[postgis-users] st_union says geometries have different srids

[mchapman at texelinc.com]

Mike,

I'm not sure if this is related but I was using the same version of geos today but from c++ directly and I was also getting an error when trying to use geos union in a loop where I continually union the previous resulting union geometry to the next in the table.  It would always throw an error when parsing a multipolygon wkb saying that one of the linear rings was not closed (which it isn't).  When I just union two geoms then all unions work correctly.  The ring it complains about is a ring created from a previous union operation and it looks like the geos coordinate sequencer doesn't parse the ring correctly or the ring is failing on creation.  I validated my wkb binary and all rings in my table geometries are closed.  

I think this is a possible geos problem.  I will debug further tomorrow and let you know if I find the error in the geos code and send you the fix if I find one.

Best regards,
Martin
Sent via BlackBerry by AT&T

-----Original Message-----
From: Mike Leahy <mgleahy at alumni.uwaterloo.ca>

Date: Mon, 17 Mar 2008 17:14:31 
To:postgis-users at postgis.refractions.net
Subject: [postgis-users] st_union says geometries have different srids


Hello list,

I seem to be having trouble with st_union as an aggregate function. 
Sorry if this is a known issue, but a quick search didn't turn anything 
up similar.  Below is an example where I have two polygons next to each 
other

test=# create table test (gid serial primary key);
test=# select 
addgeometrycolumn('public','test','test_geom',4326,'MULTIPOLYGON',2);
test=# insert into test (test_geom) values ('srid=4326;MULTIPOLYGON(((1 
1,1 2,2 2,2 1,1 1)))');
test=# insert into test (test_geom) values ('srid=4326;MULTIPOLYGON(((2 
1,3 1,3 2,2 2,2 1)))');


The following will fail no matter what shapes/srid I'm using:

test=# select st_union(test_geom) from test;
ERROR:  Operation on mixed SRID geometries

But this works fine:

test=# select st_memunion(test_geom) from test;
 
 
 
 
    st_memunion 
 
 

------------------------
  0103000020E610000001...


Is there something in particular that I'm doing wrong?  This only seems 
to be an issue with st_union as an aggregate, if I use 
st_union(geometry,geometry), the result looks fine.  This is on a 
Fedora-7 (x86_64) machine with PostgreSQL 8.3, PostGIS 1.3.2, 
Proj-4.5.0, GEOS-3.0.0 - I have a similar setup on another machine 
running F8-x86_64 (except for Proj-4.6.0), and the failure is worse (the 
postgresql server segfaults with 'signal 11').  Would this be a bug, or 
have I committed some kind of user error?  Is there anything I should 
try to figure out where the source of my problem is?

Thanks in advance for any suggestions.

Mike
_______________________________________________
postgis-users mailing list
postgis-users at postgis.refractions.net
http://postgis.refractions.net/mailman/listinfo/postgis-users