[postgis-users] Troubleshooting topology exception errors in large datasets

[Sandro Santilli]

It's not easy to isolate the problem, my bet is on Union geometry sometimes
returning an invalid (collapsed) polygon. There's a ticket for that on GEOS,
altought it may be reported as fixed in some cases.

The best thing would likely be adding debugging lines either in GEOS or
PostGIS to print the HEXWKB of the specific input which results in invalid
output and of the input which results in an exception in output.
There _may_ be some macro already available for that in GEOS's BinaryOp.h

A definitive fix to these issues would be exposing in PostGIS the possibility
to specify a precision model to work while performing the GEOS operations.
It's on my TODO list but I've no idea when I'll have time to do that.

--strk;

  ,------o-. 
  |   __/  |    Delivering high quality PostGIS 2.1
  |  / 2.1 |    http://strk.keybit.net - http://vizzuality.com
  `-o------'


On Thu, Jul 19, 2012 at 05:29:01PM -0700, THX1138 wrote:
> I have been trying to peform an ArcGIS-like erase in postgis (see example
> query at end of post).
> 
> However when I try to union all features in one table in preparation for a
> difference operation with another table I receive the following message of
> doom and despair:
> 
> ERROR:  GEOSUnaryUnion: TopologyException: found non-noded intersection
> between LINESTRING (3.09895e+06 1.01068e+07, 3.09896e+06 1.01068e+07) and
> LINESTRING (3.09896e+06 1.01068e+07, 3.09896e+06 1.01068e+07) at
> 3098955.3520329543 10106778.227583764
> 
> I have tested my geometry for invalid polygons, ran st_makevalid and even
> tried the buffer 0 trick, but to no avail.  If I do st_buffer(geom,.01) I
> manage to get through the operation with no errors, but I have to alter all
> of the geometries, not just the geometries in question -and I would rather
> not even do the latter if at all possible so that I don't change my
> geometries.
> 
> I would like to post a bug for this problem, but I know that the
> exception-throwing features are required to diagnose the problem. However in
> a dataset of 200,000 geometries it is impossible to know which features are
> causing the issue as the aforementioned error message gives no indication. 
> 
> I am wondering how I should go about capturing the geometries that cause the
> error in order to file a bug so that this issue can hopefully be easily
> resolved and I can move on with my work.
> 
> If this is not feasible, does anyone know of any work-arounds to pre-prepare
> valid geometries so that they will not throw these  types of errors?
> 
> My version of postgis is as follows:
> 
> POSTGIS="2.0.1SVN r9761" GEOS="3.3.5-CAPI-1.7.5" PROJ="Rel. 4.8.0, 6 March
> 2012" GDAL="GDAL 1.9.1, released 2012/05/15" LIBXML="2.7.8"
> LIBJSON="UNKNOWN" TOPOLOGY RASTER
> 
> The geometries that fail are probably very tiny slivers if that helps with a
> diagnosis.
> 
> Thanks,
> 
> THX1138
> 
> ================================================================================
> Example table structure:
> 
> CREATE TABLE features(
>    feature_id serial primary key,
>    feature_geom geometry(multipolygon,2277)
> );
> 
> 
> CREATE TABLE erase_features(
>   erase_id serial primary key,
>   feature_id integer,
>   erase_geom geometry(multipolygon,2277)
> );
> 
> 
> ===================================query=========================================
> 
> SELECT st_multi(st_difference(a.feature_geom,b.unioned_erase_geom))
> 
> FROM
> 
> features a,
> 
> (SELECT 
> erase_id,
> st_multi(st_union(erase_geom)) unioned_erase_geom
> 
> FROM
> 
> erase_features
> 
> GROUP BY erase_id) b
> 
> WHERE
> 
> (a.feature_id = b.feature_id) and
> st_intersects(a.feature_geom,b.unioned_erase_geom)
> 
> --
> View this message in context: http://postgis.17.n6.nabble.com/Troubleshooting-topology-exception-errors-in-large-datasets-tp4999035.html
> Sent from the PostGIS - User mailing list archive at Nabble.com.
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users