[FIXED] Possibly NumPy array bug for element access?

Issue

I was trying to modify a numpy array after construction using the code below. (yes, I know there are functions to do that like np.random.randint(100, size=(2,3))). It looks like there's no problem of reading using the Python style syntax, but I cannot make changes to the ndarray.

Is this a possible bug? Or should we call it a feature like most developers would do?

import random
import numpy as np

def print_data(arr):
    for row in arr:
        for col in row:
            print(f'{col}\t', end=' ')
        print()

# Both creation methods have the same bug     
##a = np.ndarray(shape=(2,3),dtype=int)
a = np.zeros((2,3), dtype=int)
print_data(a)
print()
for row in a:
    for cell in row:
        # cell was printed out
        cell = random.randrange(99)
        print(cell, end=' ')
    # row was not changed
    print(row)

print()
print_data(a)
print()

for row in a:
    for i in range(len(row)):
        row[i] = random.randrange(99)
        print(row[i], end=' ')
    print(row)
print()
print_data(a)

The output:

1    0   -447458960  
32763    -447580336  32763   

6 2 77 [         1          0 -447458960]
50 20 75 [     32763 -447580336      32763]

1    0   -447458960  
32763    -447580336  32763   

95 61 3 [95 61  3]
76 23 82 [76 23 82]

95   61  3   
76   23  82

As suggested by the output, the first double loop prints out the random number without any problem. However, the matrix element was not modified. In the second code block, the a[i] worked. I also tested that it worked if a[i,j] was used in a double loop.

Any idea or a link to some detailed explanation would be appreciated.

Solution

It's working as intended:

• To change a single cell in a 2D numpy array, use a[i,j] = .... This is an item assignment, which for numpy arrays changes a single cell within the array object.

• cell = ... only changes the variable cell. Think of a variable as a reference to an object; a variable assingment such as cell = ... and for cell in ...: makes the variable refer to a different object.

Answered By - pts