[geos-devel] QD (Quad-Double) Approach to Robustness

[Martin Davis]

> Ok, so back to original problem: do you obtain a successful 
> result with JTS against the XML test I attached ? GEOS is 
> successfull or not depending on wheter or not Coordinate copy 
> constructor and/or assignment operator is inlined.

Ok, I've run this case.  JTS produces an incorrect result for this test
case (using "intersection").  However, it throws an exception when using
"difference".

My theory (and hope) is that for cases which produce a wrong answer for
a particular overlay op, one of the other overlay ops will actually
throw an exception.  (This is because the underlying GeometryGraph is
topologically incorrect, it just doesn't cause a failure for some of the
overlay ops).  In this case maybe there's a way of checking the sanity
of the GeometryGraph, to produce an exception in ALL cases.

By the way, using my prototype snapping code this test case succeeds for
all overlay ops...

Martin Davis, Senior Technical Architect
Vivid Solutions Inc.      www.vividsolutions.com
Suite #1A-2328 Government Street Victoria, B.C. V8T 5G5
Phone: (250) 385 6040 - Local 308 Fax: (250) 385 6046