Concatenation And Splitting

1. CONCATENATING ARRAYS
np.concatenate() constructor is used to concatenate or join two or more arrays into one. The only required argument is list or tuple of arrays.
💡For np.concatenate() to work, all the input array dimensions for the concatenation axis must match exactly, other wise you will get ValueError
1.1. 1D arrays
Let’s get straight to work and define two arrays and then join them:
# first, import numpy
import numpy as np
​
# making two arrays to concatenate
arr1 = np.arange(1,4)
arr2 = np.arange(4,7)
​
print("Arrays to concatenate:")
print(arr1);print(arr2)
​
print("After concatenation:")
print(np.concatenate([arr1,arr2]))
Arrays to concatenate:
[1 2 3]
[4 5 6]
After concatenation:
[1 2 3 4 5 6]
We can also join more than two arrays:
# we can concatenate more than two arrays
arr3 = np.arange(7,10)
​
np.concatenate([arr1,arr2,arr3])
array([1, 2, 3, 4, 5, 6, 7, 8, 9])
1.2. nD arrays
The same constructor is used for joining two or more nD arrays
By default, np.concat joins along the row wise (stack on top of each other). However, by providing kwarg axis=1, we can also concatenate along columns.
# creating 2D arrays to join
ar2d1 = np.arange(1,7).reshape(2,3)
ar2d2 = np.arange(7,13).reshape(2,3)
print("2D arrays to concatenate:")
print(ar2d1);print(ar2d2)
​
# case1: joining along rows
print("\nConcatenating along rows:")
print(np.concatenate([ar2d1,ar2d2]))
​
# case1: joining along columns
print("\nConcatenating along columns:")
print(np.concatenate([ar2d1,ar2d2], axis=1))
2D arrays to concatenate:
[[1 2 3]
 [4 5 6]]
[[ 7  8  9]
 [10 11 12]]
​
Concatenating along rows:
[[ 1  2  3]
 [ 4  5  6]
 [ 7  8  9]
 [10 11 12]]
​
Concatenating along columns:
[[ 1  2  3  7  8  9]
 [ 4  5  6 10 11 12]]
2. CONCATENATING ARRAYS OF MIXED DIMENSIONS
To concatenate arrays of mixed dimensions we will use np.vstack and np.hstack functions
2.1. Vertical Stack ( np.vstack)
np.vstack concatenate along rows, stack vertically.
💡 To make vstack work, all input arrays should have same number of columns
# defining two 2D arrays
v1 = np.arange(8).reshape(2,4)
v2 = np.arange(20).reshape(5,4)
​
print("We vstack these arrays:\n")
print(f"Array 1: \n{v1}");print(f"Array 2: \n{v2}") 
​
# vstack
print("\nAfter vStack-ed both arrays")
print(np.vstack([v1,v2]))
We will vstack these arrays:
​
Array 1: 
[[0 1 2 3]
 [4 5 6 7]]
Array 2: 
[[ 0  1  2  3]
 [ 4  5  6  7]
 [ 8  9 10 11]
 [12 13 14 15]
 [16 17 18 19]]
​
After vStack-ed both arrays
[[ 0  1  2  3]
 [ 4  5  6  7]
 [ 0  1  2  3]
 [ 4  5  6  7]
 [ 8  9 10 11]
 [12 13 14 15]
 [16 17 18 19]]
2.1. Horizontal Stack ( np.hstack)
np.hstack concatenate along columns, stack one after the other.
💡To make hstack work, both arrays should have same number of rows
h1 = np.arange(10).reshape(5,2)
h2 = np.arange(30).reshape(5,6)
print("We will hstack these arrays:\n")
print(f"Array 1: \n{h1}");print(f"Array 2: \n{h2}") 
​
print("\nAfter hStack-ed both arrays")
print(np.hstack([h1,h2]))
We will hstack these arrays:
​
Array 1: 
[[0 1]
 [2 3]
 [4 5]
 [6 7]
 [8 9]]
Array 2: 
[[ 0  1  2  3  4  5]
 [ 6  7  8  9 10 11]
 [12 13 14 15 16 17]
 [18 19 20 21 22 23]
 [24 25 26 27 28 29]]
​
After hStack-ed both arrays
[[ 0  1  0  1  2  3  4  5]
 [ 2  3  6  7  8  9 10 11]
 [ 4  5 12 13 14 15 16 17]
 [ 6  7 18 19 20 21 22 23]
 [ 8  9 24 25 26 27 28 29]]
np.dstack is used for stacking 3 arrays
3. SPLITTING
Opposite of concatenation is splitting and we will use np.split, np.vsplit and np.hsplit to split the arrays
3.1. np.split
np.split for N split points creates N+1, sub-arrays np.split(x,y) will split the array into three subarray. An example will make the concept clearer:
# array to splt
ns = np.arange(1,10)
​
print(f"Original array: {ns}")
print("After split:")
ns1,ns2,ns3 = np.split(ns, (3,6))
print(ns1);print(ns2);print(ns3)
Original array: [1 2 3 4 5 6 7 8 9]
After nsplit:
[1 2 3]
[4 5 6]
[7 8 9]
3.2… np.vsplit
np.vsplit splits along the vertical axis. You can either provide:
integer of equally shaped array (method 1 below), or,
by specifying the row [integer] at which the division should occur (method 2 below)
vs = np.arange(16).reshape(4,4)
print(f"Array to vSplit: \n{vs}")
​
# method 1, by specifying number of equally shaped arrays
print("\nUsing Method 1")
upper, lower = np.vsplit(vs, 2)
print(f"Upper: \n{upper}\nLower: \n{lower}")
​
# method 2, by specifying row, after which division should occur
print("\nUsing Method 2")
print(np.vsplit(vs, [2]))
Array to vSplit: 
[[ 0  1  2  3]
 [ 4  5  6  7]
 [ 8  9 10 11]
 [12 13 14 15]]
​
Using Method 1
Upper: 
[[0 1 2 3]
 [4 5 6 7]]
Lower: 
[[ 8  9 10 11]
 [12 13 14 15]]
​
Using Method 2
[array([[0, 1, 2, 3],
       [4, 5, 6, 7]]), array([[ 8,  9, 10, 11],
       [12, 13, 14, 15]])]
3.3. np.hsplit
np.hsplit splits along the horizontal axis. You can either provide:
number integer of equally shaped array, or,
by specifying the column [integer] at which the division should occur
hs = np.arange(16).reshape(2,8)
print(f"Array to hSplit: \n{hs}")
​
# method 1, by specifying number of equally shaped arrays
print("\nUsing Method 1")
right, left = np.hsplit(hs, 2)
print(f"Right: \n{right}\nLeft: \n{left}")
​
# method 2, by specifying columns, after which division should occur
print("\nUsing Method 2")
print(np.hsplit(hs, [4]))
Array to hSplit: 
[[ 0  1  2  3  4  5  6  7]
 [ 8  9 10 11 12 13 14 15]]
​
Using Method 1
Right: 
[[ 0  1  2  3]
 [ 8  9 10 11]]
Left: 
[[ 4  5  6  7]
 [12 13 14 15]]
​
Using Method 2
[array([[ 0,  1,  2,  3],
       [ 8,  9, 10, 11]]), array([[ 4,  5,  6,  7],
       [12, 13, 14, 15]])]
Numpy Fancy Indexing
Numpy's Universal Functions
Last modified 4mo ago