使用数组为多维 numpy 数组定义索引

Question

I have a multidimensional Numpy array; let's say it's

 myArray = array([[[ 0,  1,  2],
                   [ 3,  4,  5],
                   [ 6,  7,  8]],

                  [[ 9, 10, 11],
                   [12, 13, 14],
                   [15, 16, 17]],

                  [[18, 19, 20],
                   [21, 22, 23],
                   [24, 25, 26]]])

I know that running myArray[1,1,1] , for instance, will return 13. However, I want to define indx = [1,1,1] then call something to the effect of myArray[indx] . However, this does some other multidimensional indexing stuff.

I have also tried myArray[*indx] but that understandably throws a syntax error.

Currently my very ugly workaround is to define

def array_as_indices(array, matrix):
    st = ''
    for i in array:
        st += '%s,' % i
    st = st[:-1]

    return matrix[eval(st)]

which works but is quite inelegant and presumably slow.

Is there a more pythonic way to do what I'm looking for?

Answer 1

This is a duplicate of Unpacking tuples/arrays/lists as indices for Numpy Arrays , but you can just create a tuple

import numpy as np


def main():
    my_array = np.array(
        [
            [[0, 1, 2], [3, 4, 5], [6, 7, 8]],
            [[9, 10, 11], [12, 13, 14], [15, 16, 17]],
            [[18, 19, 20], [21, 22, 23], [24, 25, 26]],
        ]
    )
    print(f"my_array[1,1,1]: {my_array[1,1,1]}")
    indx = (1, 1, 1)
    print(f"my_array[indx]: {my_array[indx]}")


if __name__ == "__main__":
    main()

will give

my_array[1,1,1]: 13
my_array[indx]: 13

Answer 2

The indices of a numpy array are addressed by tuples, not lists. Use indx = (1, 1, 1) .

As an extension, if you want to call the indices (1, 1, 1) and (2, 2, 2), you can use

>>> indx = ([1, 2], [1, 2], [1, 2])
>>> x[indx]
array([13, 26])

---

The rationale behind the behavior with lists is that numpy treats lists sequentially, so

>>> indx = [1, 1, 1]
>>> x[indx]
array([[[ 9, 10, 11],
        [12, 13, 14],
        [15, 16, 17]],
       [[ 9, 10, 11],
        [12, 13, 14],
        [15, 16, 17]],
       [[ 9, 10, 11],
        [12, 13, 14],
        [15, 16, 17]]])

It returns a list of three elements, each equal to x[1].

Answer 3

I have a multidimensional Numpy array; let's say it's

 myArray = array([[[ 0,  1,  2],
                   [ 3,  4,  5],
                   [ 6,  7,  8]],

                  [[ 9, 10, 11],
                   [12, 13, 14],
                   [15, 16, 17]],

                  [[18, 19, 20],
                   [21, 22, 23],
                   [24, 25, 26]]])

I know that running myArray[1,1,1] , for instance, will return 13. However, I want to define indx = [1,1,1] then call something to the effect of myArray[indx] . However, this does some other multidimensional indexing stuff.

I have also tried myArray[*indx] but that understandably throws a syntax error.

Currently my very ugly workaround is to define

def array_as_indices(array, matrix):
    st = ''
    for i in array:
        st += '%s,' % i
    st = st[:-1]

    return matrix[eval(st)]

which works but is quite inelegant and presumably slow.

Is there a more pythonic way to do what I'm looking for?