... use the start time of the script ($^T) as a base index, and increment from there. And if you restart the script, it won't collide with any names from before

How much would you like to bet on that idea?

$^T has second resolution. Start the script at some point T in time. $filename=T at start of script. Process 100 files in 60 seconds, with $filename++ for each file. $filename=T+100 at end of script. Start again for the next set of files at T+65 seconds. First $filename=$^T=T+65. Instant collision.

Alexander

OK, flaw in my logic there - which you could have simply pointed out without the challenge to a bet. Hope you feel superior now.

The point of my original post is that there may be a sufficiently simple way of doing it, without resorting to databases or with the caveats associated with random numbers and hashing.

Maybe the Time::HiRes module would be an option - microsecond resolution, if available, should be enough:

use strict;
use warnings;
use Time::HiRes qw(gettimeofday);

my($seconds, $microseconds);
my $index;

for (1..10) {
   ($seconds, $microseconds) = gettimeofday;
   $index = sprintf("%d%06d", $seconds, $microseconds);
   print "$index\n";
}
[download]

Output:


1290951829553400
1290951829553437
1290951829553448
1290951829553457
1290951829553467
1290951829553477
1290951829553487
1290951829553496
1290951829553505
1290951829553515
[download]

And what happens if the site where these images are being uploaded is, or becomes, a FaceBook or Flickr?

Or any site that has multiple upload servers; and/or uses threaded servers.

Now any scheme that tries to use upload time--even to the microsecond; and with or without combination with current pid--is almost guaranteed to create duplicates.

easy to fix :)

$^T . rand($^T) . $$
$^T . int( rand($^T) ) . $$
[download]

heck you could even append

crypt $^T.$$.$^T, rand($^T)
[download]