Optimizing a 'for' loop

Question

djr 2015-2-10

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/177047-optimizing-a-for-loop

评论： Titus Edelhofer 2015-2-10

Hello,

I have a cell containing two cell columns. Column 1 is date and Column 2 is hour. I want to merge them together and transfrom date/time to the Matlab format. I am using a 'for' loop to do this, because I wasn't able to do it using 'cellfun'. This is the code:

for i=1:385521
PT{1}{i,1} = datenum(strcat(PT{1,1}{i,1},PT{1,2}{i,1}),'dd.mm.yyyyHH');
end

This works, but it takes about 6 minuts to do this operation. Could you give me some suggestions on how to optimize this operation. Even better if it is not in a loop.

Thanks, DjR

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

djr 2015-2-10

在 MATLAB Online 中打开

test.txt

This is the sample of the input file (first few lines) and the whole for loop:

fid = fopen('test.txt');
PT = textscan(fid, '%s %s %f %f','Delimiter',',', 'HeaderLines',1);
fclose(fid);
clear ans fid
% Organize data (remove 00:00:00) and convert date to Matlab format
PT{1} = cellfun(@(x) strrep(x, ' 00:00:00',''), PT{1},'UniformOutput',false);
PerRow = cell2mat(strfind(PT{:,1},'.'));
for i=1:length(PT{:,1})
% add two zeros if format is m.d.yyyy
    if PerRow(i,1) == 2 && PerRow(i,2) == 4
        PT{1}{i,1} = ['0', PT{1,1}{i,1}(1:2), '0', PT{1}{i,1}(3:end)];
% Add one zero if format is d.mm.yyyy
    elseif PerRow(i,1) == 2 && PerRow(i,2) ~=4
        PT{1}{i,1} = ['0', PT{1}{i,1}];
% Add one zero if format is dd.m.yyyy
    elseif PerRow(i,2) == 5
        PT{1}{i,1} = [PT{1}{i,1}(1:3), '0', PT{1}{i,1}(4:end)];
    end
% Add zero if format is H    
    if size(PT{1,2}{i},2) == 1
        PT{1,2}{i}=['0', PT{1,2}{i}];
    end
% Combain columns and transform date/time to Matlab format
      PT{1}{i,1} = datenum(strcat(PT{1,1}{i,1},PT{1,2}{i,1}),'dd.mm.yyyyHH');
end
clear i PerRow

The loop without the problematic line takes only about 20 sec on my Linux computer.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Titus Edelhofer 2015-2-10

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/177047-optimizing-a-for-loop#answer_167592

在 MATLAB Online 中打开

Hi,

you should do the strcat out of the loop, something like

PT12 = strcat(PT{1,1}(:,1), PT{1,2}(:,1));

and then you don't need the loop on datenum

PTnum = datenum(PT12, 'dd.mm.yyyyHH');

Titus

3 个评论
显示 1更早的评论隐藏 1更早的评论

djr 2015-2-10

Thanks a lot. It is much more efficient. It takes like 30 seconds to finish. It took about 6 minutes when the line was in the loop.

Titus Edelhofer 2015-2-10

Your welcome. Usually MATLAB is not that bad on loops anymore. But datenum is using a MEX file in the back ground, which has some overhead when called in a (large) loop.

请先登录，再进行评论。