Using Numpy to replace values in place in arbitrary axis

Question

How can I replace array values in place if I don't know the axis beforehand?

For example, if I wanna do something like

arr[:,5]

but I don't know the axis beforehand and want to make it general I can use take:

arr.take(5, axis=1)

and it'll work.

However, if I want to something like

arr[:,5]=10

but I don't know the axis beforehand, how can I do it? I obviously can't do arr.take(5, axis=1) = 10, and I can't find a function to do it.

The function that comes the closest (that I found) would be np.put(), but I don't think it can be done with that.

Warren Weckesser · Accepted Answer · 2017-09-28 22:23:43Z

You could swap the desired axis to the first position and then do the assignment. swapaxes returns a view, so the assignment will do what you want.

For example,

In [87]: np.random.seed(123)

In [88]: a = np.random.randint(1, 9, size=(5, 8))

In [89]: a
Out[89]: 
array([[7, 6, 7, 3, 5, 3, 7, 2],
       [4, 3, 4, 2, 7, 2, 1, 2],
       [7, 8, 2, 1, 7, 1, 8, 2],
       [4, 7, 6, 5, 1, 1, 5, 2],
       [8, 4, 3, 5, 8, 3, 5, 8]])

In [90]: ax = 1

In [91]: k = 5

In [92]: val = 99

In [93]: a.swapaxes(0, ax)[k] = val

In [94]: a
Out[94]: 
array([[ 7,  6,  7,  3,  5, 99,  7,  2],
       [ 4,  3,  4,  2,  7, 99,  1,  2],
       [ 7,  8,  2,  1,  7, 99,  8,  2],
       [ 4,  7,  6,  5,  1, 99,  5,  2],
       [ 8,  4,  3,  5,  8, 99,  5,  8]])

In [95]: ax = 0

In [96]: k = 2

In [97]: val = -1

In [98]: a.swapaxes(0, ax)[k] = val

In [99]: a
Out[99]: 
array([[ 7,  6,  7,  3,  5, 99,  7,  2],
       [ 4,  3,  4,  2,  7, 99,  1,  2],
       [-1, -1, -1, -1, -1, -1, -1, -1],
       [ 4,  7,  6,  5,  1, 99,  5,  2],
       [ 8,  4,  3,  5,  8, 99,  5,  8]])

unutbu · Accepted Answer · 2017-09-28 22:13:42Z

2

I don't think there is a NumPy function for this, but it is not too hard to construct your own:

def replace(arr, indices, val, axis):
    s = [slice(None)]*arr.ndim
    s[axis] = indices
    arr[s] = val

import numpy as np

def replace(arr, indices, val, axis):
    s = [slice(None)]*arr.ndim
    s[axis] = indices
    arr[s] = val

arr = np.zeros((3,6,2))
indices = 5
axis = 1
val = 10

replace(arr, indices, val, axis)
print(np.take(arr, indices, axis))

prints

[[ 10.  10.]
 [ 10.  10.]
 [ 10.  10.]]

answered Sep 28, 2017 at 22:13

unutbu

886k197 gold badges1.9k silver badges1.7k bronze badges

2 Comments

Acccumulation Over a year ago

"replace" is already a method for several classes. I would recommend another name.

unutbu Over a year ago

@Acccumulation: Okay, but then what would you suggest?

Collectives™ on Stack Overflow

Using Numpy to replace values in place in arbitrary axis

2 Answers 2

Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related