From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Dimitri Fontaine <dimitri(at)2ndQuadrant(dot)fr> |
Cc: | Robert Haas <robertmhaas(at)gmail(dot)com>, Teodor Sigaev <teodor(at)sigaev(dot)ru>, Oleg Bartunov <oleg(at)sai(dot)msu(dot)su>, pgsql-hackers(at)postgresql(dot)org |
Subject: | Re: KNNGIST next step: adjusting indexAM API |
Date: | 2010-12-01 15:36:50 |
Message-ID: | 16286.1291217810@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Dimitri Fontaine <dimitri(at)2ndQuadrant(dot)fr> writes:
> Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> writes:
>> Right, AFAIK there is nothing in KNNGIST that would involve an on-disk
>> data change.
> But any external module relying on GiST will have to provide for the new
> function you're thinking about, right? Updating was already needed to
> cope with the newer consistent API, I guess.
ISTM that it should be possible to allow an opclass to not supply the
new hook function if it doesn't support any ordering operators. So
that's not really a serious problem. The existing patch approaches
this by having two different APIs for the Consistent function depending
on whether the opclass supports ordering operators or not. I find that
pretty icky, even though it nominally avoids breaking existing opclass
modules.
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Bruce Momjian | 2010-12-01 15:36:51 | Re: crash-safe visibility map, take three |
Previous Message | Kevin Grittner | 2010-12-01 15:26:24 | Re: profiling connection overhead |