np.argsort() for sort indices in NumPy

Introduction

We use np.argsort() to find the order of elements that would sort an array. It tells us the positions to rearrange the data instead of sorting the data itself.

When you want to sort data but keep track of original positions.

When you need to reorder one array based on the sorted order of another array.

When you want to find rankings or positions of elements in sorted order.

When you want to sort data without changing the original array.

Syntax

NumPy

np.argsort(a, axis=-1, kind='quicksort', order=None)

a is the input array to find sort indices for.

The result is an array of indices that sort a.

Examples

This finds indices that sort the array [3, 1, 2]. The output shows positions of elements in ascending order.

NumPy

import numpy as np
arr = np.array([3, 1, 2])
indices = np.argsort(arr)
print(indices)

This finds sort indices along each row of a 2D array.

NumPy

arr = np.array([[3, 1], [2, 4]])
indices = np.argsort(arr, axis=1)
print(indices)

This uses argsort to sort the array by rearranging elements using the indices.

NumPy

arr = np.array([3, 1, 2])
sorted_arr = arr[np.argsort(arr)]
print(sorted_arr)

Sample Program

This program shows how to get sort indices with np.argsort(), use them to sort the array, and find the rank (position) of each element in the sorted order.

NumPy

import numpy as np

# Original array
arr = np.array([50, 20, 30, 10, 40])

# Get indices that would sort the array
sort_indices = np.argsort(arr)
print('Sort indices:', sort_indices)

# Use indices to sort the array
sorted_arr = arr[sort_indices]
print('Sorted array:', sorted_arr)

# Example: Find rank of each element
ranks = np.empty_like(sort_indices)
ranks[sort_indices] = np.arange(len(arr))
print('Ranks of elements:', ranks)

Important Notes

np.argsort() returns indices, not sorted values.

You can use the indices to reorder the original array or related arrays.

For multidimensional arrays, specify axis to sort along rows or columns.

Summary

np.argsort() gives the order of indices to sort an array.

Use it to sort arrays without changing the original data directly.

It helps find rankings and reorder related data consistently.

Practice

1. What does the np.argsort() function return when applied to a numpy array?

easy

A. The sum of all elements in the array

B. The sorted array itself

C. The maximum value in the array

D. An array of indices that would sort the original array

Solution

1. Step 1: Understand the purpose of np.argsort()

   This function does not sort the array directly but returns the indices that would sort the array.

2. Step 2: Differentiate from sorting functions

   Unlike np.sort() which returns the sorted array, np.argsort() returns the order of indices to achieve that sorting.

3. Final Answer:

   An array of indices that would sort the original array -> Option D

4. Quick Check:

   np.argsort() = indices order [OK]

Hint: Remember: argsort returns indices, not sorted values [OK]

Common Mistakes:

• Confusing argsort with sort and expecting sorted values
• Thinking argsort returns the maximum or minimum value
• Assuming argsort returns a scalar instead of an array

2. Which of the following is the correct syntax to get the indices that would sort the array arr using a NumPy function?

easy

A. arr.sort()

B. np.argsort(arr)

C. np.sort(arr)

D. arr.argsort()

Solution

1. Step 1: Identify the numpy function for argsort

   The function np.argsort() is called with the array as argument: np.argsort(arr).

2. Step 2: Differentiate from other methods

   arr.argsort() is an array method (not the NumPy function), while np.sort(arr) returns sorted values, and arr.sort() sorts in place.

3. Final Answer:

   np.argsort(arr) -> Option B

4. Quick Check:

   Correct function call = np.argsort(arr) [OK]

Hint: Use np.argsort(array), the NumPy function, to get sort indices [OK]

Common Mistakes:

• Using arr.argsort() (array method instead of NumPy function)
• Confusing np.sort() with np.argsort()
• Using arr.sort() which sorts in place and returns None

3. Given the code:

import numpy as np
arr = np.array([40, 10, 30, 20])
indices = np.argsort(arr)
print(indices)

What will be the output?

medium

A. [1 3 2 0]

B. [3 2 1 0]

C. [0 1 2 3]

D. [1 2 3 0]

Solution

1. Step 1: Understand the array and sorting order

   The array is [40, 10, 30, 20]. Sorting it ascending gives [10, 20, 30, 40].

2. Step 2: Find indices that sort the array

   10 is at index 1, 20 at index 3, 30 at index 2, and 40 at index 0. So, indices are [1, 3, 2, 0].

3. Final Answer:

   [1 3 2 0] -> Option A

4. Quick Check:

   Sorted indices = [1 3 2 0] [OK]

Hint: Match sorted values to original indices for argsort output [OK]

Common Mistakes:

• Confusing sorted values with indices
• Reversing the order of indices
• Using sorted array instead of indices

4. What is wrong with this code snippet?

import numpy as np
arr = np.array([3, 1, 2])
indices = arr.argsort()
print(indices)

medium

A. The code will run correctly and print the sorted indices

B. The method argsort() does not exist for numpy arrays

C. The array must be sorted before calling argsort()

D. The print statement is missing parentheses

Solution

1. Step 1: Check if argsort() is a valid numpy array method

   In numpy, arrays do have an argsort() method, so arr.argsort() is valid.

2. Step 2: Verify code correctness

   The code will run and print the indices that sort the array, which are [1, 2, 0].

3. Final Answer:

   The code will run correctly and print the sorted indices -> Option A

4. Quick Check:

   arr.argsort() is valid and works [OK]

Hint: Remember numpy arrays have argsort() method too [OK]

Common Mistakes:

• Assuming argsort() is only in np module, not array method
• Thinking array must be sorted before argsort()
• Confusing Python 2 print syntax with Python 3

5. You have two related numpy arrays:

names = np.array(['apple', 'banana', 'cherry', 'date'])
prices = np.array([3.5, 2.0, 4.0, 1.5])

You want to list the fruit names sorted by their prices in ascending order. Which code snippet correctly achieves this?

hard

A. sorted_names = np.argsort(names)[prices]

B. sorted_names = np.sort(names)[np.argsort(prices)]

C. sorted_names = names[np.argsort(prices)]

D. sorted_names = names[np.sort(prices)]

Solution

1. Step 1: Use np.argsort(prices) to get indices that sort prices

   This returns indices that sort prices ascending.

2. Step 2: Use these indices to reorder names

   Indexing names with these indices sorts names by price.

3. Final Answer:

   sorted_names = names[np.argsort(prices)] -> Option C

4. Quick Check:

   Index names by argsort(prices) to sort by price [OK]

Hint: Index names by argsort of prices to sort related arrays [OK]

Common Mistakes:

• Trying to sort names directly without using indices
• Using np.sort(names) which sorts names alphabetically
• Indexing with sorted prices instead of indices