[geos-devel] New overlay validation, call for tests

[Sean Gillies]

All the Shapely tests are passing, FWIW.


On Wed, Jul 31, 2013 at 3:51 AM, Sandro Santilli <strk at keybit.net> wrote:

> I've just committed in trunk a change in the overlay operation
> heuristic to reduce the possibility of getting invalidly noded
> output.
>
> We used to get unnoded output due to common-bits-reducer
> heuristic introducing invalidities when moving the output
> geometry back to its original position (the input geometries
> were translated near the origin to have more computation bits).
>
> The new code re-checks noding after shifting back and refuses
> to accept a result if it isn't properly noded. This is still not
> 100% accurate in that results with mixed-typed components are
> not really checked for self-intersection, so you can still get
> such cases.
>
> In addition, the code now checks if the input is valid before
> trying the various heuristics, so that you would get a message
> about that, which is usually easier to read. Example:
>
>  => select st_union('POLYGON((0 0, 10 10, 10 0, 0 10, 0 0))'::geometry,
>                     'POLYGON((0 0, 5 0, 5 5, 0 5, 0 0))'::geometry);
>  ERROR:  GEOSUnion: TopologyException: Input geom 0 is invalid:
> Self-intersection at or near point 5 5 at 5 5
>
> This will be in GEOS-3.4.0, which I'm trying to close ASAP, so please
> run your tests against the trunk version in order to catch any problem
> earlier.
>
> Thank you !
>
> --strk;
> _______________________________________________
> geos-devel mailing list
> geos-devel at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/geos-devel
>



-- 
Sean Gillies
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/geos-devel/attachments/20130731/ecf4e5a8/attachment.html>