Re: [PATCH] we have added support for box type in SP-GiST index

From: David Steele <david(at)pgmasters(dot)net>
To: Teodor Sigaev <teodor(at)sigaev(dot)ru>, pgsql-hackers(at)postgresql(dot)org, Oleg Bartunov <obartunov(at)gmail(dot)com>
Cc: Alexander Lebedev <a(dot)lebedev(at)postgrespro(dot)ru>
Subject: Re: [PATCH] we have added support for box type in SP-GiST index
Date: 2016-03-14 18:01:52
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers

On 2/15/16 10:29 AM, Teodor Sigaev wrote:

> It's very pity but author is not able to continue work on this patch,
> and I would like to raise this flag.
> I'd like to add some comments about patches:
> traversalValue patch adds arbitrary value assoсiated with branch in
> SP-GiST tree walk. Unlike to recostructedValue it could be just pointer,
> it havn't to be a regular pgsql type. Also, it could be used independly
> from reconstructedValue. This patch is used both following attached
> patches.
> range patch just switchs using reconstructedValue to traversalValue in
> range opclasses. reconstructedValue was used just because it was an
> acceptable workaround in case of range type. Range opclase stores a full
> value in leafs and doesn't need to use reconstructedValue to return
> tuple in index only scan.
> See
> q4d patch implements index over boxes using SP-GiST. Basic idea was an
> observation, range opclass thinks about one-dimensional ranges as 2D
> points.
> Following this idea, we can think that 2D box (what is 2 points or 4
> numbers) could represent a 4D point. We hoped that this approach will be
> much more effective than traditional R-Tree in case of many overlaps in
> box's collection.
> Performance results are coming shortly.

It appears that the issues raised in this thread have been addressed but
the patch still has not gone though a real review.

Anybody out there willing to take a crack at a review? All three
patches apply (with whitespace issues).


In response to


Browse pgsql-hackers by date

  From Date Subject
Next Message Vladimir Borodin 2016-03-14 18:15:29 Re: Improving replay of XLOG_BTREE_VACUUM records
Previous Message Robert Haas 2016-03-14 17:58:21 Re: Proposal: "Causal reads" mode for load balancing reads without stale data