;-)

use Digest::MD5;

$pretty_darn_close_to_unique = md5_hex(md5_hex(time() . $$ . rand . {}
+));
[download]

-stvn

I love the subtle inclusion of a memory reference.

---

Examine what is said, not who speaks.

"Efficiency is intelligent laziness." -David Dunham
"Think for yourself!" - Abigail
"Memory, processor, disk in that order on the hardware side. Algorithm, algoritm, algorithm on the code side." - tachyon

You can get guaranteed uniqueness by replacing the rand call with a "hold this pid until the time changes":

  $unique = time() . $$;
  sleep 1;
[download]

Hugo

Guarenteed with a given process, but if multiple copies of the process are running, with time() limited to 1 second resolution?

---

Examine what is said, not who speaks.

"Efficiency is intelligent laziness." -David Dunham
"Think for yourself!" - Abigail
"Memory, processor, disk in that order on the hardware side. Algorithm, algoritm, algorithm on the code side." - tachyon

Then the sleep() should ensure the pid is not reused within the same tick, so it should still remain unique.

It does depend a bit on the precise sleep() implementation though; possibly safer (and a bit faster) would be:

  my $t = time;
  select(undef, undef, undef, 0.1) while time <= $t;
  $unique = "$t$$";
[download]

It also assumes a monotonically increasing time: if the sysadmin may correct a fast clock backwards, it would be very difficult to cope with. (Note that ntpd will adjust the time in increments of less than a second except when the clock error is "large"; once initially synchronised, a running system corrected by ntpd should never normally skip a full second forwards or backwards.)

Hugo

Or use a counter:

{
  my $cnt = 0;
  sub unique {
     return $time() . $$ . $cnt++;
  }
}
$unique = unique();
[download]