numpy. ndarray swaps the row and column method for the multidimensional array of matrix

As shown below:

>> import numpy as np
>> P = np.eye(3)
>> P
array([[ 1., 0., 0.],
    [ 0., 1., 0.],
    [ 0., 0., 1.]])

Swap lines 0 and 2:

>> P[[0, 2], :] = P[[2, 0], :]
    # P[(0, 2), :] = P[(2, 0), :]
>> P
array([[ 0., 0., 1.],
    [ 0., 1., 0.],
    [ 1., 0., 0.]])

Then swap column 1 and column 3:

>> P[:, [0, 2]] = P[:, [2, 0]]
>> P
array([[ 1., 0., 0.],
    [ 0., 1., 0.],
    [ 0., 0., 1.]])

Note the following is the wrong approach:

>> P[0, :], P[2, :] = P[2, :], P[0, :]
>> P
array([[ 0., 0., 1.],
    [ 0., 1., 0.],
    [ 0., 0., 1.]])

    #  Not to be wordy, and the meaning of the representation is not exchange