CGAL users discussion list

[Andreas Fabri <>]

Thomas Zangl - Home wrote:
>  
> Hi!
>
> Another triangulation problem :-) I have this points, forming a shape:
>
> 50 10.2013
> 45.2806 6.60342e-07
> 49.4532 -1.88372e-06
> 47.419 -7.77607e-07
> 51.4652 2.69164e-06
> 53.5365 -1.68059e-06
> 55.2482 -0.0703464
>
> I insert() this points into a RegularTriangulation (with weight 1). I
> call Triangulation.is_valid();. Then I iterate over all edges and get
> their dual edge (Segment or Ray). E.g. the edge: 47.419 -7.77607e-07
> 55.2482 -0.0703464 has its dual edge: 48.4359 -322.535 46.3494 -554.754
>
> Can this be correct? The y-coordinate is a way toooo low?
>
> Best regards,


The Delaunay triangulation constructs correctly.
The dual vertices are constructed entities, that
is centers of circumcircles. With an .._exact_constructions
kernel they are probably correct in the exact arithmetic
but the double approximation is just too approximate,
because these predefined kernels use interval arithmetic
to approximate the coordinates of the vertex and the
to_double just takes the midpoint of the interval.

andreas