CGAL users discussion list

[Andreas Fabri <>]

Eric Berberich wrote:
> Ben Strasser wrote:
> > Thanks,
> >
> > #include <CGAL/Cartesian.h>
> > #include <CGAL/MP_Float.h>
> > #include <CGAL/Quotient.h>
> >
> > typedef CGAL::Quotient<CGAL::MP_Float>        Number_type;
> > typedef CGAL::Cartesian<Number_type>        Kernel;
> > ...
> >
> > did the trick :)
>
> This is not a trick - it's essential for reliable geometric computing. 
> 'doubles' are evil for that. See the other mail I just sent a few 
> seconds ago.
>
> eriC

The maybe better trick would be to use 
CGAL::Exact_predicates_exact_constructions_kernel
because this should be faster than using an arbitrary precision number type 
all the time.

andreas