[postgis-devel] New node splitting algorithm for GiST

[Paul Ramsey]

OK, yes testing 2.0 against 2.1svn makes loads of sense. I've made the
new split the default, please test test test.
P.

On Wed, Nov 14, 2012 at 2:48 AM, Paragon Corporation <lr at pcorp.us> wrote:
> I don't see why we can't just enable it by default or Paul -- if you really
> don't want it default make it a configuration variable.
>
> Then it would be easy to have our buildbots just build trunk with a
> different configure command. Fussing with .c files is much messier since it
> requires changing source and remembering to change back if changes made to
> the file (so too risky for buildbot testing).
>
> I think for benchmarking comparing 2.0 behavior with new 2.1 will be
> sufficient.
>
> Just my 2 cents,
> Regina
>
> ________________________________
> From: postgis-devel-bounces at lists.osgeo.org
> [mailto:postgis-devel-bounces at lists.osgeo.org] On Behalf Of Alexander
> Korotkov
> Sent: Wednesday, November 14, 2012 1:19 AM
> To: PostGIS Development Discussion
> Cc: PostGIS Development Discussion
>
> Subject: Re: [postgis-devel] New node splitting algorithm for GiST
>
> Paul,
>
> Who are expected to try it? I think in current state it's unlikely that
> someone would try it until you ask him in person.
>
> Also, ITSM you selected extremely cautious strategy. It would be reasonable
> to have a macro if we had some good algorithm of split before. For example,
> split algorithm from R*-tree. But "New linear node splitting" sucks so
> much...
>
> ------
> With best regards,
> Alexander Korotkov.
>
> On Wed, Nov 14, 2012 at 2:14 AM, Paul Ramsey <pramsey at opengeo.org> wrote:
>>
>> This is now on trunk, with a #define KOROTKOV_SPLIT in
>> gserialized_gist_2d.c to turn it on/off. Some extra testing (you have
>> to build a new index to get the new split in effect, just querying an
>> existing index will make no difference at all) to ensure we have
>> more/better performance would be great.
>>
>> http://trac.osgeo.org/postgis/ticket/1895
>>
>> P.
>>
>> On Sun, Jul 1, 2012 at 8:00 AM, Alexander Korotkov <aekorotkov at gmail.com>
>> wrote:
>> > Done: http://trac.osgeo.org/postgis/ticket/1895
>> >
>> > ------
>> > With best regards,
>> > Alexander Korotkov.
>> >
>> > On Sun, Jul 1, 2012 at 5:54 PM, Paragon Corporation <lr at pcorp.us> wrote:
>> >>
>> >> Alex,
>> >>
>> >> Can you attach this to a ticket.  That way Paul won't forget or one of
>> >> us
>> >> can remind him when he does :)
>> >>
>> >> http://trac.osgeo.org/postgis/
>> >>
>> >> Slate it for 2.1.0
>> >>
>> >> Thanks,
>> >> Regina
>> >>
>> >> ________________________________
>> >> From: postgis-devel-bounces at postgis.refractions.net
>> >> [mailto:postgis-devel-bounces at postgis.refractions.net] On Behalf Of
>> >> Alexander Korotkov
>> >> Sent: Sunday, July 01, 2012 8:33 AM
>> >> To: PostGIS Development Discussion
>> >> Subject: Re: [postgis-devel] New node splitting algorithm for GiST
>> >>
>> >> On Sat, Jun 30, 2012 at 10:59 PM, Paul Ramsey <pramsey at opengeo.org>
>> >> wrote:
>> >>>
>> >>> No progress, still an "item of interest" in my list of things I might
>> >>> do. If you prepare a compile-time conditional patch, I'd be interested
>> >>> in reviewing and applying it.
>> >>
>> >>
>> >> OK, it's here.
>> >> Just an small example. Without patch.
>> >>
>> >> test=# create index geonames_idx on geonames using gist (point);
>> >> CREATE INDEX
>> >> Time: 113796,848 ms
>> >>
>> >> test=# explain (analyze, buffers) select * from geonames where point &&
>> >> 'BOX(34.4671 126.631,34.5023 126.667)'::box2d::geometry;
>> >>
>> >> QUERY PLAN
>> >>
>> >>
>> >> ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>> >>  Index Scan using geonames_idx on geonames  (cost=0.00..14.05 rows=2
>> >> width=128) (actual time=0.589..0.876 rows=11 loops=1)
>> >>    Index Cond: (point &&
>> >>
>> >> '0103000000010000000500000016FBCBEEC93B4140AAF1D24D62A85F4016FBCBEEC93B4140A69BC420B0AA5F40F163CC5D4B404140A69BC420B0AA5F40F163CC5D4B404140AAF1D24D62A85F4016FBCBEEC93B4140AAF1D24D62A85F40'::geometry)
>> >>    Buffers: shared hit=26
>> >>  Total runtime: 0.927 ms
>> >> (4 rows)
>> >>
>> >> With patch.
>> >>
>> >> test=# create index geonames2_idx on geonames2 using gist (point);
>> >> CREATE INDEX
>> >> Time: 118382,076 ms
>> >>
>> >> test=# explain (analyze, buffers) select * from geonames2 where point
>> >> &&
>> >> 'BOX(34.4671 126.631,34.5023 126.667)'::box2d::geometry;
>> >>
>> >>
>> >> QUERY PLAN
>> >>
>> >>
>> >> ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>> >>  Index Scan using geonames2_idx on geonames2  (cost=0.00..22.40 rows=4
>> >> width=128) (actual time=0.209..0.225 rows=11 loops=1)
>> >>    Index Cond: (point &&
>> >>
>> >> '0103000000010000000500000016FBCBEEC93B4140AAF1D24D62A85F4016FBCBEEC93B4140A69BC420B0AA5F40F163CC5D4B404140A69BC420B0AA5F40F163CC5D4B404140AAF1D24D62A85F4016FBCBEEC93B4140AAF1D24D62A85F40'::geometry)
>> >>    Buffers: shared hit=11
>> >>  Total runtime: 0.276 ms
>> >> (4 rows)
>> >>
>> >> ------
>> >> With best regards,
>> >> Alexander Korotkov.
>> >>
>> >>
>> >> _______________________________________________
>> >> postgis-devel mailing list
>> >> postgis-devel at postgis.refractions.net
>> >> http://postgis.refractions.net/mailman/listinfo/postgis-devel
>> >>
>> >
>> >
>> > _______________________________________________
>> > postgis-devel mailing list
>> > postgis-devel at postgis.refractions.net
>> > http://postgis.refractions.net/mailman/listinfo/postgis-devel
>> >
>> _______________________________________________
>> postgis-devel mailing list
>> postgis-devel at lists.osgeo.org
>> http://lists.osgeo.org/cgi-bin/mailman/listinfo/postgis-devel
>
>
>
> _______________________________________________
> postgis-devel mailing list
> postgis-devel at lists.osgeo.org
> http://lists.osgeo.org/cgi-bin/mailman/listinfo/postgis-devel
>