Array Manipulation Functions

NumPy comes with lots of functions for manipulating arrays. Some of them are needed more often, others almost never. A comprehensive list is provided in array manipulation routines. Here we only mention some of the more important ones.

import numpy as np

Modifying Shape with reshape

A NumPy array’s reshape method yields an array of different shape, but with identical data. The new array has to have the same number of elements as the old one.

a = np.ones(5)         # 1d (vector)
b = a.reshape(1, 5)    # 2d (row matrix)
c = a.reshape(5, 1)    # 2d (column matrix)

print(a, '\n')
print(b, '\n')
print(c)

[1. 1. 1. 1. 1.] 

[[1. 1. 1. 1. 1.]] 

[[1.]
 [1.]
 [1.]
 [1.]
 [1.]]

One dimension may be replaced by -1 indicating that the size of this dimension shall be computed by NumPy:

a = np.ones((8, 8))
b = a.reshape(4, -1)

print(a.shape, b.shape)

(8, 8) (4, 16)

Mirrowing with fliplr and flipud

To mirrow a 2d array on its vertical or horizontal axis use fliplr and flipud, respectively.

a = np.array([[1, 2, 3], [4, 5, 6]])
b = np.fliplr(a)

print(a, '\n')
print(b)

[[1 2 3]
 [4 5 6]] 

[[3 2 1]
 [6 5 4]]

Joining Arrays with concatenate and stack

Arrays of identical shape (except for one axis) may be joined along an existing axis to one large array with concatenate.

a = np.ones((2, 3))
b = np.zeros((2, 5))
c = np.full((2, 2), 5)
d = np.concatenate((a, b, c), axis=1)

print(d)

[[1. 1. 1. 0. 0. 0. 0. 0. 5. 5.]
 [1. 1. 1. 0. 0. 0. 0. 0. 5. 5.]]

If identically shaped array shall be joined along a new axis, use stack.

a = np.ones(2)
b = np.zeros(2)
c = np.full(2, 5)
d = np.stack((a, b, c), axis=1)

print(d)

[[1. 0. 5.]
 [1. 0. 5.]]

Appending Data with append

Like Python lists NumPy arrays may be extended by appending further data. The append method takes the original array and the new data and returns the extended array.

a = np.ones((3, 3))
b = np.append(a, [[1, 2, 3]], axis=0)

print(b)

[[1. 1. 1.]
 [1. 1. 1.]
 [1. 1. 1.]
 [1. 2. 3.]]