Re: RTREE on points

Jeff Hoffmann <jeff(at)propertykey(dot)com> writes:
> Tom Lane wrote:
>> How the heck did the GIST index code get developed/tested without some
>> opclasses?

> doing some digging at berkeley, i found the original pggist patch file
> that created the gist access method & gist_box_ops opclass (among
> others). i'm assuming that patch was the basis for what was originally
> introduced, so i don't know why it didn't get included with everything
> else.

The CVS logs show there was some confusion about applying the whole of
the pggist patch, so it's possible that this was an unintentional
omission.

> it looks like there are a lot of calls to internal postgresql box
> comparison functions that would need to get converted to the new calling
> convention, but it should be pretty straightforward to get it to work
> with a recent version of postgresql. it does seem pretty silly to have
> it in there if you don't have any built-in way of using it, if for no
> other reason than to be able to test if the feature even works.

Yes. I had been thinking of migrating one or more of the new contrib
GIST opclasses into the main distribution for 7.2, if only to be able to
create a regress test for GIST. Updating this old Berkeley code would
be another path to create/extend a GIST regress test. If you (or
someone else) have time to do that, it'd be great.

regards, tom lane