CGAL users discussion list

[Peter Hachenberger <>]

Hi Bart,

you found the main problem---the facets must be coplanar. I guess it's
not an option for you to first convert the coordinates to exact
coordinates and then do the rotation? That's easier than doing the
triangulation. 

BTW, why are you using the Extended_homogeneous kernel? This kernel is
very slow. Do you need to represent halfspaces? Otherwise, you should
use another kernel. 

Peter

On Sun, 2007-08-26 at 14:25 +0200, Bart Janssens wrote:
> On Saturday 25 August 2007 00:14:45 Bart Janssens wrote:
> > Hi all,
> >
> > I have made some more progress in using CGAL to do booleans in K-3D. I 
> > have
> > written a converter from k3d::mesh to CGAL::Nef_polyhedron_3, and it
> > appears to work on two cubes that have their edges parallel to the
> > coordinate system. I can calculate the difference of two such cubes, as
> > shown in the attached image. The cubes consist of 6 quadrilateral 
> > polygons,
> > so there is no triangulation prior to passing them to CGAL.
> >
> > The problem arises when I complicate matters only slightly, and attempt to
> > rotate one of the cubes. When I do that, I get the following error from
> > create_volumes() called by build_external_structure() called on the 
> > rotated
> > cube:
> 
> I've done some more checking, and after removing some glaring errors from 
> the 
> normal calculation part in my code, I can now rotate the cubes around a 
> single axis, though more complicated rotations definitely result in 
> non-planar polygons to appear.
> 
> I think I basically have two options here:
> 1. Attempt to make the corners of the polygons coplanar, i.e. by 
> calculating 
> intersections between planes surrounding the corner and then moving the 
> point 
> to the intersection. I am not sure if this can work in general, though.
> 
> 2. Triangulate all polyhedra before calling the boolean operations.
> 
> 3. A combination of the above: attempt 1, if it is detected to be 
> impossible, 
> use 2.
> 
> The reason I'm not going for triangulation right away is that the CGAL 
> booleans result in nice meshes that use a minimum of geometry to represent 
> the boolean mesh, which is much nicer to work with than a triangulated mesh.
> 
> Any recommendations welcome!
> 
> regards,
> 
> Bart