partition() – Partial Sort in NumPy
Hi Enthusiastic Learners! In this article we will be studying about “partition() – partial sort in NumPy“. Have you ever wondered, if you are given an array and you only want n smallest numbers from your array instead of completely sorting it? Well, that is exactly the case where we should use partition() - partial sort & and we are calling partition() a partial sort because we actually will be sorting only few parts of given array or you can say getting a partition of n smallest values. Before jumping on to partition() – partial sort in NumPy, please check out other articles related to sorting in NumPy:
Watch video tutorial here:
partition() – Partial Sort Syntax
np.partition(arr, pos, axis=-1, kind='introselect', order=None)
Where,
- arr is the array which has to be sorted
- pos position in array.
- axis defines the axis along which you need to do sorting & -1 value means last axis.
- kind gives you liberty to choose sorting algorithm. Default algo ‘introselect’
- order — if arr is a structured array, this argument specifies which fields to compare first, second, etc.
Let’s begin with creating a array arr.
import numpy as np
arr = np.array([3, 7, 4, 2, 8, 9, 0, 18, 1])
print("-- Base Array --")
print(arr)
Sorting array for pos = 3. Now, partition() will create a copy of original array and will arrange number in such a way such that till pos all values will be the smallest values in array & after that position other values will be as it is.
Note that they may not be ordered. Its just that, till pos = 3, that is 4 elements as indexing in arrays starts from 0, we will be getting 4 smallest values & remaining values will be shifted as it is towards right and there is no order in sorted values.
Syntax for it will be — np.partition(arr, 2)
sort_for_3 = np.partition(arr, 3)
print("-- Sorted Array till 3rd Position --")
print(sort_for_3)
As, we can clearly from above example, that 1st four elements (till pos = 3) are smallest among complete array & other values are just shifted towards right in the same order they naturally were. Also, there is no order in sorted values.
partition() on 2-dimensional array
We can do partial sort for 2-D arrays as well & we can choose axis along which we want values to be sorted.
Let’s begin with creating a 2-D array
arr_2d = np.random.randint(0, 10, (4, 6))
print("-- 2D Array --\n")
print(arr_2d)
Get 2 smallest values for each row of array.
Syntax for it will be — np.partition(arr_2d, 1, axis=1)
sorted_2d = np.partition(arr_2d, 1, axis=1)
print("-- Partially Sorted 2D Array -- \n")
print(sorted_2d)
From above example it clear, that for each row we are getting smallest value till index (pos) value = 1, that is 2 smallest elements in each row.
Stay tuned & keep learning! In our next post we will be covering more sorting techniques that can be achieved in NumPy.