How to create a vector from two vectors?

I have two vectors with the same length:

A = [2 4 5 4 9 8 1 6];

B = [3 6 2 5 4 4 3 5];

Each element in vector B defines how many times the corresponding element in vector A duplicates itself. For example, here I want the final vector C to have 3 twos, 6 fours, 2 fives, 5 fours, 4 nines, 4 eights, 3 ones, and 5 sixes:

C = [2 2 2, 4 4 4 4 4 4, 5 5, 4 4 4 4 4, 9 9 9 9, 8 8 8 8, 1 1 1, 6 6 6 6 6];

I can do it with a for-loop. However in practice, vectors A and B are long. So is there a way to do this without a loop.

Thanks!

Answer 1

[Stephen23]

If you have MATLAB 2015a or more recent then use repelem:

 repelem(A,B)

otherwise:

 cell2mat(arrayfun(@(a,b)repmat(a,1,b),A,B,'Uni',0))

although a loop may well be faster.

[DS Chen]

Is repelem(A,B) in 2015a faster than for-loop?

I'm using 2013b. I guess I'll use for-loop then.

Thank you!

[Stephen23]

Important: do NOT expand the array in the loop! Use a preallocated cell array:

>> C = cell(size(A));
>> for k = 1:numel(C), C{k} = repmat(A(k),1,B(k)); end
>> cell2mat(C)

Or a preallocated numeric array (this is faster):

 >> X = cumsum([0,B]);
 >> D = NaN(1,X(end));
 >> for k = 1:numel(A), D(1+X(k):X(k+1)) = A(k); end

[DS Chen]

Good Point!

Using preallocated cell array is 40 times faster than expanding array in the loop.

I also test the repelem(A, B) on the other machine. repelem(A,B) is indeed the fastest! It's at least 10 times faster than for-loop. I hope I can use it later:)