[boost] Re: Trait style (Was: container traits (was: Pre-submission: container algorithms))

From: Valentin Bonnard (Bonnard.V_at_[hidden])
Date: 1999-07-07 09:26:53

Ed Brey wrote:
> 
> Email seemed somewhat boring, until Valentin Bonnard wrote:
> >
> >See
> >http://www.eleves.ens.fr:8080/home/bonnard/ContAlgo/ContTraits.hpp
> >for an _example_ (I will probably rewrite it).
> 
> [And a quote from that example:
> 
> namespace TypeLogic {
> struct yes {};
> struct no {};
> struct maybe {};
> } // TypeLogic
> 
> // struct Container
> 
> template <typename T>
> struct container_trait
> {
>   static const bool is_specialised = false;
>   typedef TypeLogic::maybe is_sequence;
>   typedef TypeLogic::maybe is_associative;
>   typedef TypeLogic::maybe is_node_based;
> };
> 
> -- end quote]
> 
> What is the rational behind using the TypeLogic.  I would have expected
> something like this for the tri-state variables:
> 
> enum tristate {yes, no, maybe};
> 
> template <typename T>
> struct container_trait
> {
>   static const bool is_specialised = false;
>   static const tristate is_sequence = maybe;
>   static const tristate is_associative = maybe;
>   static const tristate is_node_based = maybe;
> };

Certainly a possible solution. But then auto-selection 
is less simple:

template <typename T, T value>
struct holder {};

void foo (holder<tristate, maybe>);

vs

void foo (maybe);

OTOH holder is more general (holder<int, N> vs 
struct one, struct two, struct three...).

FWIW, SGI uses my style (or I use SGI style).

-- 
Valentin Bonnard
------------------------------------------------------------------------
eGroups.com home: http://www.egroups.com/group/boost
http://www.egroups.com - Simplifying group communications