CGAL users discussion list

[Eric Berberich <>]

Flaminigo wrote:
> I use 
> OutputIterator	
> intersection (   Polygon_2<Kernel, Container> p1,
> Polygon_2<Kernel, Container> p2,
> OutputIterator oi)
> to compute intersection of two polygons. How ever it reports errors when
> computing the following two data:
>
>         Point data [] = {
>                         Point 
> (0.0779032708430652393039039793621,0.122371377632631991860812092909
> ),
>                         Point (13.2581684036192424258615574217, 
> 0.959769766188581341381791389722
> ),
>                         Point (1.83288664437794701633777094685 
> ,0.617138373712451504538023527857
> ),
>                         Point (0.693014198934081226965986388677 
> ,0.39778520059801764086060416048
> ),
>                         Point (0.463650350407879274783340406429 
> ,0.340347544290135750610915010839
> ),
>                         Point (0.0922023183796141926338307825972,
> 0.229473196453353733836877381691),
>                         Point( 0.0727179173985912918887208888918
> ,0.195167801755973285837697517309)
>         };
>
> Polygon p2(data, data + 4);
>
> Point points_b[] = { Point(0,0), Point(10,0), Point(10,20), Point(0,20)};
> Polygon p1 (points_b, points_b+4);
>
> std::ostringstream ostr(std::ostringstream::out);
> std::ostream_iterator<Polygon_h> sit = ostr;
> intersection(p1, p2, sit); 
>
> My kernel is 
> typedef CGAL::Cartesian<double> K;
> typedef CGAL::Polygon_2<K>                                                
> Polygon;
> typedef CGAL::Polygon_with_holes_2<K>                                     
> Polygon_h;
>
> The error is
> CGAL error: assertion violation!
> Expr: *slIter == curve
> File: /usr/include/CGAL/Basic_sweep_line_2.h
> Line: 675
> Explanation:
> terminate called after throwing an instance of 'CGAL::Assertion_exception'
>   what():  CGAL ERROR: assertion violation!
> Expr: *slIter == curve
> File: /usr/include/CGAL/Basic_sweep_line_2.h
> Line: 675
> Aborted

doubles are evil for geometry - as we deal with non-continuous functions 
that are in (near-)degenerate cases (probably here as well) extremely 
sensitive because of double's rounding.

Please check http://www.cgal.org/philosophy.html or 
http://www.mpi-inf.mpg.de/~kettner/proj/NonRobust/index.html for the 
anatomy of 'double' "arithmetic" in geometric settings.

eriC