Indexing returns views not copies in NumPy

Introduction

When you select parts of data in numpy arrays, it often gives you a view, not a new copy. This means changes to the view affect the original data. This helps save memory and time.

You want to quickly look at or change a part of a big dataset without copying it.

You want to save memory when working with large arrays.

You want to update parts of your data and see the changes reflected everywhere.

You want to avoid slow operations caused by copying data unnecessarily.

Syntax

NumPy

view = original_array[start:stop]

Indexing with slices usually returns a view, not a copy.

Modifying the view changes the original array too.

Examples

Changing the view changes the original array because the view shares the same data.

NumPy

import numpy as np
arr = np.array([1, 2, 3, 4, 5])
view = arr[1:4]
view[0] = 20
print(arr)

The slice from index 2 to end is a view. Changing it updates the original array.

NumPy

import numpy as np
arr = np.array([10, 20, 30, 40])
slice_view = arr[2:]
slice_view[1] = 99
print(arr)

Using .copy() creates a new array. Changing it does not affect the original.

NumPy

import numpy as np
arr = np.array([5, 6, 7, 8])
copy = arr[1:3].copy()
copy[0] = 100
print(arr)

Sample Program

This program shows that slicing returns a view that changes the original array. Using .copy() makes a separate array that does not affect the original.

NumPy

import numpy as np

# Create an array
arr = np.array([1, 2, 3, 4, 5])

# Get a view by slicing
view = arr[1:4]

# Change the first element of the view
view[0] = 20

# Print original array to see the change
print(arr)

# Now create a copy
copy = arr[1:4].copy()

# Change the first element of the copy
copy[0] = 99

# Print original array again to show no change
print(arr)

Important Notes

Not all indexing returns views. Fancy indexing (like using lists or boolean arrays) returns copies.

Always use .copy() if you want to avoid changing the original data.

Summary

Indexing slices usually return views that share data with the original array.

Changing a view changes the original array.

Use .copy() to make a separate array if needed.