Split cell array taking into account difference between elements

Hello! I have a cell array which contains indices from another cell array.Lets say for example

TES={[1;2;3;4;5];[12;13;14];[1;2;3;4;8;9;10;22;23;24];[4;5;6;8;9];[5;6;7]};

Where the indices have no jump at their sequence (cases 1,2 & 5) I want to keep the cell as it is. Where the jump is smaller than 2 I also want to keep the cell as it is(cell 4) But where there is a jump(or more) greater than 2 I need to split each cell into other cells.Their Number should be (jumbs+1).

The desirable result is

TES_new={[1;2;3;4;5];[12;13;14];[1;2;3;4];[8;9;10];[22;23;24];[4;5;6;8;9];[5;6;7]};

Answer 1

[Guillaume]

A fairly simple way of doing it:

TES={[1;2;3;4;5];[12;13;14];[1;2;3;4;8;9;10;22;23;24];[4;5;6;8;9];[5;6;7]};

runlengths = cellfun(@(m) diff([0; find(diff(m) > 2); numel(m)]), TES, 'UniformOutput', false);
splits = cellfun(@(m, runs) mat2cell(m, runs), TES, runlengths, 'UniformOutput', false);
TES_new = vertcat(splits{:})

You can combine both cellfun for a little bit more speed at the expense of readability:

splits = cellfun(@(m) mat2cell(m, diff([0; find(diff(m) > 2); numel(m)])), TES, 'UniformOutput', false);
TES_new = vertcat(splits{:})

[Dimitris Kokkinos]

I prefer the first one which is simpler for a rookie.Thanks.

[Dimitris Kokkinos]

If we had another column at cell array TES{i,2}= n where n is an index referring to another cell array...

Can we preserve the this index to our split cell?

For example if it was TES{3,2}=6 is it possible to get TES_new{3,2}=6 TES_new{4,2}=6 TES_new{5,2}=6

[Guillaume]

If I understood correctly, this is what you want:

TES={[1;2;3;4;5] 10;[12;13;14] 2;[1;2;3;4;8;9;10;22;23;24] 6;[4;5;6;8;9] 4;[5;6;7] 8} %for example
runlengths = cellfun(@(m) diff([0; find(diff(m) > 2); numel(m)]), TES(:, 1), 'UniformOutput', false);
splits = cellfun(@(m, runs) mat2cell(m, runs), TES(:, 1), runlengths, 'UniformOutput', false);
indices = cellfun(@(i, runs) repmat(i, numel(runs), 1), TES(:, 2), runlengths, 'UniformOutput', false);
TES_new = [vertcat(splits{:}) num2cell(vertcat(indices{:}))]

[Dimitris Kokkinos]

Thanks a lot!!!It is exactly what I was looking for.

Answer 2

[Thorsten]

This is somewhat similar to Star Strider's solution but with an extra index ind_new to get rid of the subarrays and with a copy of those arrays that need not to be split to the new cell:

 TES={[1;2;3;4;5];[12;13;14];[1;2;3;4;8;9;10;22;23;24];[4;5;6;8;9];[5;6;7]};        
i_new = 1;
for i=1:numel(TES)
  split = find(diff(TES{i}) > 2);
  if isempty(split)
    TES_new{i_new} = TES{i};
    i_new = i_new + 1;
  else
    ind = [0 split' numel(TES{i})];
    for ii = 1:numel(ind2)-1
      TES_new{i_new} = TES{i}(ind(ii)+1:ind(ii+1));
      i_new = i_new + 1;
    end
  end
end

 for i=1:numel(TES_new)
  disp(TES_new{i})
end

Answer 3

[Andrei Bobrov]

n = cellfun(@numel,TES);
ii = accumarray(cumsum([1;n]),1);
ii = ii(1:end-1);
T = cat(1,TES{:})
jj = [1; abs(diff(T))>2];

out = accumarray(cumsum([jj | ii]),T,[],@(x){x})