Deleting multiple slices from a numpy array

Question

I have a given numpy array and a list containing a number of slice objects (alternatively containing (start, end) tuples). I am looking to remove the slice object positions from the original array and get a second array with the remaining values.

Toy example:

myarray = np.arange(20)

array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16,
       17, 18, 19])

mylist=(slice(2,4),slice(15,19))

Do something and result should be

array([0, 1, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14])

The array can be in a few hundred thousand large, the list of slice objects can contain a few thousand elements and I need to run the operation often, so speed is somewhat important.

Numpy delete does not take a list of slices as far I can see?

For now I am generating the complement of my slice object list and slicing that, but generating the complement is a somewhat awkward process where I am sorting my slice list then iterating through it, creating the complement slice objects as needed. I am hoping there is a more elegant way I have not figured!

HYRY · Accepted Answer · 2014-07-09 04:31:16Z

3

You can use np.r_[] to join the slices into an array:

myarray = np.arange(20)
mylist=(slice(2, 4),slice(15, 19))
np.delete(myarray, np.r_[tuple(mylist)])

output:

array([ 0,  1,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 19])

But I think it's not very fast.

edited Jul 9, 2014 at 4:31

answered Jul 9, 2014 at 4:19

HYRY

97.8k28 gold badges197 silver badges192 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Saullo G. P. Castro · Accepted Answer · 2014-07-09 06:49:35Z

1

You can use set() to identify which positions will be kept and np.take() to obtain the corresponding values, doing something like:

ind = np.indices(myarray.shape)[0]
rm = np.hstack([ind[i] for i in mylist])

ans = np.take(myarray, sorted(set(ind)-set(rm)))

Note that np.hstack() is used to obtain a single array with all the indices that will be removed. This takes about half the time of @HYRY's solution.

answered Jul 9, 2014 at 6:49

Saullo G. P. Castro

59.4k28 gold badges191 silver badges244 bronze badges

Comments

AMacK · Accepted Answer · 2014-07-08 22:43:12Z

1

I can't think of a way to join the slices cleanly; however, I think using the composite is the way to go. Perhaps try something like this:

import numpy as np

# Create test data
n_data = 1000000
n_slices = 10000

data = np.arange(n_data)
slices = []
for i in range(n_slices):
    r = np.random.randint(n_data-1000)
    slices.append(slice(r,r + np.random.randint(1000)))

# Remove slices
keep_mask = np.ones_like(data, dtype=bool)
for slice in slices: keep_mask[slice] = False
data = data[keep_mask] # or np.take, etc.

answered Jul 8, 2014 at 22:43

AMacK

1,3969 silver badges9 bronze badges

Collectives™ on Stack Overflow

Deleting multiple slices from a numpy array

3 Answers 3

Comments

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related