converting from cell 2 matrix

i have a cell like

a{1,1}={160;45;58;89}% in the form of 1*50char a{2,1}={160;45;58;89}% in the form of 1*50char a{3,1}={160;45;58;89}% in the form of 1*50char a{4,1}={160;45;58;89}% in the form of 1*50char

like this i have a cellsize of 50000 rows(a{50000,1})

what is the best way to convert elements inside cell to matrix form

eg o/p : desired matrix to be

b(1,:)= 160 45 58 89 b(2,:)= 160 45 58 89 b(3,:)= 160 45 58 89 b(4,:)= 160 45 58 89

problem here is the forloop is taking lot of time to convert to serired matrix way

Answer 1

[Jan]

I do not believe that the FOR loop is slow. If you post the code you are using, we will most likely find another problem. Without seeing the code, guessing the reasons is impossible.

testData    = cell(1, 1e6);
testData(:) = {'12; 123; 45; 45; 0.1'};

dummy  = sscanf(testData{1}, '%g;', [1, Inf]);  % To get the width
width  = numel(dummy);
Result = zeros(width, numel(testData));  % Pre-allocate!!!
for ii = 1:numel(testData)
  Result(:, ii) = sscanf(testData{ii}, '%g;', width);
end

Another approach, which consumes more memory - as long as it fits in the processor cache, this could be faster than the loop:

dummy   = sscanf(testData{1}, '%g;', [1, Inf]);  % To get the width
width   = numel(dummy);
dataStr = sprintf('%s;', testData{:});
Result  = sscanf(dataStr, '%g;', [width, Inf]);

It looks like you need to transpose(Result) at the end.

The simple FOR-loop follows the important KISS programming paradigm:

Keep it simple stupid!

When the time required for programming and debugging exceeds the runtime, there is no better concept.

[shaz]

great.....it works.....thanks

i have a matrix of size(1000000*25) but getting some memory issues

for eg: if testData(:) = {'12; 123; 45; 45; 0.1;12; 123; 45; 45; 0.1;12; 123; 45; 45; 0.1'};

how to handle this situation

[Jan]

This matrix requires 1e6*25*8 bytes of free contiguos RAM, otherwise swapping to disk will destroy any performance. This is only 200 MB and should be available on modern computers. Please explain exactly, which line of code creates which kind of "some memory issues".

[shaz]

preallocation of memory is throwing out of memory

Result = zeros(width, numel(testData)); % Pre-allocate!!!

Answer 2

[Azzi Abdelmalek]

b=cell2mat(a)

[Azzi Abdelmalek]

a={'12' '123' '4';'5' '45' '0.1'}
b=cellfun(@str2double,a)

[Azzi Abdelmalek]

what is the problem with

b=cellfun(@str2double,a)

I gave just an example, it should work with your case

[Azzi Abdelmalek]

please give a sample of your data

[Azzi Abdelmalek]

If you mean

a={'12  123  4';'5 45 0.1'}  
b=cell2mat(cellfun(@str2num,a,'un',0))

[Azzi Abdelmalek]

I said provide a small example, a=?

 a={'12  123  4';'5 45 0.1'}

or

a={'12' '123' '4';'5' '45' '0.1'}

or what?

[Azzi Abdelmalek]

b=cell2mat(cellfun(@(x) cellfun(@str2double, regexp(x,';','split')),a,'un',0))

[Jan]

@Azzi: str2double accepts cell strings as input. Therefore I'd omit at least the innermost cellfun. cell2mat contains an expensive loop. Therefore it is most likely naticably faster to create this loop manually, because this will reduce the amount of temporarily used memory.