Why does Meshgrid differ from the for loop here for filling in this tensor?

I’m trying to fill in this 3dimensional tensor efficiently using Meshgrid instead of using for loops, and it seems like the Meshgrid way does not give me what I expect( for loop way).

import numpy as np

def s(x,y,z,w):
    return x + y + z + w + .1*.1


arr_x = [5,6,3,4,3]
arr_y = [5,3,3,10,34]
T_1 = [5,6,10]
nx = len(arr_x)
ny = len(arr_y)

zx, zy, zt = np.meshgrid(arr_x, arr_y, T_1, sparse = True)
Z = s(zx, zy, zt, 0)
print(Z[4])

for i in range(nx):
    for j in range(ny):
        for k in range(3):
            Z[i,j,k] = s(arr_x[i],arr_y[j],T_1[k],0)
print("----------")
print(Z[4])

Why do I get this as an output, I would think they would be the same?

[[44.01 45.01 49.01]
 [45.01 46.01 50.01]
 [42.01 43.01 47.01]
 [43.01 44.01 48.01]
 [42.01 43.01 47.01]]
----------
[[13.01 14.01 18.01]
 [11.01 12.01 16.01]
 [11.01 12.01 16.01]
 [18.01 19.01 23.01]
[42.01 43.01 47.01]]

>Solution :

Add indexing='ij' to the meshgrid call, and all will be well. It’s a difference between matrix indexing and image indexing, where x and y are swapped.