CGAL users discussion list

[Sylvain Pion <>]

Chaman Singh Verma wrote:
>     Does this means that we should never use CGAL::Cartesian<double> in
>     future ?

It depends on the problem.

My advice is: if you do not know what you are doing, you should most probably
not be using Cartesian<double>.

By "know what you are doing", I mean that one should understand all the
robustness issues potentially linked with the algorithms that one uses.
It can also depends on the kind of degeneracies you can expect from the data
sets your algorithms are going to run on.

If you know what you are doing, well, you don't need my advice :)

There is a good deal of literature on these robustness issues already, e.g. :

 "Classroom Examples of Robustness Problems in Geometric Computations"
 Lutz Kettner, Kurt Mehlhorn, Sylvain Pion, Stefan Schirra and Chee Yap,
 In Computational Geometry: Theory and Applications, volume 40, issue 1, 
pages 61-78, May 2008.

You can find other links on cgal's web site.

> With the modifications you made, the code worked. But I have one 
> question, with
> this kernel, how we specify the datatype i.e. double, float, int etc ?

Exact_predicates_inexact_constructions_kernel is only a typedef.
It's meant to quickly answer many robustness issues.  The type
used for the coordinates is double.  If you need something else,
just have a look at how this typedef is defined, and you will
find how to change this.  Namely, for choosing float, it will be:
Filtered_kernel< Simple_cartesian<float> >.

--
Sylvain Pion
INRIA Sophia-Antipolis
Geometrica Project-Team
CGAL, http://cgal.org/

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature