[geos-devel] Re: [GEOS] #356: multilinestring buffer artifacts

[Sandro Santilli]

On Sun, Feb 27, 2011 at 01:40:41PM -0800, Martin Davis wrote:
> The reason that precision reduction is invoked is probably because the 
> generated buffer curve has some very close vertices - which tends to 
> result in noding failures.  So this is operating according to current 
> design (although it might be nice to find a way to reduce the number of 
> very close vertices, e.g. by snapping).

Yeah. Only problem is that the final output in GEOS is just incorrect.
Rather than a polygon with an hole, it's a multipolygon of two 
polygons (one within the other). No, the inner polygon doesn't match
the expected hole either (unfortunately).

> While investigating this issue, I came across another one, which is more 
> severe.  If the output from this buffer operation is itself buffered by 
> 1, the result is incorrect - it contains a hole which should not be 
> present.  I"ve attached a test case. At the moment I don't have any 
> ideas about how to fix this.

Your test succeeds in GEOS (no hole):
 Polygon[ZM] with 1 rings
    ring 0 has 74 points

Maybe is some recent refactoring fixing mine and breaking yours ?

--strk;

  ()   Free GIS & Flash consultant/developer
  /\   http://strk.keybit.net/services.html