '22 at 23:38
Use the above pls. The issue is there is it always returns 5 elements. Basically ignoring the mask, even though `indices` is a masked array. — Anonymous, Feb 17 '22 at 23:40
Do you mean `array[indices[~indices.mask]]`? I just thought of it. Seems like it's working. Is this the standard syntax? — Anonymous, Feb 17 '22 at 23:43
What do you mean? `array[indices.mask]` is returning different numbers of elements every time I use it. *Please provide a concrete example* of your input and expected output. Please see this post on providing a [minimal reproducible example](https://stackoverflow.com/help/minimal-reproducible-example). Using randomly generated numbers is, almost by definition, not reproducible. — ddejohn, Feb 17 '22 at 23:43

John Giorgio · Answer 1 · 2022-02-17T23:56:48.853

1

EDIT Try to play around with n_elements. Mask and data size must be the same.

import numpy as np

n_elements = 5
x = 3

array = np.arange(x, x + n_elements)

indices = np.ma.masked_array(np.arange(n_elements), 
                             np.random.randint(0, 2, n_elements))

print('array data: %s' % array)
print('indices: %s' % indices)
print('-')
print('max index: %s - ok' % indices.max())
print('selecting on indices %s' % array[indices.mask])
print('excluding on indices %s' % array[~indices.mask])

ORIGINAL Maybe you're looking for the below?

array[indices.mask]

edited Feb 17 '22 at 23:56

answered Feb 17 '22 at 23:27

John Giorgio

634
3
10

Won't work if we use different indices: `np.ma.masked_array(np.arange(2,4),np.random.randint(0,2,2))` – Anonymous Feb 17 '22 at 23:35
1

It doesn't work because you are producing a mask of 2 elements while your data has 5 elements. – John Giorgio Feb 17 '22 at 23:43
Wrong. arange 2,4 is 2 element array. – Anonymous Feb 17 '22 at 23:45
1

Are you sure you're not still using `array = np.arange(3,8)`, which is made of 5 elements? – John Giorgio Feb 17 '22 at 23:48
I do, but from array of 5 elements you can select 2 indices. duh – Anonymous Feb 17 '22 at 23:50
1

Yes, but data and mask must be the same size. I edited my answer, hope it helps – John Giorgio Feb 17 '22 at 23:57

score -2 · Answer 2 · answered Feb 17 '22 at 23:49

-2

Apparently, I should be using array[indices[~indices.mask]] instead of array[indices]. So instead of just specifying the indices array I should also explicitly apply it's mask on it.

answered Feb 17 '22 at 23:49

Anonymous

4,692
8
61
91

1

Actually `array[~indices.mask]` is enough. – John Giorgio Feb 18 '22 at 00:00
No it's not, if indices don't start with 0. – Anonymous Feb 18 '22 at 00:12
There is still a problem though. The result has fewer elements. It should have 5 and have some of them masked as well. – Anonymous Feb 18 '22 at 00:26

How to select on masked indices from non-masked array?

2 Answers2