On 10/10/06, Stephen Frost <sfrost@snowman.net> wrote:
> * Stephen Frost (sfrost@snowman.net) wrote:
> > I'm going to be running a very large query shortly using
> > this aaccum and will report back how it goes.
>
> It went *very* well, actually much better than I had originally
> expected. This query used to take over 12 hours to complete (about 11
> of which was after the main table involved was sorted). With this new
> aaccum in place the whole query only took about an hour, most of which
> was the sort and join required by the query. The aggregation (aaccum)
> and r_hist() (R histogram function generating PNGs) took only a few
> minutes.
very cool, and definately useful imo. i use array_accum all the time,
and have not always been happy with its performance. if your stuff
passes muster, would be nice to see it move into core :-).
merlin
