Copy array into part of another array in NumPy

Question

I'd like to copy smaller array A into bigger array B, like so: The obvious way to do this is to calculate which part of A would fit into B and copy only this part to the also precalculated part of the destination array B. This is a tedious and error-prone task, especially in higher dimesions. Is there an easier (i.e. without having to calculate all the indices) way to achive this?

A word of clarification - normally one would have to:

in X axis: calculate how many elements of A fit in B part, subtract this from A size, recalculate starting and ending point (indices) in B where they fit, recalculate indices in A. Repeat in second dimension (possibly more). Insert looong slicing formula, which is in no way readable and then be sure not to miss anything in the process. This is what I mean by tedious and error-prone. It can be done, but is there a better way?

Besides A and B what are the givens? How is the region of overlap specified? — unutbu
– unutbu, Commented Oct 8, 2018 at 13:01
@unutbu basically everything needed is given (bounding boxes, sizes, indices where you want to put A in B). What I'm after is only simplifying the whole process, so there is less places to make errors while calculating all the indices. The overlap may be defined as (x_start, y_start, x_end, y_end). — alex
– alex, Commented Oct 8, 2018 at 13:05
@unutbu ok, I might have said too much - not all indices are given, only the starting and ending point. Sorry for not being specific enough. — alex
– alex, Commented Oct 8, 2018 at 13:15
Why do you consider this to be a tedious and error-prone task? Calculating the start and end indices in each dimension for both A and B seems pretty straight forward. — user2653663
– user2653663, Commented Oct 8, 2018 at 13:16

unutbu · Accepted Answer · 2018-10-08 15:04:22Z

3

You can take advantage of NumPy array arithmetic to handle N dimensions the same as you would 1 dimension. Moreover, computing the slices for N dimension can be handled with calls to map(slice, start_indices, end_indices), and once these slices are generated, the actual copying becomes a one-liner: B[B_slices] = A[A_slices]:

import numpy as np

def copy_from(A, B, A_start, B_start, B_end):
    """
    A_start is the index with respect to A of the upper left corner of the overlap
    B_start is the index with respect to B of the upper left corner of the overlap
    B_end is the index of with respect to B of the lower right corner of the overlap
    """
    A_start, B_start, B_end = map(np.asarray, [A_start, B_start, B_end])
    shape = B_end - B_start
    B_slices = tuple(map(slice, B_start, B_end + 1))
    A_slices = tuple(map(slice, A_start, A_start + shape + 1))
    B[B_slices] = A[A_slices]

A = np.zeros((21,15))
B = np.ones((16,15))
A_start = [11, 5]
B_start =  [6, 0]
B_end =  [15, 9]
copy_from(A, B, A_start, B_start, B_end)
print(B)

yields

[[1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.]
 [1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.]
 [1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.]
 [1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.]
 [1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.]
 [1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 1. 1. 1. 1.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 1. 1. 1. 1.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 1. 1. 1. 1.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 1. 1. 1. 1.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 1. 1. 1. 1.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 1. 1. 1. 1.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 1. 1. 1. 1.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 1. 1. 1. 1.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 1. 1. 1. 1.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 1. 1. 1. 1.]]

edited Oct 8, 2018 at 15:04

answered Oct 8, 2018 at 14:04

unutbu

886k197 gold badges1.9k silver badges1.7k bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

norok2 Over a year ago

You should change list(map(... to tuple(map(... or you'll get a

FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.

with newer versions of NumPy like 1.15.2.

Collectives™ on Stack Overflow

Copy array into part of another array in NumPy

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related