Slicing with numpy in a for loop-CodePudding

I have to work with x, x[:, :-1], x[:, :-2], etc. How to do this in a for loop?

for i in range(100):
    for j in range(100):
        y = x[:-i, :-j]

This works as desired if i and j are different than 0. But it fails if i or j is 0:

x[:, :-0]

is not x[:, :].

What is the standard Numpy way to do this?

Is there better than x[:x.shape[0]-i, :x.shape[1]-j]?

TL;DR:

y[:-1] is "all elements except the last one",
y[:-2] is "all elements except the two last ones",
but y[:-0] is NOT "all elements"

CodePudding user response：

One of possible solutions is to revert the order of iteration in both loops (start from the last index, in decreasing order).

Then you can refer to a range of elements starting from 0 (no number before the colon), ending at the index expressed by the loop control variable (after the colon).

If you want "original" values from your original loops, compute them as 10 - i and 10 - j.

To sum up, try the below code:

n = x.shape[0]
for i in range(n, 0, -1):
    for j in range(n, 0, -1):
        i2 = 10 - i
        j2 = 10 - j
        print(f'{i2}, {j2},  {i}, {j}\n{x[:i, :j]}')

Of course, for test purpose, use a smaller array, e.g. 10 * 10.

CodePudding user response：

One possible option is to use the array-length len inside:

y[:len(y)-2] #is all except the last two
y[:len(y)-1] #is all except the last one
y[:len(y)-0] #is all

CodePudding user response：

This is essentially the same as the prior answer but I would just create a reverted version of the array. It could be memory-inefficient, but slightly more readable.

x = np.arange(0, 25).reshape(5,5)
x_r = x[::-1, ::-1]
for i in range(5):
    for j in range(5):
        y = x_r[i:, j:]