Who said it had to be consecutive? This is sequential, unique (unless you can create more than one row per microsecond in a single process), and portable.
use Time::HiRes qw(gettimeofday);
my ($secut,$musec) = gettimeofday;
my ($uniqueid) = sprintf("%010d%06d%05d", $secut, $musec, $$);
Notes:
- The lead 0s are there to pad the digits to the proper resolution. If your system supports >16-bit process-IDs, that last 5 may have to change.
- This can also scale across multiple machines (eg. multiple webservers writing to the same database), but to do so, each machine should have a unique numeric machine identifier as part of the id as well; add an extra 2 or 3 digit (or bigger, if you have to!) integer to the expansion to account for it.
- Keeping the first two variables in that order will assure that the records have linearity -- ie. sorting on this ID will put things in proper chronological order, with one-microsecond granularity.
Spud Zeppelin * spud@spudzeppelin.com