Check if all sides of a multidimensional numpy array are arrays of zeros

Here's how you can do it:

assert(all(np.all(np.take(x, index, axis=axis) == 0)
           for axis in range(x.ndim)
           for index in (0, -1)))

np.take does the same thing as "fancy" indexing.

Here's an answer that actually examines the parts of the array you're interested in, and doesn't waste time constructing a mask the size of the whole array. There's a Python-level loop, but it's short, with iterations proportional to the number of dimensions instead of the array's size.

def all_borders_zero(array):
    if not array.ndim:
        raise ValueError("0-dimensional arrays not supported")
    for dim in range(array.ndim):
        view = numpy.moveaxis(array, dim, 0)
        if not (view[0] == 0).all():
            return False
        if not (view[-1] == 0).all():
            return False
    return True

I reshaped the array and then iterated through it. Unfortunately, my answer assumes you have at least three dimensions and will error out for normal matrices, you would have to add a special clause for 1 & 2 dimensional shaped arrays. In addition, this will be slow so there are likely better solutions.

x = np.array(
        [
            [
                [0 , 1, 1, 0],
                [0 , 2, 3, 0],
                [0 , 4, 5, 0]
            ],
            [
                [0 , 6, 7, 0],
                [0 , 7, 8, 0],
                [0 , 9, 5, 0]
            ]
        ])

xx = np.array(
        [
            [
                [0 , 0, 0, 0],
                [0 , 2, 3, 0],
                [0 , 0, 0, 0]
            ],
            [
                [0 , 0, 0, 0],
                [0 , 7, 8, 0],
                [0 , 0, 0, 0]
            ]
        ])

def check_edges(x):

    idx = x.shape
    chunk = np.prod(idx[:-2])
    x = x.reshape((chunk*idx[-2], idx[-1]))
    for block in range(chunk):
        z = x[block*idx[-2]:(block+1)*idx[-2], :]
        if not np.all(z[:, 0] == 0):
            return False
        if not np.all(z[:, -1] == 0):
            return False
        if not np.all(z[0, :] == 0):
            return False
        if not np.all(z[-1, :] == 0):
            return False

    return True

Which will produce

>>> False
>>> True

Basically I stack all the dimensions on top of each other and then look through them to check their edges.

You can make use of slice and boolean masking to get the job done:

def get_borders(arr):
    s=tuple(slice(1,i-1) for i in a.shape)
    mask = np.ones(arr.shape, dtype=bool)
    mask[s] = False
    return(arr[mask])

This function first shapes the "core" of the array into the tuple s, and then builds a mask that shows True only for the bordering points. Boolean indexing then delivers the border points.

Working example:

a = np.arange(16).reshape((4,4))

print(a)
array([[ 0,  1,  2,  3],
       [ 4,  5,  6,  7],
       [ 8,  9, 10, 11],
       [12, 13, 14, 15]])

borders = get_borders(a)
print(borders)
array([ 0,  1,  2,  3,  4,  7,  8, 11, 12, 13, 14, 15])

Then, np.all(borders==0) will give you the desired information.

Note: this breaks for one-dimensional arrays, though I consider those an edge case. You're probably better off just checking the two points in question there

maybe the ellipsis operator is what you are looking for, which will work for many dimensions:

import numpy as np

# data
x = np.random.rand(2, 5, 5)
x[..., 0:, 0] = 0
x[..., 0, 0:] = 0
x[..., 0:, -1] = 0
x[..., -1, 0:] = 0

test = np.all(
    [
        np.all(x[..., 0:, 0] == 0),
        np.all(x[..., 0, 0:] == 0),
        np.all(x[..., 0:, -1] == 0),
        np.all(x[..., -1, 0:] == 0),
    ]
)

print(test)