CGAL users discussion list

["Sebastien Loriot (GeometryFactory)" <>]

Try using a different kernel like this example:
http://doc.cgal.org/latest/Apollonius_graph_2/Apollonius_graph_2_2ag2_filtered_traits_no_hidden_8cpp-example.html

Sebastien.

On 12/18/2013 07:53 PM, 

 wrote:
> First off, just want to say that CGAL is an impressive piece of software!
>
> I'm having trouble with colinear points in an ApolloniusGraph2.
>
> It appears that inserting colinear points with an almost integer-valued weight
> can lead to a bad Apollonius graph.
>
> CGAL: 4.3
> compiler: gcc 4.8.1
> uname -a: Linux thistle 3.11.0-13-generic #20-Ubuntu SMP Wed Oct 23 07:38:26
> UTC 2013 x86_64 x86_64 x86_64 GNU/Linux
> cpu: intel xeon x5570
>
> installation:
> cmake -DCMAKE_C_COMPILER=gcc -DCMAKE_CXX_COMPILER=g++ .
> make
>
> To compile the test program, called cgal_create_CMakeLists, then same cmake
> command as above.
>
> test program:
> //----------------------
> #include <iostream>
> #include <CGAL/Exact_predicates_inexact_constructions_kernel.h>
> #include <CGAL/Apollonius_graph_2.h>
> #include <CGAL/Apollonius_graph_traits_2.h>
>
> typedef CGAL::Exact_predicates_inexact_constructions_kernel EPIC_Kernel;
> typedef EPIC_Kernel::Point_2 Point_2;
> typedef CGAL::Apollonius_graph_traits_2<EPIC_Kernel>   Traits;
> typedef CGAL::Apollonius_graph_2<Traits> AG;
> typedef AG::Site_2 Site_2;
>
> int main(int argc,char **argv) {
>    double weight;
>    AG          ag;
>
>    weight = 179.99999999999983;
>
>    ag.insert( Site_2( Point_2(0,-12),weight));
>    ag.insert( Site_2( Point_2(0,12),weight));
>    ag.insert( Site_2( Point_2(0,0),weight));
>
>    std::cout << ag ;
>
>    return 0;
> }
> //---------------
>
>
> The output:
> ----------
> 4 4 2
>
> 0 -12 180 0
> 0 12 180 0
> 0 0 180 0
>
> 1 0 3
> 1 2 0 # !! this edge is suspicious
> 0 1 3
> 2 1 0 # !! this edge is suspicious
>
> 2 2 1
> 3 0 3
> 0 0 3
> 2 1 1
> -------------
>
>  From a handful of test, it *appears* that the problem only arises when (a)
> weights are large compared to distance between vertices, (b) weight is close
> but not exactly an integer, and (c) the vertices are colinear.
>
> I don't know CGAL that well, but when my code was using a previous version of
> CGAL (but also older g++), it worked fine, so I'm wondering if this is a
> regression in CGAL.
>
> thanks for any help -
> Rusty