How to get cumsum to work on consecutive values and restart if there is a 0 value?

Good Afternoon,

Say I had a a vector

G = [ 0 0 1 1 1 0 0 0 0 1 1 1 1 1 1 1 0 NaN 0 NaN 0 0 0 0]

and I wanted to get a vector

H = [0 0 1 2 3 0 0 0 0 1 2 3 4 5 6 7 0  NaN 0 NaN 0 0 0 0 ]

from this. Is there an easy way to go about this? I essentially just want to restart cumsum everytime there is a break in the sequence.

Thank you for any help/advice/time you can offer!

Answer 1

[Jos (10584)]

Here is a rather easy approach:

G = [ 0 0 1 1 1 0 0 0 0 1 1 1 1 1 1 1 0 NaN 0 NaN 0 9 9 9 9] % data
ix = cumsum([true diff(G)~=0])                               % index the sections
tmp = arrayfun(@(k) cumsum(G(ix==k)), 1:ix(end), 'un', 0)    % cumsum each section
H = cat(2,tmp{:})                                            % concatenate the cells

Answer 2

[Guillaume]

There is no built-in function for this. I think there may be one on the FileExchange, otherwise, you'll have to write your own. A loop is probably the easiest and fastest:

H = zeros(size(G));
count = 0;
for idx = 1:numel(G)
  if G(idx)
    count = count + 1;
  else
    count = 0;
  end
  H(idx) = count;
end

[Birdman]

What if there are 10000 elements in G? Would for loop be faster then?

[Guillaume]

Probably not faster than a mex. But most likely faster than anything else. That's a very simple loop so the jit compiler should optime that result (as long as it's in a function)