Re: The Few Extents Myth(?) Revisited

Paul Drake wrote:
>
> Connor McDonald <connor_mcdonald_at_yahoo.com> wrote in message news:<3D01100C.5AE6_at_yahoo.com>...
> > This is true, but only applies to dict managed tspaces which of course
> > no-one is using anymore :-) Similarly, its only an issue for tables
> > that you truncate or drop - and of course, we're using global temp
> > tables for that :-)
> >
> > hth
> > connor
>
> Connor,
>
> How are you handling the subject of statistics for global temporary
> tables where the average number of rows for such tables varies widely?
>
> If you're using the CBO (of course) you would have to use 1 set of
> values for statistics for all child tables of the global temp table -
> not a very practical result.
>
> 8.1.7.3.0
>
> thanks,
>
> Paul

In the (few) cases I've used them, I've stolen a technique from Tom Kyte and used dbms_stats to set them directly. Hardly ideal, but in most cases I would propose that people at least know the number of rows in the gtt since you've got to have code to populate it (eg insert into gtt select ..., and then using sql%rowcount). For other stats, such as column level cardinality a 'guestimate' should suffice.

hth
connor

-- 
==============================
Connor McDonald

http://www.oracledba.co.uk

"Some days you're the pigeon, some days you're the statue..."