CGAL users discussion list

[Bart Janssens <>]

On Thu, Jan 29, 2009 at 7:04 PM, Peter Hachenberger
<>
 wrote:
> there are actually two sources of randomness in the construction of Nef
> polyhedra, which cause
> the orders of vertices, edges, facets, etc. to be inconsistent. I have one
> client who explicitly asked
> me to remove it, which I did for him. But it's still in the class as
> provided by us. That's on purpose.
> We use a search data structure that uses randomness to guarantee a good
> running time with
> an extremely high probability. Is this a problem?

Hi Peter,

Unfortunately, for K-3D it is a big problem. Due to the "pipelined"
architecture, any modifiers that are applied to the mesh after a
boolean operation rely on the order of the facets, edges and vertices.
For example, an extrude modifier is applied to face 3. When the
document is reloaded, the "pipeline" is executed, which means the
boolean is executed again, and then the extrusion. If face 3 is
another face now, the document will not look the same anymore.

This is in fact the reason why we moved away from GTS to CGAL, since
somewhere deep inside GTS points are ordered by memory pointers, which
resulted in (unfixable) inconsistent results. For us, it would be very
nice if there were an option to switch the randomness off.

Cheers,

-- 
Bart