[geos-devel] Disconnected interior failure?

Mika Heiskanen mika.heiskanen at fmi.fi
Sat Sep 5 06:58:34 PDT 2015


After a long busy summer I finally had the time to revisit this
problem, and I'm happy to report that geos got it right. There was an
extra polygon inside the hole which our topology building code did not
handle correctly. Sorry for the noise.


Mika Heiskanen / FMI

On 06/17/2015 08:15 AM, Mika Heiskanen wrote:
> No. However, the input set is such that any intersections can occur
> only at vertices. The holes touch at a single vertex, as does
> the exterior with another exterior. I will see if I can reduce
> the problem to a simpler test case.
> - Mika
> On Tue, 2015-06-16 at 21:11 +0300, Sandro Santilli wrote:
>> Had you checked if the two shell's interior-interior intersection
>> is empty and the boundary-boundary intersection is puntual,
>> with the Relate operation ?
>> --strk;
>> On Fri, Jun 05, 2015 at 01:38:39PM +0300, Mika Heiskanen wrote:
>>> Hi,
>>> We recently encountered a temperature isoband which GEOS isValidOp
>>> claims is invalid:
>>>    Interior is disconnected at or near point 16.56825286557833 -6
>>> The geometry consists of a shell plus two holes. The holes touch at
>>> a single point, but do not touch the shell. The description does not
>>> match to what the ConnectedInteriorTester class reference says:
>>>     An area Geometry is invalid if the interior is disconnected.
>>>     This can happen if:
>>>     * one or more holes either form a chain touching the shell
>>>       at two places
>>>     * one or more holes form a ring around a portion of the interior
>>> I extracted the problematic shell plus the holes from the global data,
>>> and made them temporarily available from http://beta.fmi.fi/geos/
>>> QGis and MapShaper validators accept the shape made from the WKT
>>> printed by GEOS. Could there be a bug in the GEOS validator?
>>> Regards,
>>> Mika Heiskanen / FMI
>> _______________________________________________
>> geos-devel mailing list
>> geos-devel at lists.osgeo.org
>> http://lists.osgeo.org/mailman/listinfo/geos-devel

More information about the geos-devel mailing list