NumPy ndarray.view() Method

NumPy ndarray.view()

The numpy.ndarray.view() method allows you to create a new view of an existing array with a different data type or as a different array subclass. This operation does not create a copy of the data but instead provides a new perspective on the same memory buffer.

Syntax

</>

Copy

ndarray.view([dtype][, type])

Parameters

Parameter	Type	Description
`dtype`	data-type or ndarray sub-class, optional	Specifies the data type for the new view, such as `float32` or `int16`. If omitted, the view retains the original data type.
`type`	Python type, optional	Specifies the type of the returned view (e.g., `numpy.ndarray` or `numpy.matrix`). If omitted, the original type is preserved.

Return Value

Returns a new view of the array with the specified data type or subclass, while sharing the same data buffer as the original array.

Examples

1. Creating a View with a Different Data Type

In this example, we create an integer array and view it as an array of 32-bit floating-point numbers.

</>

Copy

import numpy as np

# Create an array of integers
arr = np.array([1, 2, 3, 4], dtype=np.int32)

# Create a view of the array with a different data type (float32)
view_arr = arr.view(np.float32)

print("Original array:", arr)
print("Data type of original array:", arr.dtype)
print("Viewed array:", view_arr)
print("Data type of viewed array:", view_arr.dtype)

Output:

Original array: [1 2 3 4]
Data type of original array: int32
Viewed array: [1.e-45 3.e-45 4.e-45 6.e-45]
Data type of viewed array: float32

Since the memory layout remains unchanged, interpreting the integer values as floating-point numbers produces seemingly arbitrary small numbers.

2. Creating a View as a Different Array Type

Here, we create a view of an array as a numpy.matrix object.

</>

Copy

import numpy as np

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

# Create a view as a matrix
matrix_view = arr.view(np.matrix)

print("Original array:\n", arr)
print("Type of original array:", type(arr))
print("Viewed matrix:\n", matrix_view)
print("Type of viewed matrix:", type(matrix_view))

Output:

Original array:
 [[1 2]
 [3 4]]
Type of original array: <class 'numpy.ndarray'>
Viewed matrix:
 [[1 2]
 [3 4]]
Type of viewed matrix: <class 'numpy.matrix'>

The data remains the same, but the view is now an instance of numpy.matrix instead of numpy.ndarray.

3. Demonstrating that Views Share Memory

Modifying elements in a view affects the original array because they share the same memory.

</>

Copy

import numpy as np

# Create a NumPy array
arr = np.array([10, 20, 30, 40])

# Create a view of the array
view_arr = arr.view()

# Modify the view
view_arr[0] = 99

print("Original array after modification:", arr)
print("Viewed array:", view_arr)

Output:

Original array after modification: [99 20 30 40]
Viewed array: [99 20 30 40]

The change in the view affects the original array, proving that they share the same data buffer.

4. Viewing an Array with a Different Data Layout

We can use view() to reinterpret the memory layout of an array.

</>

Copy

import numpy as np

# Create an array of 4 integers
arr = np.array([1, 2, 3, 4], dtype=np.int16)

# View the same data as bytes
byte_view = arr.view(np.uint8)

print("Original array:", arr)
print("Data type of original array:", arr.dtype)
print("Viewed array (as bytes):", byte_view)
print("Data type of viewed array:", byte_view.dtype)

Output:

Original array: [1 2 3 4]
Data type of original array: int16
Viewed array (as bytes): [1 0 2 0 3 0 4 0]
Data type of viewed array: uint8

Each 16-bit integer is stored as two consecutive bytes, and view(np.uint8) allows us to interpret them as raw bytes.

TutorialKart

NumPy ndarray.view() Method