NumPy副本和视图用法详解

输入数组的副本实际上存储在其他某个位置, 并且返回存储在该特定位置的内容, 这是输入数组的副本, 而在视图的情况下, 返回同一内存位置的不同视图。

在本教程的这一部分中, 我们将考虑从某个内存位置生成不同副本和视图的方式。

数组分配

将numpy数组分配给另一个数组不会直接复制原始数组, 而是使另一个数组具有相同的内容和相同的id。它表示对原始数组的引用。在此引用上所做的更改也会反映在原始数组中。

id()函数返回数组的通用标识符, 类似于C中的指针。

考虑以下示例。

例子

import numpy as np

a = np.array([[1, 2, 3, 4], [9, 0, 2, 3], [1, 2, 3, 19]])

print("Original Array:\n", a)

print("\nID of array a:", id(a))

b = a 

print("\nmaking copy of the array a")

print("\nID of b:", id(b))

b.shape = 4, 3;

print("\nChanges on b also reflect to a:")
print(a)

输出

Original Array:
 [[ 1  2  3  4]
 [ 9  0  2  3]
 [ 1  2  3 19]]

ID of array a: 139663602288640

making copy of the array a

ID of b: 139663602288640

Changes on b also reflect to a:
[[ 1  2  3]
 [ 4  9  0]
 [ 2  3  1]
 [ 2  3 19]]

ndarray.view()方法

ndarray.view()方法返回新数组对象, 该对象包含与原始数组相同的内容。由于它是一个新的数组对象, 因此对该对象所做的更改不会反映原始数组。

考虑以下示例。

例子

import numpy as np

a = np.array([[1, 2, 3, 4], [9, 0, 2, 3], [1, 2, 3, 19]])

print("Original Array:\n", a)

print("\nID of array a:", id(a))

b = a.view()

print("\nID of b:", id(b))

print("\nprinting the view b")
print(b)

b.shape = 4, 3;

print("\nChanges made to the view b do not reflect a")
print("\nOriginal array \n", a)
print("\nview\n", b)

输出

Original Array:
 [[ 1  2  3  4]
 [ 9  0  2  3]
 [ 1  2  3 19]]

ID of array a: 140280414447456

ID of b: 140280287000656

printing the view b
[[ 1  2  3  4]
 [ 9  0  2  3]
 [ 1  2  3 19]]

Changes made to the view b do not reflect a

Original array 
 [[ 1  2  3  4]
 [ 9  0  2  3]
 [ 1  2  3 19]]

view
 [[ 1  2  3]
 [ 4  9  0]
 [ 2  3  1]
 [ 2  3 19]]

ndarray.copy()方法

它返回原始数组的深层副本, 该副本不与原始数组共享任何内存。对原始数组的深层副本所做的修改不会反映原始数组。

考虑以下示例。

例子

import numpy as np

a = np.array([[1, 2, 3, 4], [9, 0, 2, 3], [1, 2, 3, 19]])

print("Original Array:\n", a)

print("\nID of array a:", id(a))

b = a.copy()

print("\nID of b:", id(b))

print("\nprinting the deep copy b")
print(b)

b.shape = 4, 3;

print("\nChanges made to the copy b do not reflect a")
print("\nOriginal array \n", a)
print("\nCopy\n", b)

输出

Original Array:
 [[ 1  2  3  4]
 [ 9  0  2  3]
 [ 1  2  3 19]]

ID of array a: 139895697586176

ID of b: 139895570139296

printing the deep copy b
[[ 1  2  3  4]
 [ 9  0  2  3]
 [ 1  2  3 19]]

Changes made to the copy b do not reflect a

Original array 
 [[ 1  2  3  4]
 [ 9  0  2  3]
 [ 1  2  3 19]]

Copy
 [[ 1  2  3]
 [ 4  9  0]
 [ 2  3  1]
 [ 2  3 19]]