CGAL users discussion list

["FEYEL Frederic." <>]

Hello,

I made several experiments around your suggestions.

First of all, I really had to use an extended precision kernel,
otherwise all creation of the second
nef fail (this is normal, since it describes a thin structure).

Both the constructor (from a Polyhedron) and the OFF_to_nef_3 work.

On my problems, the computation of the substraction (N2-N1) is
reasonably fast. The creation
of the Nefs from the Polyhedrons using the constructor is much faster
that the OFF_to_nef_3 (of course,
the limitation are more restrictive using the constructor).

So I would say that in my case, the bottleneck is not the boolean
operation itself, but clearly the
creation of the Nefs.

Now, I want to mesh the surface of the resulting Nef. I transformed it
back to a Polyhedron.

I had a look at the mesher class, and a solution would perhaps to define
an implicit surface (and mesh
the 0 isovalue) by computing the signed distance (using a projection
operator) of any point to the polyhedron.

It seems however a little bit complex.

Do you have any ideas to change my Polyhedron so that it contains only
triangles (and not general N-nodes faces) ?

Thank you for everything !






Peter Hachenberger a écrit :
> Hi Frederic,
>
> your files do not work for two reasons:
>
> 1. If you use the homogeneous kernel and load double coordinates, the
> coordinates will all be rounded to integers. 
>
> 2. We only convert Polyhedron_3 to Nef_polyhedron_3, if the former is a
> 2-manifold, but your second file consists of several facets, if I see
> correctly. 
>
> You can do the following:
>
> - You can try the Kernels Cartesian<Gmpq> and Simple_cartesian<Gmpq>.
> Unfortunatly, Nef_3 is quite slow in combination with cartesian kernels
> at the moment. In the next release we will change that. 
>
> - You can try the function OFF_to_nef_3 (see documentation). This
> converter can handle nearly every off file, i.e., it handles double
> coordinate, wrong orientations, several surfaces with boundary, self
> intersecting surfaces, etc. Because of it's versatility, this function
> is not very fast.
>
> - We provide further undocumented constructors. Try
> Polygon_constructor.h for your second file. The usage is shown in
> examples/Nef_3/polygon_construction.h.
>
> -  I just wrote new I/O routines and can now convert coordinates. If I
> find time, I send you a small program that converts off files with
> doubles into our nef3 file format with homogeneous integer coordinates
> as long as the off file is a 2-manifold.
>
> Peter
>
>