Making Copies Of An Numpy.ndarray Object

Create one or more references to an existing ndarray:

 

  • To create one are more references to an existing numpy.ndarray object the assignement operator can be used.

 

  • Once another reference of an existing array object is created any change made to the array through new reference affects the original contents of the array.

 

  • The following example creates three references to an existing numpy.ndarray object.

 

Example: Creating multiple references to an existing ndarray object

import numpy as np

 

# Create an ndarray object

reference0 = np.arange(5)

print("ndarray object printed through reference0:")

print(reference0)

 

# Make two more references to the same ndarray object

reference1 = reference2 = reference0

 

print("ndarray object printed through reference1:")

print(reference1)

 

print("ndarray object printed through reference2:")

print(reference2)

 

# Change ndarray contents through the reference1

reference1[0] = 10

print("ndarray contents changed through reference1 and printed through reference0:")

print(reference0)

 

# Change the contents of the ndarray through the second reference2

reference2[1] = 20

print("ndarray contents changed through reference2 and printed through reference1:")

print(reference1)

 

  • In the above Python program, the references reference0, reference1 and reference2 all point to the same ndarray object.
  • Changes made through any of references reflect in the original ndarray object.

 

Output:

ndarray object printed through reference0:

[0 1 2 3 4]

ndarray object printed through reference1:

[0 1 2 3 4]

ndarray object printed through reference2:

[0 1 2 3 4]

ndarray contents changed through reference1 and printed through reference0:

[10  1  2  3  4]

ndarray contents changed through reference2 and printed through reference1:

[10 20  2  3  4]

 

Create a deep copy of an ndarray object using : the slicing operator:

  • A deep copy of ndarray elements can be created using the slicing operator.

 

  • For this to work there should be an already existing target array with the same shape as the source array.

 

Example:

import numpy as np

 

# Define a class

class Container:

    myVal = 0

   

    def __init__(self, val):

        self.myVal = val

   

    def __repr__(self):   

        return "%d"%(self.myVal)

 

    def __str__(self):   

        return "%d"%(self.myVal)

 

# Create container instances

c0 = Container(10)

c1 = Container(20)

c2 = Container(30)

       

# Create an ndarray of containers

array1 = np.ndarray(3, dtype='object')

 

array1[0] = c0

array1[1] = c1

array1[2] = c2

 

print("array1 contents:")

print(array1)

 

array2      = np.ndarray(3, dtype='object')

array2[:]   = array1

 

array2[0]   = 1

array2[1]   = 2

array2[2]   = 3

 

print("array2 contents after deep copy and modifications to it:")

print(array2)

 

print("array1 contents after deep copy and modifications to array2:")

print(array1)

 

 

Output:

array1 contents:

[10 20 30]

array2 contents after deep copy and modifications to it:

[1 2 3]

array1 contents after deep copy and modifications to array2:

[10 20 30]

 

Create a deep copy of the numpy.ndarray object using the copy() method:

  • The copy method returns a deep copy of the numpy.ndarray object.

 

  • While using the copy() method the memory layout can specified as any of the following using the parameter order:
    • ’C’: ‘C’ Style or Row Major – meaning the row elements of the array are stored in subsequent array indices.
    • ’F ’: Fortran Style or Column Major – meaning the column elements of the array are stored in subsequent array indices.
    • ’A’: If the source array is Fortran style the same style will be applied. Else ‘C’ style will be applied.
    • ’K ’: A memory layout that matches close to the memory layout of the source array.

Example:

import numpy as np

 

# Define a class

class Container:

    myVal = 0

   

    def __init__(self, val):

        self.myVal = val

   

    def __repr__(self):   

        return "%d"%(self.myVal)

 

    def __str__(self):   

        return "%d"%(self.myVal)

 

# Create Container instances

c0 = Container(10)

c1 = Container(20)

c2 = Container(30)

       

# Create an ndarray of Containers

array1 = np.ndarray(3, dtype='object')

 

array1[0] = c0

array1[1] = c1

array1[2] = c2

 

print("Array1 contents:")

print(array1)

 

array2      = array1.copy()

 

print("array1:")

print(array2)

 

print("array2:")

print(array1)

 

array2[0] = 9

array2[1] = 8

array2[2] = 7

 

print("Array2 contents after modifying its elements:")

print(array2)

 

print("Array1 contents after modifying array2:")

print(array1)

 

 

Output:

Array1 contents:

[10 20 30]

array1:

[10 20 30]

array2:

[10 20 30]

Array2 contents after modifying its elements:

[9 8 7]

Array1 contents after modifying array2:

[10 20 30]

 


Copyright 2022 © pythontic.com