CGAL users discussion list

[Stephane Tayeb <>]

Ianic wrote:
> Good day to you,

Hi Yan,

> I would like to use the CGAL 3D meshing capabilities for a subsequent finite
> element analysis. Our FE code needs to read in information about the
> neighbors of every element in the mesh. So I would like to add this bit of
> output as a separate .neigh file to complement the .mesh file. The format
> that I have in my mind is what a tetgen mesher generates:
>
> ********************************************
> *  First line: <# of tetrahedra> <# of nei. per tet (always 4) >
> * Remaining lines list of # of neighbors:
> <tetrahedron #> <neighbor1> <neighbor2> <neighbor3> <neighbor4>
> ... 
>
> A .neigh file associates with each tetrahedron its neighbors (adjacent
> tetrahedra), which are indices into the corresponding .ele file. An index of
> -1 indicates no neighbor (because the tetrahedron is on a boundary of mesh
> domain). The first neighbor of tetrahedron i is opposite the first corner of
> tetrahedron i, and so on. It is output by TetGen when -n switch is used. 
> ********************************************
>
> Obviously, in my case there is no .ele file, but a corresponding section of
> .mesh file. Could you please give me a hint as to how to do this? I am new
> to CGAL and don't know yet how to iterate over cells and obtain the
> necessary connectivity data.

The mesh is stored in an object called c3t3 (of class 
Mesh_complex_3_in_triangulation_3).

If you have a look at the documentation of concept 
MeshComplex_3InTriangulation_3 you should find how to iterate over 
cells. To navigate from one Cell_handle to its neighbors, you could have 
a look in the Triangulation_3 manual.

Here is an incomplete bunch of code which may help you:
=======================
for ( C3T3::Cell_iterator cit = c3t3.cells_begin() ;
      cit != c3t3.cells_end() ; ++cit ) {
  const C3T3::Cell_handle& c = cit;
  // Iterate over c neighbors
  for ( int i=0 ; i<4 ; ++i ) {
    const C3T3::Cell_handle& neighbor = c->neighbor(i);
    if ( c3t3.is_in_complex(neighbor) { ... }
    [...]
  }
}
=======================

> Also, a question that I have is this - is it correct to assume that in the
> .mesh file cell indices are implicitly included by means of order in which
> cells are listed in a file?

I'm not sure to understand what you mean. The mesh file is constructed 
using the above Cell_iterator. Hence, the implicit indice of each cell 
(in the .mesh file) corresponds to its position following this iterator.


> And the last question - I work with CGAL on Windows and I have noticed that
> the mesh output is not unique. If I run it several times, it has roughly the
> same amount of vertices and cells, but they slightly differ every time. Is
> it a feature or a bug? :-)

This is probably not a bug. There are two sources of randomness in the 
mesh generation algorithm:
 - the initial points of the mesh are discovered shooting random rays.
 - the sliver perturbation algorithm (CGAL v3.6) contains some randomness.

If you insert the initial points by yourself and if you don't use sliver 
perturbation, then the output mesh should not differ from one run to 
another one.

> Sincerely,
> Yan

Best,
Stéphane.

--
Stephane Tayeb
Software engineer - INRIA Sophia Antipolis
Geometrica Project-Team