matchfiles

 function f = matchfiles(patterns,sorttype)

 <patterns> is
   (1) a string that matches zero or more files or directories (wildcards '*' okay)
   (2) the empty matrix []
   (3) a cell vector of zero or more things like (1) or (2)
 <sorttype> (optional) is how to sort in each individual match attempt.
   't' means sort by time (newest first)
   'tr' means sort by time (oldest first)
   default is [], which means to sort alphabetically by explicitly using MATLAB's sort function.
   (note that MATLAB's sort function may sort differently than UNIX's ls function does!)

 return a cell vector of strings containing paths to the matched files and/or directories.
 if there are no matches for an individual match attempt, we issue a warning.

 this function should be fully functional on Mac and Linux.  however, on Windows,
 we have the following limitations:
 - you cannot use the '?' operator
 - you can use the '*' operator only once and at the end of the expression
   (not in an intermediate directory)

 on Mac and Linux, if we run into the too-many-files limitation of the ls command,
 we will resort to the alternative mode described above, and this inherits the 
 same limitations.

 history:
 2011/09/28 - if ls returns too many files, resort to alternative.  also, the alternative mode now allows sorttype to be specified.
 2011/08/07 - allow empty matrix as an input
 2011/04/02 - now, works on Windows (in a limited way)
 2011/04/02 - oops, time-sorting behavior did not work.  bad bug!!!
 2011/02/24 - escape spaces in patterns using \.  this fixes buggy behavior.
 2011/01/21 - explicitly use MATLAB's sort function to ensure consistency across platforms.

0001 function f = matchfiles(patterns,sorttype)
0002 
0003 % function f = matchfiles(patterns,sorttype)
0004 %
0005 % <patterns> is
0006 %   (1) a string that matches zero or more files or directories (wildcards '*' okay)
0007 %   (2) the empty matrix []
0008 %   (3) a cell vector of zero or more things like (1) or (2)
0009 % <sorttype> (optional) is how to sort in each individual match attempt.
0010 %   't' means sort by time (newest first)
0011 %   'tr' means sort by time (oldest first)
0012 %   default is [], which means to sort alphabetically by explicitly using MATLAB's sort function.
0013 %   (note that MATLAB's sort function may sort differently than UNIX's ls function does!)
0014 %
0015 % return a cell vector of strings containing paths to the matched files and/or directories.
0016 % if there are no matches for an individual match attempt, we issue a warning.
0017 %
0018 % this function should be fully functional on Mac and Linux.  however, on Windows,
0019 % we have the following limitations:
0020 % - you cannot use the '?' operator
0021 % - you can use the '*' operator only once and at the end of the expression
0022 %   (not in an intermediate directory)
0023 %
0024 % on Mac and Linux, if we run into the too-many-files limitation of the ls command,
0025 % we will resort to the alternative mode described above, and this inherits the
0026 % same limitations.
0027 %
0028 % history:
0029 % 2011/09/28 - if ls returns too many files, resort to alternative.  also, the alternative mode now allows sorttype to be specified.
0030 % 2011/08/07 - allow empty matrix as an input
0031 % 2011/04/02 - now, works on Windows (in a limited way)
0032 % 2011/04/02 - oops, time-sorting behavior did not work.  bad bug!!!
0033 % 2011/02/24 - escape spaces in patterns using \.  this fixes buggy behavior.
0034 % 2011/01/21 - explicitly use MATLAB's sort function to ensure consistency across platforms.
0035 
0036 % input
0037 if ~exist('sorttype','var') || isempty(sorttype)
0038   sorttype = '';
0039 end
0040 if ~iscell(patterns)
0041   patterns = {patterns};
0042 end
0043 % % if ~isunix
0044 % %   assert(isempty(sorttype),'due to current implementation limitations, <sorttype> must be [] on Windows');
0045 % % end
0046 
0047 % do it
0048 f = {};
0049 for p=1:length(patterns)
0050   if isempty(patterns{p})
0051     continue;
0052   end
0053   
0054   % if UNIX, try to use ls
0055   doalternative = 0;
0056   if isunix
0057     [status,result] = unix(sprintf('/bin/ls -1d%s %s',sorttype,regexprep(patterns{p},' ','\\ ')));
0058     if status==126  % oops, too many files
0059       doalternative = 1;
0060     elseif status~=0
0061       warning(sprintf('failure in finding the files or directories for %s',patterns{p}));
0062     else
0063       temp = strsplit(result,sprintf('\n'));
0064       temp = temp(~cellfun(@(x) isempty(x),temp));  % remove empty entries
0065       if isempty(sorttype)
0066         temp = sort(temp);
0067       end
0068       f = [f temp];
0069     end
0070   end
0071   
0072   % if not UNIX or if we failed by matching too many files using ls, we have to do the alternative
0073   if ~isunix || doalternative
0074     if exist(patterns{p},'dir')
0075       f = [f {patterns{p}}];
0076     else
0077       tempdir = stripfile(patterns{p});
0078       dmatch = dir(patterns{p});
0079       if isempty(dmatch)
0080         warning(sprintf('failure in finding the files or directories for %s',patterns{p}));
0081       else
0082         if isequal(sorttype,'t')
0083           [ss,ii] = sort(cat(2,dmatch.datenum),2,'descend');
0084         elseif isequal(sorttype,'tr')
0085           [ss,ii] = sort(cat(2,dmatch.datenum));
0086         else
0087           [ss,ii] = sort(cat(2,{dmatch.name}));
0088         end
0089         dmatch = dmatch(ii);
0090         temp = cat(2,{dmatch.name});
0091         temp = temp(~cellfun(@(x) isempty(x),temp));  % remove empty entries
0092         temp = temp(~cellfun(@(x) isequal(x(1),'.'),temp));  % remove things starting with .
0093         if ~isempty(tempdir)
0094           temp = cellfun(@(x) [tempdir x],temp,'UniformOutput',0);  % add directory
0095         end
0096         f = [f temp];
0097       end
0098     end
0099   end
0100 end