RE: [boost] Re: Re: Preprocessor Docs Comments

From: Brock Peabody (brock.peabody_at_[hidden])
Date: 2003-10-14 17:37:39

> -----Original Message-----
> From: boost-bounces_at_[hidden] [mailto:boost-bounces_at_[hidden]]
> On Behalf Of Edward Diener
> Sent: Tuesday, October 14, 2003 5:17 PM
> To: boost_at_[hidden]
> Subject: [boost] Re: Re: Preprocessor Docs Comments

[...]
> 
> What were you trying to do, and what did you find ?

I was trying to find a way to keep from writing N versions of
functions/templates when you need to have 0-N parameters:

   template <typename T0> struct t_0 {...};
   template <typename T0, typename T1> struct t_1 {...};

   ...

   template <typename T0, typename T1, ... typename Tn> struct t_n {...};

and

   template <typename T0> t_0<T0> f(T0) {...}
   template <typename T0, typename T1, t_1<T0,T1> f(T0,T1) {...}

   ...

   template <typename T0, typename T1, ... typename TN>
      t_n<T0,T1,... TN> f(T0,T1,... TN);

I ended up using CAT, ENUM, and ENUM_PARMS.  Boost PP made my code much
shorter and easier to maintain.

> I don't know of anyone who believes the preprocessor is evil, although I
> know Mr. Stroustrup has sought to steer C++ away from its reliance on the
> preprocessor. If it is worth it producing a preprocessor library which is
> usable by others, it is worth it producing a decent roadmap of its use and
> functionality.

IMO almost any solution is better than copy + paste.  Most of my reasons for
using it would go away if we had variable number of template parameters, and
typesafe variable number of function arguments.  I think the "Motivation"
section of the documentation gives a pretty good rational for this use.
Maybe there are other common uses for Boost PP that aren't as well
documented.

Brock