abellgithub
Delaunator creates badly formed models on some data
I've been testing Dealuanator-cpp and found it produces badly formed triangulations on certain data sets, example attached. Specifically, larger data sets with millimeter resolution that have closely spaced points along feature lines with large gaps inbetween, for example points along walls on the inside of a building. This can be mitigated to a large extent, but not entirely, by adding very small random offsets to the vertex position which leads me to suspect it could relate to the issue raised here, https://github.com/abellgithub/delaunator-cpp/issues/27 It doesn't seem to happen on smaller samples I've tried, so apologies for the large test case. I'll try to hunt down a much simpler data set to illustrate this problem better for testing purposes
