numpy array slicing produces an error indexing arrays could not be broadcast together-CodePudding

I have an np.array :

a = np.array([x for x in range(10) for y in range(10)]).reshape(10, 10)

I want to get the 3rd and the 6th rows columns 4,7,10 - marked in green.

I tried:

a[[2,5]] # gives me the rows  I want
a[[2,5], [3,6,10]]  # produces an an error

IndexError: shape mismatch: indexing arrays could not be broadcast together with shapes (2,) (3,)

In the end the result should look like:

[[2,2,2],
[5,5,5]]

Where is my mistake?

CodePudding user response：

the first index list must the (2,1) shape (or list equivalent):

In [31]: a[[[2], [5]], [3, 6, 9]]
Out[31]: 
array([[2, 2, 2],
       [5, 5, 5]])

Do you understand what the error message means by broadcasting?

For simple addition, a (2,1) array broadcasts with a (3,) to produce a (2,3) result:

In [32]: I, J = np.array((2, 5)), np.array((3, 6, 9))
In [33]: I, J
Out[33]: (array([2, 5]), array([3, 6, 9]))
In [34]: I   J
Traceback (most recent call last):
  Input In [34] in <module>
    I   J
ValueError: operands could not be broadcast together with shapes (2,) (3,) 

In [35]: I[:, None]   J
Out[35]: 
array([[ 5,  8, 11],
       [ 8, 11, 14]])

The same idea applies to indexing with several arrays.

Your a can be created with the same logic:

In [38]: np.arange(10)[:, None]   np.zeros(10,int)
Out[38]: 
array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
       [2, 2, 2, 2, 2, 2, 2, 2, 2, 2],
       [3, 3, 3, 3, 3, 3, 3, 3, 3, 3],
       [4, 4, 4, 4, 4, 4, 4, 4, 4, 4],
       [5, 5, 5, 5, 5, 5, 5, 5, 5, 5],
       [6, 6, 6, 6, 6, 6, 6, 6, 6, 6],
       [7, 7, 7, 7, 7, 7, 7, 7, 7, 7],
       [8, 8, 8, 8, 8, 8, 8, 8, 8, 8],
       [9, 9, 9, 9, 9, 9, 9, 9, 9, 9]])

With 2 arrays (lists) of matching size, the effect is to select a "diagonal", or 1d array of values (rather than the block that you were trying to get):

In [39]: a[[2, 3], [5, 6]]
Out[39]: array([2, 3])
In [40]: a[2, 5], a[3, 6]
Out[40]: (2, 3)

CodePudding user response：

I believe this link will explain everything since the question is the same and was given great answer: Selecting specific rows and columns from NumPy array