How to avoid linear indexing in operations involving matrices of different sizes

If I carry out an operation of matrices with different sizes using indexing, the end result tends to be a column matrix with linear indexing. For example:

A = rand(3,3);
B = rand(3,2);
idx = logical([0 1 1; 0 1 1; 0 1 1]);

If I want to add A with idx indices to B, the only way I can seem to make this work is if I do:

C = A(idx) + B(:)

C = 6×1

1.5283 1.5581 1.0464 1.4741 0.5704 0.2374

Is there any way to carry out the above operation and end up with a matrix the same shape as B? My initial attempt was to simply do C = A(idx) + B.

Answer 1

[Stephen23]

RESHAPE is very efficient, because no data gets moved in memory:

A = rand(3,3);
B = rand(3,2);
idx = logical([0,1,1;0,1,1;0,1,1]);
C = reshape(A(idx),size(B)) + B

C = 3×2

1.6889 0.4419 0.7161 0.1917 1.4049 0.7527

Answer 2

[Matt J]

A = rand(3,3);
B = rand(3,2);
idx = logical([0 1 1; 0 1 1; 0 1 1]);


C=B;
C(:)=A(idx)+B(:)

C = 3×2

1.1539 0.5087 1.1585 1.0807 1.4266 0.7590