[postgis-devel] 1.3.4SVN Testers Needed

Tue Oct 14 07:57:13 PDT 2008

Good Job Paul.  The ST_Intersects performance seems much better. My
ST_Dwithin I'll need to retest.  I got very mixed results on that.
Seems my ST_DWithin query caches in memory much more than my
ST_Intersects query so my second run always seems to run under 345ms on
both (vs. first which ran in 8000ms).  Also it seemed to be worse on the
new install sometimes and sometimes not - so I have to figure out what's
up with that.

I just ran a quick test to test out small multipoly in small numbers of
large multipolys intersects

 comparing my postgresql Windows 8.2.1 1.3.3 against same database on VM
OpenSUSE 10.3 8.2.6 1.3.4SVN.  

My test server is a VM running OpenSUSE 10.3, so its not exactly a one
to one compare but the windows box has performed faster.

But for this query  - my OpenSUSE VM takes -  8,234 ms

"POSTGIS="1.3.4SVN" GEOS="3.1.0-CAPI-1.5.0" PROJ="Rel. 4.6.0, 21 Dec
2007" USE_STATS"  (though hm I guess I should recompile Geos since this
is the version from last friday)
 SELECT count(p.the_geom) as tot, n.name 
 FROM neighborhoods n 
 INNER JOIN assessing.parceltime_2007 p 
 ON ST_Intersects(n.the_geom, p.the_geom) 
 WHERE  ST_IsValid(n.the_geom) 
 GROUP BY  n.name
 ORDER BY n.name; 

My Windows 2003 (which before used to be a little faster than the old
1.3.3 install on VM) takes
"POSTGIS="1.3.3" GEOS="3.0.0-CAPI-1.4.1" PROJ="Rel. 4.6.0, 21 Dec 2007"
USE_STATS"
65,031 ms.

Both results give
  tot  |       name       
-------+------------------
 14976 | Allston/Brighton
  6124 | Charlestown
  9355 | Hyde Park
  9610 | Jamaica Plain
  7214 | Mattapan
  9133 | Roslindale
 11528 | Roxbury
  9109 | South End
 11444 | West Roxbury
----For the ST_DWithin test  ----
I have a view that looks like this

CREATE OR REPLACE VIEW foreclosure.vwfit_parcelswithhalfmilebuffer AS 
 SELECT DISTINCT ON (p.parcel_id) f.fit_name, f.start_date, f.end_date,
p.parcel_id, p.pid, p.cm, p."owner", p.st_num, p.st_name, p.st_nam_suf,
p.unit_num, p.zipcode, p.lu, p.ptype, p.lot_size, p.mail_addre,
p.mail_cs, p.mail_zip, p.own_occ, p.code, p.bldg_area, p.total,
p.value_land, p.value_bldg, p.gross_tax, p.story_hg, p.xcoord, p.ycoord,
p.neighborhood, st_distance(p.the_point, f.the_geom) AS
distance_ft_from_fit
   FROM landparcels p
   JOIN foreclosure.fit_area f ON st_dwithin(p.the_point, f.the_geom,
2640)
  ORDER BY p.parcel_id, st_distance(p.the_point, f.the_geom);

On the Windows 2003 - 1.3.3 install doing a query like

SELECT count(pid) as numparcels, fit_name
FROM foreclosure.vwfit_parcelswithhalfmilebuffer
GROUP BY fit_name
ORDER BY fit_name;

On Windows 1.3.3 install Takes 7,178 ms first run and 345ms each
consecutive.

On OpenSUSE 1.3.4 SVN Geos 3.1 takes I think it took 2,167 ms first run
and 345 ms each consecutive.

Both return
 numparcels | fit_name 
------------+----------
       3644 | A
       4000 | B
       2740 | C

But can't be sure.  Maybe it caches so well because I have the
ST_DWithin clause in a view.

-----Original Message-----
From: postgis-devel-bounces at postgis.refractions.net
[mailto:postgis-devel-bounces at postgis.refractions.net] On Behalf Of Paul
Ramsey
Sent: Monday, October 13, 2008 2:57 PM
To: PostGIS Development Discussion
Subject: [postgis-devel] 1.3.4SVN Testers Needed

1.3.4 is very close to release, but we can release it faster, with
less intestinal distress, if we get some solid pre-release testing
from folks. You can download the current SVN version directly from the
downloads page, no SVN client required:

http://postgis.refractions.net/download/postgis-1.3.4SVN.tar.gz

If you want to also try the latest preparedgeometry enhancements,
you'll need to upgrade your GEOS version to the SVN trunk before
compiling and installing1.3.4SVN,
http://download.osgeo.org/geos/geos-svn.tar.bz2

Even if you don't, you should see some performance increases in 1.3.4,
particularly in point-and-polygon predicate tests, and large joins
using point-ant-polygon predicates.

Let us know how it goes. We are particularly  interested in any
compilation difficulties or regression failures on less common unix
variants (Solaris on Sparc, AIX on Power, etc)

Paul
_______________________________________________
postgis-devel mailing list
postgis-devel at postgis.refractions.net
http://postgis.refractions.net/mailman/listinfo/postgis-devel
-----------------------------------------
The substance of this message, including any attachments, may be
confidential, legally privileged and/or exempt from disclosure
pursuant to Massachusetts law. It is intended
solely for the addressee. If you received this in error, please
contact the sender and delete the material from any computer.