How can I do cumulative sum separately in one array?

For example, there is an array

A=[1 2 0 3 4 0 5 6 7 0 8 9]

Every time there is an zero in A, i.e. restart signal, I want the cumulative sum reset as 0 and restart the sum. The result I want from A is array

B=[1 3 0 3 7 0 5 11 18 0 8 17]

Please help me :) Cheers.

Answer 1

[Image Analyst]

Did you think of using a simple, intuitive, and fast for loop:

A=[1 2 0 3 4 0 5 6 7 0 8 9]
theSum = 0;
for k = 1 : length(A)
  if A(k) == 0
    theSum = 0;
  end
  theSum = theSum + A(k);
  B(k) = theSum;
end
% Show in command window
B

[the cyclist]

You'll want to preallocate B if A is long. Put

B = zeros(size(A));

ahead of the algorithm.

[the cyclist]

In limited testing, this simple solution is fastest, by a large margin (assuming you put in the preallocation).

Answer 2

[Guillaume]

A = [1 2 0 3 4 0 5 6 7 0 8 9];
subarraylengths = diff([0 find(~A)-1 numel(A)]);
subarrays = mat2cell(A, 1, subarraylengths);
cumsubarrays = cellfun(@cumsum, subarrays, 'UniformOutput', false);
B = [cumsubarrays{:}]

or use a loop

Answer 3

[the cyclist]

Here's another algorithm. The best one might depend on the size of A.

A = [1 2 0 3 4 0 5 6 7 0 8 9];

% Append start and end zeros temporarily
B = [0 A 0];

% Find the zeros (including "artificial" zeros tagged at the end).
zeroLocations = find(B==0);

numberZeros = numel(zeroLocations);

for ii = 1:numberZeros-1
    segmentIndex = zeroLocations(ii)+1:zeroLocations(ii+1)-1;
    B(segmentIndex) = cumsum(B(segmentIndex));
end

% Remove the temporary zeros
B = B(2:end-1);