CGAL users discussion list

[Andreas Fabri <>]

On 23/05/2011 14:45, Costas Tsirogiannis wrote:
>   Greetings,
>
> Suppose that one uses CGAL::Lazy_exact_nt<CGAL::Gmpq> as a number type
> for some algorithm.
>
> If one asks for the byte size of an object of this class, then what is
> returned is a constant which is equal to the minimum number of bytes
> that are needed to represent a Gmpq object + the size of the fixed
> precision part of the number type.
>
> On the other hand, If one asks for the PRECISE byte size of the exact
> part + the approximate part of the lazy_exact number type if he uses a
> function like this:
>
>   int byte_size( NT a)
> { return sizeof(CGAL::Lazy_exact_nt<CGAL::Gmpq>) -
> sizeof(CGAL::Gmpq)+a.exact().size(); }  (That I learned from a previous
> question here, thanks :) )
>
>   But, in fact, the "exact().size()"  function forces the exact part of
> the object to be computed even if algebraic filtering was sufficient to
> avoid this computation so far. Thus, then we get an overestimation of
> what is computed by the algorithm.  It's like a "schrodinger"
> phenomenon: to read the byte-size, you have to force the computation of
> the exact part.
>
> Is there a way then to detect for a lazy_exact_nt object if the exact
> part is not computed so far? In other words, given a geometric structure
> whose coordinates are represented with this number type,  can one
> measure the number of bytes which are used for the structure's
> representation, no more, no less?

You don't ask for the size of the representation. You ask for the size
of the representation with GMP.

When you have the GMP 1/3 you also only want to know that it takes 2 
limbs, and not infinitely many characters if you write it as a string.

andreas


--
Andreas Fabri, PhD
Chief Officer, GeometryFactory
Editor, The CGAL Project

phone: +33.492.954.912    skype: andreas.fabri