How to swap columns of a given NumPy array?

Swapping columns of a NumPy array means exchanging the positions of two specified columns across all rows. For example, if you have a 3x3 array with values like [[0, 1, 2], [3, 4, 5], [6, 7, 9]] and you swap column 0 with column 2, the array becomes [[2, 1, 0], [5, 4, 3], [9, 7, 6]]. Let’s explore different ways to do this efficiently using NumPy.

Using indexing

This is the simplest method, where we use slicing to specify the column indices to swap like a[:, [0, 2]] = a[:, [2, 0]]. It swaps the columns in-place, making it fast and ideal for quick operations on the original array.

import numpy as np
a = np.arange(12).reshape(4, 3)
print("Original array:\n", a)
a[:, [0, 2]] = a[:, [2, 0]]
print("After swapping:\n",a)

Original array:
 [[ 0  1  2]
 [ 3  4  5]
 [ 6  7  8]
 [ 9 10 11]]
After swapping:
 [[ 2  1  0]
 [ 5  4  3]
 [ 8  7  6]
 [11 10  9]]

Explanation: a[:, [0, 2]] = a[:, [2, 0]] selects all rows and swaps columns 0 and 2. It replaces column 0 with column 2 and vice versa, modifying the original array in-place without creating a new one.

Using column reordering

This method uses list comprehension to create a new column order, offering flexibility for dynamic or reusable functions. Define i and j, build the new index list and reindex the array. It requires slightly more code but provides greater control.

import numpy as np
a = np.arange(12).reshape(4, 3)
print("Original array:\n", a)
i, j = 0, 2
res = a[:, [j if x == i else i if x == j else x for x in range(a.shape[1])]]
print(res)

Original array:
 [[ 0  1  2]
 [ 3  4  5]
 [ 6  7  8]
 [ 9 10 11]]
[[ 2  1  0]
 [ 5  4  3]
 [ 8  7  6]
 [11 10  9]]

Explanation: List comprehension swaps columns 0 and 2 by generating a new column order. It selects all rows and returns a new array with the specified columns swapped, leaving the original array unchanged.

Using np.take()

np.take() lets you reorder or extract elements along a specific axis. For column swapping, create a new column order and use np.take() to apply it. It returns a new array, keeping the original unchanged.

import numpy as np
a = np.arange(12).reshape(4, 3)
print("Original array:\n", a)
i, j = 0, 2
col_order = [j if x == i else i if x == j else x for x in range(a.shape[1])]
res = np.take(a, col_order, axis=1)
print("After swapping:\n", res)

Original array:
 [[ 0  1  2]
 [ 3  4  5]
 [ 6  7  8]
 [ 9 10 11]]
After swapping:
 [[ 2  1  0]
 [ 5  4  3]
 [ 8  7  6]
 [11 10  9]]

Explanation: np.take(a, col_order, axis=1) swaps columns 0 and 2 by applying the new column order col_order to all rows using np.take(). This returns a new array with the specified columns swapped, while keeping the original array unchanged.

Using np.copy()

To preserve the original data, use np.copy() and swap columns on the copy using direct indexing. It’s a safe and simple method that leaves the original array untouched.

import numpy as np
a = np.arange(12).reshape(4, 3)
print("Original array:\n", a)
res = a.copy()
res[:, [0, 2]] = res[:, [2, 0]]
print("After swapping:\n", res)

Original array:
 [[ 0  1  2]
 [ 3  4  5]
 [ 6  7  8]
 [ 9 10 11]]
After swapping:
 [[ 2  1  0]
 [ 5  4  3]
 [ 8  7  6]
 [11 10  9]]

Explanation: a.copy() creates a copy to preserve the original array. The line res[:, [0, 2]] = res[:, [2, 0]] swaps columns 0 and 2 in the copy, returning a modified array while leaving the original unchanged.