Numpy - get mask from shape and indices

Question

After using np.instersect1d I have retrieved the indices that I want to retrieve. I want to turn these indices into a mask so I can then combine it with another mask.

What I am wanting:

>>> times_1.shape
... (160,)

>>> _, time_indices, _ = np.intersect1d(times_1, times_2, return_indices=True)

>>> time_indices.shape
... (145,)

    # Some way to turn return mask from original shape and indices
>>> time_mask = get_mask_from_shape(time_1.shape, time_indices)

>>> time_mask.shape
... (160,)

What would be the simplest way to implement the function get_mask_from_shape?

score 1 · Accepted Answer · answered Jan 16 '20 at 18:48

1

The solution I've found is:

time_mask = np.zeros(times_1.shape, dtype=bool)
time_mask[time_indices] = True

answered Jan 16 '20 at 18:48

Max Collier

573
8
24

mathfux · Answer 2 · 2020-01-15T23:31:56.143

-1

If I get an idea right, I would apply this pattern I've been looking for a few days before:

intersection = np.intersect1d(times_1, times_2)
intersection = np.sort(intersection)
locations = np.searchsorted(intersection, times_1)
locations[locations==len(intersection)] = 0
time_mask = intersection[locations]==times_1

searchsorted method finds indices where elements of times_1 should be inserted into intersection to maintain order, according to numpy documentation.

edited Jan 15 '20 at 23:31

answered Jan 15 '20 at 23:26

mathfux

5,759
1
14
34

Answer is fairly long winded, would be good to have the answer be a few lines... – Max Collier Jan 16 '20 at 00:50
Was the output as you expected? – mathfux Jan 16 '20 at 08:17
This is massive overkill. You are performing very expensive operations unnecessarily. Why are you sorting and searching? – Mad Physicist Jan 16 '20 at 18:54
Sorry for misunderstandings. My method is efficient only if intersections were used repeatedly with the same array. – mathfux Jan 17 '20 at 13:19

Numpy - get mask from shape and indices

2 Answers2