replace comma with dot in large array

Please help, I'm new to matlab coding. I have a array like this

data={'250,1000'    '-6,52796'       '322,683'    '1,50367' 
  '250,0500'    '-6,54959'       '322,695'    '1,50379' 
  '250,0000'    '-6,63267'       '322,711'    '1,50395'}

except that it is very large (4500 lines). I want to replace all the commas with dots and convert to a data matrix. So far I have this:

data = strrep(data, ',', '.');%replace ´,´ with ´.´
data=cellfun(@str2num,data);

which can do the trick but it is very slow. Any suggestions to speed up the process?

Answer 1

[Guillaume]

Well, the slow bit is probably your cellfun(@str2num). Replacing that by:

data = str2double(data);

would probably be faster. The strrep step is already as fast as it can be in matlab.

If you want faster, you need to modify your import process but if I recall correctly matlab is not very good as coping with ',' decimal separator.

[JB]

the code: dataS1 = str2double(strrep(dataS1, ',', '.'));

did the job in 0.21 sec. Thanks a lot. My old code was 3.6 sec...

Answer 2

[Image Analyst]

4500 is not very big. I ran this code

% Create sample data
n = 4500
data = cell(1, n);
for k = 1 : n
  data{k} = '123,456';
end

% Now do the replacement.
tic;
data = strrep(data, ',', '.');   % Replace ´,´ with ´.´
toc;
% celldisp(data);
fprintf('Done!\n');

and got this:

Elapsed time is 0.002293 seconds.

Two milliseconds seems fast to me. Exactly how fast do you need it?

[JB]

Hi Image analyst. It sure is faster than my code, but it does not convert to a data matrix. I might have made a mistake...

[Image Analyst]

"data matrix" is so ambiguous that we don't know what it is. ANY kind of matrix in MATLAB could be considered a data matrix, from a character matrix to a cell array to a table to a regular double matrix. Your "data" is a cell array and so is, in fact, already a data matrix. If you want a vector of doubles, you can do

d = str2double(data);

as Guillaume first showed. Did you ever figure out why yours took so long while my running of your code took only 2 ms?