Vectorised cell search and logical indexing

Hello, I would like to know if there is a vectorised/cellfun/better approach to the loop below:

% Find Key 4 in block_new{:}(1,1)
Key4 = '=x';
index4 = cellfun(@(x)regexp(x, Key4), block_new, 'UniformOutput', false);
Logic = false(length(index4),1);
for i = 1 : length(index4)
    Logic(i) = ~isempty(cell2mat(index4{i}(1,1)));
end
index4   =   Logic;
block_set = block_new(:,index4);

Where the cells in the 1 X m cell array, block_new are in themselves cell arrays of various sizes.

I have already tried the following:

find(~cellfun(@isempty,index4{:}(1,1)))

Thanks in advance.

[Jos (10584)]

what does block_new contain? Is it a cell array of strings?

[Joshua]

Yes. A typical example, let's say block_new{i} would contain:

block_new{i}(1,1) -> String
block_new{i}(1,2:n} -> String of comma delimited doubles

[Jan]

~isempty(cell2mat(index4{i}(1,1)));

If the output of cell2mat is empty, the cell must have been empty already. What about:

~isempty(index4{i}{1})

instead?

Please post a relevant input array.

Answer 1

[Jos (10584)]

regex can work on a cell array of strings directly:

block_new = {'abcd=x','abcd','a=xbc=x'}
index4 = regexp('block_new','=x')

[Joshua]

Apparently they are not all strings..

regexp(block_new,'=x')
Error using regexp
All cells must be char row vectors.

[Jos (10584)]

Some cells are cell array of strings themselves? Kind of complicated data structure ...

[Cedric]

As mentioned by Jan, it would be simpler if you would attach a sample data set. Also, if you could describe in words what you are trying to achieve, that could help because you are doing a lot of operations that seem unnecessary (but that may in fact be necessary depending the final purpose).