22
votes

I have two NumPy arrays, e.g.:

a = [1,2,3,4,5]

and a filter array, e.g.:

f = [False, True, False, False, True]

len(a) == len(f)

How can I get a new numpy array with only the values in a where the same index in f is True? In my case: [2, 5].

According to the accepted solution (with different values):

>>> a = numpy.array([ 1,  2,  3,  4,  5,  6,  7,  8,  9, 10])
>>> b = numpy.array([True, False, True, False, True, False, True, False, True, False])
>>> a[b]
array([1, 3, 5, 7, 9])
2
it looks like b is a list not an array, b must be a boolean array. Try b = np.asarray(b, 'bool') - Bi Rico

2 Answers

33
votes

NumPy supports boolean indexing

a[f]

This assumes that a and f are NumPy arrays rather than Python lists (as in the question). You can convert with f = np.array(f).

4
votes

If you don't already need numpy arrays, here's with a plain list:

import itertools
print itertools.compress(a, f)

For pre-2.7 versions of python, you must roll your own (see manual):

def compress(data, selectors):
    return (d for d, s in itertools.izip(data, selectors) if s)