If you're going to be converting multiple date strings during the course of a script run, and if the dates are reasonably well bounded, then another approach is to precalculate a set of   "yyyy/mm/dd" => integer mappings, set them up to load into a hash (e.g., pre-generate a separate .pl file that does nothing except load up a hash, then require that file into your script), then use table lookup to do the conversion. This is less of a space problem than it might seem: 3653 entries covers a ten year span.