dabroz / csgjs-cpp Goto Github PK

csgjs_vector pos;
csgjs_vector normal;
csgjs_vector uv;****

If I have a cube that has a uniform scale of 1.0 and perform a union with a smaller cube that only intersects with one plane (side) of the first cube, the opposite side of the first cube also ends up with added vertices despite having no intersection. Is there a way to restrain the algorithm to the plane or planes that intersect?

The splits seem OK on the faces that required splitting, so I feel the algorithm could be improved instead of adding any post functions to reduce excessive vertices on a plane.

Hi !

I experienced csgjs to get stuck in build trying to split one polygon against itself over and over because the coplanar test fails in splitPolygon() because of precision issues. I was able to enhance it by making epsilon larger. But what solved it for me was not trying to split the polygon against itself at all. So in csgjs_csgnode::build() I replaced

	for (size_t i = 1; i < list.size(); i++)
		me->plane.splitPolygon(list[i], me->polygons, me->polygons, list_front, list_back);

with

	me->polygons.push_back(list[0]);
	for (size_t i = 1; i < list.size(); i++)
		me->plane.splitPolygon(list[i], me->polygons, me->polygons, list_front, list_back);

splitPolygons() would theoretically add the polygon list[0] to me->polygons if there weren't precision errors. So I just do this upfront and start with i = 1. I'm not sure if this breaks somethign else but it still looks good to me and I don't get these infinite loops.

I just want to say, thanks for this library, even if it's just a port from other language. You really saved me many hours of work (even though I already been struggling with CSG for past month). Because I found a bug in it - float variables makes too large meshes to crash on boolean operations - soon I will make some pull request to your library, even fork it with some examples. So for now - thanks from other Warsaw dweller! :) (I owe you a big beer)

Cheshire Cat