in reply to Quick file search

It would be interesting to see how much faster the win/dos "tree" utility is than the File::Find module (or any of its relatives on CPAN). I know that in unix, File::Find takes much longer than the standard "find" utility for a directory tree of any considerable size.

The only problem with the OP code is that you have to remember to run "tree" yourself -- and save its output in a specific file whose name is hard-coded in your perl script -- before running the script, so that you know your file list is up-to-date. But that's unnecessary -- just run "tree" inside the perl script (it'll still run just as fast): 

my $music_path = ".\\music";  # probably should make this an absolute 
+path...

open my $fh, "D:\\tree /f /a $music_path |"
    or die "cannot run tree on $music_path: $!";

while ( <$fh> ) {
   # everything else stays the same...
}

[download]
(not tested -- I'm not a windows user)

Actually, it might also be interesting to see whether a win/dos port of the unix "find" utility is slower or faster than "tree"...

Replies are listed 'Best First'.
Re^2: Quick file search
by zentara (Cardinal) on Jan 05, 2006 at 11:46 UTC
    I know that in unix, File::Find takes much longer than the standard "find" utility for a directory tree of any considerable size.

    This was discussed in considerable length in Myth busted: Shell isn't always faster than Perl and was wondering if you can back up the claim of "much longer" . My tests show differences of only a fraction of a second on 100 meg tree.

    I'm not really a human, but I play one on earth. flash japh
[reply]
      The evidence I've gathered about unix "find" vs. File::Find is at least a couple years old, and I haven't checked to see whether there has been an update to File::Find since then, nor tested it more recently to see whether it has improved within the last couple years. (I assume unix "find" remains about the same.)

      In any case, I posted some benchmarks here and here. Looking at those again just now, my statement above about "5- or 6-to-1" should have simply been "5-to-1".

[reply]
Re^2: Quick file search
by sh1tn (Priest) on Jan 05, 2006 at 05:36 UTC
    The lack of unix "find" utility is enormous but "tree" filtered results solve to some extent this problem: 
    # rough comparison
D:\music>perl -MFile::Find -e "find(sub{$File::Find::name}, '.');print
+ time-$^T"
6

D:\music>perl -e "`tree /f /a`;print time-$^T"
1

    [download]
    Where "music" contains 5547 files (without the directory inodes).


[reply]
[d/l]
      The lack of unix "find" utility is enormous

      I'm wondering what you mean by that. There are windows ports of "find" available (google "unix tools for windows"); the ATT Research Labs version and the cygwin version are both authentically "unix-like" (or maybe "gnu-ish" is the better term).

      "tree" filtered results solve to some extent this problem

      Oh yes. That seems consistent with what I've seen in the unix domain -- about a 5- or 6-to-1 speed ratio comparing File::Find to "find". For really big directory trees, that multiplier becomes devastatingly significant.

[reply]
        One thing to note is that you need to put the Cygwin before Windows in the PATH for this to work, since there already is a "find" (although it's really a rather weak "grep") on the system.

        /J

[reply]
        No need for tool from the "outside" world, maybe just: 
        my $search_what  = shift 
    || die "no search criteria provided\n";
my $search_where = shift;
my $file  = qr/\|(?!\+\S{3})(\s.*?)$argv/i;
my $clean = qr/^\W+(.+?)\W+$/;

$dir = -d $dir 
    ? $dir 
    : '.';

for(`tree /a /f $search_where`){
    /(\\|\+)-{3}/ 
           and $dir= $_
           and next;
    if (/$search_what/) {
        s/$clean/$1/ for $_, $dir;
           $dir 
               ? print "dir:   $dir$/","file:  $_$/$/"
               : print "dir: . $/","file:  $_$/$/";           
    }
}

        [download]


[reply]
[d/l]

In Section Cool Uses for Perl