"In" operator for numpy arrays?

Question

How can I do the "in" operation on a numpy array? (Return True if an element is present in the given numpy array)

For strings, lists and dictionaries, the functionality is intuitive to understand.

Here's what I got when I applied that on a numpy array

a
array([[[2, 3, 0],
    [1, 0, 1]],

   [[3, 2, 0],
    [0, 1, 1]],

   [[2, 2, 0],
    [1, 1, 1]],

   [[1, 3, 0],
    [2, 0, 1]],

   [[3, 1, 0],
    [0, 2, 1]]])

b = [[3, 2, 0],
    [0, 1, 1]]

b in a
True
#Aligned with the expectation

c = [[300, 200, 0],
    [0, 100, 100]]

c in a
True
#Not quite what I expected

For 1d arrays there is a np.in1d, but applying it to rows of a 2d array requires some tricks. Look at it's code to see what is involved. — hpaulj
– hpaulj, Commented Sep 12, 2016 at 16:13

Divakar · Accepted Answer · 2016-09-12 14:50:25Z

7

You could compare the input arrays for equality, which will perform broadcasted comparisons across all elements in a at each position in the last two axes against elements at corresponding positions in the second array. This will result in a boolean array of matches, in which we check for ALL matches across the last two axes and finally check for ANY match, like so -

((a==b).all(axis=(1,2))).any()

Sample run

1) Inputs :

In [68]: a
Out[68]: 
array([[[2, 3, 0],
        [1, 0, 1]],

       [[3, 2, 0],
        [0, 1, 1]],

       [[2, 2, 0],
        [1, 1, 1]],

       [[1, 3, 0],
        [2, 0, 1]],

       [[3, 1, 0],
        [0, 2, 1]]])

In [69]: b
Out[69]: 
array([[3, 2, 0],
       [0, 1, 1]])

2) Broadcasted elementwise comparisons :

In [70]: a==b
Out[70]: 
array([[[False, False,  True],
        [False, False,  True]],

       [[ True,  True,  True],
        [ True,  True,  True]],

       [[False,  True,  True],
        [False,  True,  True]],

       [[False, False,  True],
        [False, False,  True]],

       [[ True, False,  True],
        [ True, False,  True]]], dtype=bool)

3) ALL match across last two axes and finally ANY match :

In [71]: (a==b).all(axis=(1,2))
Out[71]: array([False,  True, False, False, False], dtype=bool)

In [72]: ((a==b).all(axis=(1,2))).any()
Out[72]: True

Following similar steps for c in a -

In [73]: c
Out[73]: 
array([[300, 200,   0],
       [  0, 100, 100]])

In [74]: ((a==c).all(axis=(1,2))).any()
Out[74]: False

edited Sep 12, 2016 at 14:50

answered Sep 12, 2016 at 14:40

Divakar

222k19 gold badges273 silver badges374 bronze badges

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

2 Comments

akilat90 Over a year ago

Thanks for saving the day! So no straightforward way like "in" to check the existence on a numpy array?

Divakar Over a year ago

@akilat90 I am afraid there isn't. Or just wrap it into a function and there you go, a custom made straightforward way!

soo-zie · Accepted Answer · 2020-11-05 07:27:24Z

Since I do not have enough reputation to comment but the answer suggested by @rp0 using np.isin and then np.all will not work with numpy arrays. Counter Example:

a = np.array([[[2, 3, 0],
               [1, 0, 1]],
              [[3, 2, 0],
               [0, 1, 1]],
              [[2, 2, 0],
               [1, 1, 1]],
              [[1, 3, 0],
               [2, 0, 1]],
              [[3, 1, 0],
               [0, 2, 1]]])
b = [[3, 2, 0],
     [0, 1, 1]]

c = [[3, 3, 3],
     [3, 3, 3]]

print(np.isin(b,a))
[[ True  True  True]
 [ True  True  True]]

print(np.isin(c,a))
[[ True  True  True]
 [ True  True  True]]

For completeness, I converted the array to python list using tolist and then used normal 'in' operator and its worked just as expected.

print(b in a.lolist())
True

print(c in a.tolist())
False

rp0 · Accepted Answer · 2019-07-10 15:25:59Z

1

This question is fairly old, but if you're like me, you might have thought there was no numpy equivalent for in by reading it.

Numpy 1.13 was released in 2017, and with it came the function isin(), which now nicely solves the problem:

import numpy as np

a = np.array([[[2, 3, 0],
               [1, 0, 1]],
              [[3, 2, 0],
               [0, 1, 1]],
              [[2, 2, 0],
               [1, 1, 1]],
              [[1, 3, 0],
               [2, 0, 1]],
              [[3, 1, 0],
               [0, 2, 1]]])

b = [[3, 2, 0],
     [0, 1, 1]]

print np.isin(b,a)

# [[ True  True  True]
#  [ True  True  True]]

c = [[300, 200, 0],
    [0, 100, 100]]

print np.isin(c,a)

# [[False False  True]
#  [ True False False]]

You'll probably want to use np.all() at the end if you're looking for the entire element to be in the test array.

answered Jul 10, 2019 at 15:25

rp0

616 bronze badges

1 Comment

pabouk - Ukraine stay strong Over a year ago

@AturSams This is not probably what is expected. The matches are regardles the item position and np.all() with np.isin() will not do what you probably expect as soo-zie pointed out

Collectives™ on Stack Overflow

"In" operator for numpy arrays?

3 Answers 3

2 Comments

Comments

1 Comment

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

2 Comments

Comments

1 Comment

Your Answer

Sign up or log in

Post as a guest

Linked

Related