[postgis-users] Unexplained results when using Intersection?

[Craig Feuerherdt]

I have been attempting to intersect on set of geometries (bio.evc) with
another geometry (bio.bnd_buf). All the geometries are valid according to
isvalid and the query below successfully returns a table.

create table bio.bnd_evc as
select a.buf_dist, b.bio_no, b.bio_code, b.evc, b.constcde, b.consstat,
b.density, intersection(a.the_geom, b.the_geom) as the_geom
from bio.bnd_buf as a, bio.evc as b
where a.the_geom && b.the_geom
and a.buf_dist = 1
and overlaps(a.the_geom, b.the_geom)

When I view the resulting table (bnd_evc) in QGIS there are several missing
polygons (image 1a<http://www.bendigo-orienteers.com.au/study/evc1_over.png>).
One of the polygons lies entirely within bnd_bnf while the other has a small
portion which overlaps the boundary.

When I try intersecting the same geometries but this time selecting buf_dist
= 1000 (the 1000m buffer), additional evc polygons (which are wholly within
buf_dist=1 and therefore also within buf_dist=1000) are missing from the
output table and the whole thing is a 'mess' nad is in no way a
representation of what the intersection shoul dlook like (image
2a<http://www.bendigo-orienteers.com.au/study/evc1000_over.png>
).

The other thing that occurs in both cases is that some polygon geometries
are created where there are no evc polygons (image
1b<http://www.bendigo-orienteers.com.au/study/evc1_under.png>&
2b <http://www.bendigo-orienteers.com.au/study/evc1000_under.png>).

Both tables have a GIST index on the geometry column (the_geom).

Are the issues associated with having multiple polygon geometries in the one
table (unlikely in my opinion)?
Is it something to do with the query?
Are these issues associated with PostGIS or GEOS?
Is there a work around to solve the issue?

Thanks in advance,

Craig
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-users/attachments/20071031/ed83750e/attachment.html>