Also, if you are using a Unix platform, you'd probably be better off using the -x (can be executed) file test rather than -e (exists).

You seem to have the fastest solution. Unless I miss my guess, File::Find is too much for an operation like this:

#!/usr/bin/perl -w

use strict;
use Benchmark;
use File::Find;

my @paths = (split/\:/,$ENV{PATH});

sub which1 {
        my $filename= "test.txt";
        foreach my $pathname (@paths) {
                print "$pathname/$filename\n" if ( -e "$pathname/$file
+name" );
        };
}

sub which2 {
        my $filename = "test.txt";
        finddepth sub {print "$_\n" if ( /$filename$/)} , @paths;
}

timethese (100, {'whichone' => \&which1, 'whichtwo' => \&which2});
[download]

Yields:

[10:27:15 jhorner@gateway scripts]$ ./ch0as-1.pl test.txt
Benchmark: timing 100 iterations of whichone, whichtwo...
  whichone:  0 wallclock secs ( 0.01 usr +  0.02 sys =  0.03 CPU) @ 33
+33.33/s (n=100)
            (warning: too few iterations for a reliable count)
  whichtwo: 14 wallclock secs (12.54 usr +  1.55 sys = 14.09 CPU) @  7
+.10/s (n=100)
[download]

Unless someone can tell me why the File::Find version takes so long, I'm stumped. I have a feeling it has to do with File::Find traversing the entire filetree under the @path directories. If someone can let me know how to make it search ONLY the @path directories and not subdirectories, I'll be grateful.

I also have a feeling that File::Find is overkill in this, since you aren't recursing the directory structure.

Anyone else have good info on this?

J. J. Horner
Linux, Perl, Apache, Stronghold, Unix
jhorner@knoxlug.org http://www.knoxlug.org/

I looked at the code for File::Find, and two things stand out immediately that I think make a big difference:
• Code Size - File::Find is about 9 pages of code!
• @ISA - Being OO, it uses CWD in it's @ISA, which is more code to load.
Those two things alone will slow down execution compared to 3 lines of code making up a foreach loop.
Of course, the File::Find makes up for it's size in it's flexibility and features.... :)

hth, Maurice

How fixed are the contents of your path?

You might consider building a hash where the keys are the filenames and the values are the paths. You could build the hash using a program (why not call it rehash like the csh does) and tie it to a DBM file. Your which script would then only have to look up the filename in the tied hash.

--
<http://www.dave.org.uk>

European Perl Conference - Sept 22/24 2000 <http://www.yapc.org/Europe/>

sub Which($) {
my $file = shift;
my $loc = `which $file`;
chomp $loc;
return $loc;
}
[download]

sub Which($)
  {
   my $Filename=shift;
   my @Arr = ( split /\:/,$ENV{PATH} )
   for(my $i = 0; $i <= #$Arr; $i += 2)
   {
    return "$Arr[$i]/$Filename\n" if ( -e "$Arr[$i]/$Filename" );
    return "$Arr[$i+1]/$Filename\n" if ( -e "$Arr[$i+1]/$Filename" );
   };
  }
[download]

If my C memory can come back to me I remember a spiffy trick was to unroll loops. Right now you are doing a foreach, try something like this.

sub Which($)
  {
   my $Filename=shift;
   my @Arr = ( split /\:/,$ENV{PATH} )
   for(my $i = 0; $i <= #$Arr; $i += 2)
   {
    return "$Arr[$i]/$Filename\n" if ( -e "$Arr[$i]/$Filename" );
    return "$Arr[$i+1]/$Filename\n" if ( -e "$Arr[$i+1]/$Filename" );
   };
  }
[download]

This code's sorta ugly but you get the picture I'm sure, have it check 3 or 4 at a time and you'd be amazed how much faster it will run.

Edit Petruchio Wed May 22 02:05:51 UTC 2002 - Added code tags, other markup.