1 Sep 2008 07:38
Re: optimizing intersection of two large feature classes
Paul Ramsey <pramsey <at> cleverelephant.ca>
2008-09-01 05:38:57 GMT
2008-09-01 05:38:57 GMT
In general, overlays are going to be slow in PostGIS, until we have prepared geometries for intersection as well as predicates. However, since one side of your overlay is a grid, perhaps you can use the grid property to cut your problem down in size. Try testing the containment of the envelope of your complex polygons by the grid squares, and anything which passes that test, add to your resultant table without change. I think there is even an index operator for it... "A <at> B The " <at> " operator returns true if A's bounding box is completely contained by B's" So you could use that index op to remove all the complete containment cases from your workload, leaving only the cases that actually cross lines to be computed. Paul On Mon, Aug 25, 2008 at 12:43 AM, andreas_bfw <Andreas.Schild <at> bfw.gv.at> wrote: > > Dear List, > > I'm a newby to Postgres/Postgis. > I have 2 feature classes, #)soilunits of austria 500000+ polygons which I > reduced to about 13000 multipolygons, > #)statistic units of austria 1300000 polygons (regular grid ) > I have to do an intersection on those feature classes. >(Continue reading)
The performance is exactly the same which actually puzzles me for
functions (namely ST_Within) that should return almost instantly ...
Problem being that as far as I can see there's no other function that
replaces these for the same effect...
(Call to developers)
This leads me to asking: isn't there room for improvement in these
functions?
The above functions are suitable for very small scale implementations
(like for fleets < 50 vehicles) but when it comes to real-world
implementations with fleets >200 vehicles using these functions almost
grinds the location implementation to a screeching halt...
All the best,
Pedro.
On Sat, 2008-08-30 at 08:59 -0400, Obe, Regina wrote:
> Pedro,
>
> Try using st_intersects instead of ST_within. In general I think it
> performs better than within and for point in poly will be just as
> good.
>
> Yah Guido's suggestion would be at best give same timing since
> ST_Within, ST_Intersects have an implicit && call in them so they can
> limit the number of more painful _ST_Within, _ST_Intersects checks
> they actually need to do.
>
> Hope that helps,
> Regina
>
>
>
> -----Original Message-----
> From: postgis-users-bounces <at> postgis.refractions.net on behalf of Pedro
> Doria Meunier
> Sent: Sat 8/30/2008 8:16 AM
> To: PostGIS Users Discussion
> Subject: RE: [postgis-users] st_within too slow?
>
> Dear Guido,
>
> Thanks for replying.
> Your suggested mod actually worsens the performance by another
> ~15-20ms
>
> Best regards,
> --
> Pedro Doria Meunier <pdoria <at> netmadeira.com>
>
> > Pedro,
> >
> > Try this instead:
> >
> > SELECT road_network, toponymy FROM maps AS m WHERE m.b_box &&
> > geomfromtext('POINT($lon $lat)',4326) and
> > st_within(geomfromtext('POINT($lon $lat)',4326), m.b_box)
> >
> > of course, if you already know that the point is in Portugal,
> > you could do:
> >
> > SELECT road_network, toponymy FROM maps AS m WHERE
> > m.country = 'PT' and
> > m.b_box && geomfromtext('POINT($lon $lat)',4326)
> >
> > (assuming that the maps table has the 'country' column with
> > country names).
> >
> > GL
>
>
>
>
>
> ______________________________________________________________________
>
> 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.
>
>
> ______________________________________________________________________
>
> Help make the earth a greener place. If at all possible resist
> printing this email and join us in saving paper.
>
> _______________________________________________
> postgis-users mailing list
> postgis-users <at> postgis.refractions.net
> 
RSS Feed