Re: Patch: add GiST support for BOX @> POINT queries
Hitoshi Harada <umi.tanuki@gmail.com>
From: Hitoshi Harada <umi.tanuki@gmail.com>
To: Andrew Tipton <andrew.t.tipton@gmail.com>
Cc: pgsql-hackers@postgresql.org
Date: 2011-07-09T14:29:49Z
Lists: pgsql-hackers
2011/6/19 Hitoshi Harada <umi.tanuki@gmail.com>: > 2011/6/17 Andrew Tipton <andrew.t.tipton@gmail.com>: >> >> At this point I'm a bit lost -- while pg_amop.h has plenty of examples >> of crosstype comparison operators for btree index methods, there are >> none for GiST. Is GiST somehow a special case in this regard? > > It was I that was lost. As Tom mentioned, GiST indexes have records in > pg_amop in their specialized way. I found gist_point_consistent has > some kind of hack for that and pg_amop for point_ops records have > multiple crosstype for that. So, if I understand correctly your first > approach modifying gist_box_consistent was the right way, although > trivial issues should be fixed. Also, you may want to follow point_ops > when you are worried if the counterpart operator of commutator should > be registered or not. > > Looking around those mechanisms, it occurred to me that you mentioned > only box @> point. Why don't you add circly @> point, poly @> point as > well as box? Is that hard? > It looks like the time to wrap up. I marked "Return with Feedback" on this patch, since response from author has not come for a while. You may think the fix was pretty easy and the patch be small, but more general approach was preferred, I guess. Looking forward to seeing it in better shape next time! Thanks, -- Hitoshi Harada