On Sun, 13 Mar 2005 00:18:39 +0100, Martijn van Oosterhout
<kleptog@svana.org> wrote:
> On Sat, Mar 12, 2005 at 08:24:20PM +0000, Mike Rylander wrote:
> > There is now way to have PG completely skip the unused partitions.
> > However, with an index on the "partitioner" column of each partition
> > table PG will basically fall of the end of the index almost
> > imediately, and then use an appropiate plan for the useful partitions.
> > It's not the same as Oracle partitioning, but without expicit Planner
> > work it can't be.
>
> Note, the planner does have most of the intelligence needed to
> determine when it can skip a partition. It has code that can determine
> if the intersection of:
>
> Partition constraint: lower bound < value < upper bound
> Query clause: value [operator] constant
>
> Will be always false or can be true. It's just a matter of getting the
> clause to the right place to make it skip...
I see. I should have qualified my statement with "*as of now* there
is no way". If I felt comfortable digging into the planner I'd offer
to take that on, but...
Thanks for pointing out that detail!
--
Mike Rylander
mrylander@gmail.com
GPLS -- PINES Development
Database Developer
http://open-ils.org