There are two general ways to this:
- Read the data all at once, then do your analysis on the full file, then write the complete result.
- Read the data line by line, analyzing as you go, writing as needed.
Option 1 is generally recommended, as it will prevent errors during processing to lock down files. It is generally easier to keep everything ordered and structured.
Option 2 will work for any size file, even if it doesn't fit in memory.
% store the link to the data you posted:
URL='https://www.mathworks.com/matlabcentral/answers/uploaded_files/906200/matlabdata.txt';
Option 1:
%If you don't have R2020b or newer, use readfile:
%https://www.mathworks.com/matlabcentral/fileexchange/68780
data=cellstr(readlines(URL));
%remove empty lines
data(cellfun('prodofsize',data)<3)=[]; % the legacy syntax for cellfun is actually faster than a loop
L=false(size(data));
for n=1:numel(data)
%find '1 x'
L(n)=strcmp(data{n}(1:3),'1 x');
%keep only the number
data{n}(1:max(strfind(data{n},' ')))='';
end
ind=find(L);
%confirm the file is complete
if (max(ind)+2)>numel(data),error('file incomplete'),end
%write the output file
fid=fopen('output.txt','w');
for n=1:numel(ind)
fprintf(fid,'%s %s %s\n',data{ind(n)+(0:2)});
end
fclose(fid);
Option 2:
fid_data=fopen(websave('metlabdata.txt',URL),'r');
fid_output=fopen('output_2.txt','w');
data=cell(1,3);
while ~feof(fid_data) %there are still lines to be read
line=fgetl(fid_data);
if strcmp(line(1:(min(end,3))),'1 x')
% found a line to be written
data{1}=line;
data{2}=fgetl(fid_data);
data{3}=fgetl(fid_data);
if any(cellfun('isclass',data,'double'))
%fgetl returns -1 on error
fclose(fid_data);fclose(fid_output);
error('file incomplete')
end
for n=1:3
%keep only the number
data{n}(1:max(strfind(data{n},' ')))='';
end
%write line
fprintf(fid_output,'%s %s %s\n',data{:});
end
end
fclose(fid_data);fclose(fid_output);
Now we confirm the two results are the same:
type('output.txt')
type('output_2.txt')
