CGAL users discussion list

[Marc Glisse <>]

On Fri, 27 Apr 2018, houes wrote:

> Marc,
>
> Thanks for the reply. I am using Epick.
>
> The planes are created separately. For example, a pentagon in a 3D plane has
> 5 vertices. I might use any 3 vertices of it to define a 3D plane. Such
> planes should be identical. And if there are two such planes, I need to
> remove them.

Well, you are in trouble because of the 'i' in Epick. You constructed a 
Plane_3, and even assuming that your points are exactly coplanar (not that 
likely), that construction was not exact. Even using the same 3 points in 
a different order might yield a slightly different plane.

If we assume that you were lucky and your Plane_3 are equivalent, you 
could consider hashing or sorting based on the tuple (p.a()/p.d(), 
p.b()/p.d(), p.c()/p.d()) (handle the case d=0 separately, or normalize 
differently). Then if you don't play with the rounding mode and don't hit 
any x87 weirdness, it may work.

Actually, operator== for PlaneC3 (the base of Plane_3) looks suspicious to 
me, I don't see where it handles the fact that the representation is not 
canonical (multiplying a, b, c and d by 2 keeps the same plane).

--
Marc Glisse