[geos-devel] [GEOS] #699: Optimize Geometry->Intersection with rectangular argument

[Sandro Santilli]

On Wed, Sep 10, 2014 at 08:08:27PM +0300, Mika Heiskanen wrote:
> Could you comment on how the returned multipolygons are invalid?
> This is of major interest to us too at FMI. I can pull the
> code and study the results myself, but likely not until
> next week.

The invalidity I was referring to in the ticket had to do with the
attempt at creating linear rings by closing linestring of 2 vertices,
creating a 3-vertex ring, which is obviously invalid (no area).

I've fixed that in code by keeping those lines as such rather than
attempting to create areas out of them.

Right now I'm trying to fix other cases that arise by running the
current RectangleIntersection code against the testcases
in tests/general/TestFunctionAA.xml (for example case13 test3).
Some of them may be due to my changes of considering boundary as
being "internal", if you have code your side to test it, the input is:

  Rectangle(0,0,10,10);
  "POLYGON((5 2,5 8,15 8,15 2,5 2),(8 4, 8 6, 10 6, 10 4, 8 4))"

--strk;