----- Mensaje original ----- De: Arkadiy Vertleyb <vertleyb@hotmail.com> Fecha: Domingo, Marzo 28, 2004 5:47 pm Asunto: [boost] Re: Re: Formal Review: Indexed Set

...

"JOAQUIN LOPEZ MU?Z" <joaquin@tid.es> wrote:

...

IMHO nth_iterator_type and familiy are user-level facilities (similar vg to Boost.Tuple get<>) and accepting mpl integral constants instead of raw constants does not make much sense here. Please somebody correct me if I'm wrong.

One positive thing that could result from this would be the possibility of specializing the class to determine the type of the index, such as insteadof

nth_index_type<1>::type

to use

index_type<mpl::int_<1> >::type //rather clumzy, but doesn't clash with index_type<tag>::type

If you then create some typedefs (possibly with the preprocessor), somethinglike:

typedef mpl::int_<1> _1; ... typedef mpl::int_<10> _10;

This might look like:

index_type<_1>::type;

I more or less get your approach, but still don't see what would gain us wrt the current scheme. And those typedefs will probably we disliked by many here at the mailing list. Joaquín M López Muñoz Telefónica, Investigación y Desarrollo