CGAL users discussion list

["Fred Dorosh" <>]

> I am getting really confused with this... because this mean that the
> problem only occurs when I convert my result Nefs back to Polyhedron
> then to my geometry, and then do the opposite process to get back to nef
> structure.

Hi Laure,

This paragraph sets off warning bells for me - if you are converting your
intermediate results back and forth from your own geometrical
representation, and that representation uses floating point coordinates,
you're likely to run into trouble.

The reason is that very small features can be created as byproducts of a
boolean operation. If you extract vertex coordinates from the precise
geometry kernel and round them into e.g. 32-bit floating point, these
small features often become corrupted and result in degenerate geometry.
For instance, two nearby vertices may become merged and create a
zero-length edge. Other less detectable problems can also occur. If you
keep the geometry in Nef representation until the final operation, you
avoid this problem.

Regards,
Fred