[boost] Re: typos in serialization/is_abstract.hpp

From: David Abrahams (dave_at_[hidden])
Date: 2003-12-10 13:51:02

"Rani Sharoni" <rani_sharoni_at_[hidden]> writes:

> Jonathan Turkanis wrote:
>> This is a small point, since the code in question is hidden with '#if
>> 0'; however, there seem to be several typos in the definition of
>> is_abstract.
> I posted possible implementation for this important trait (and bother Jason
> Shirk to implement it in VC7.1) while ago and wondered where can I see the
> one that you are referring to?
>
> Anyway, I simplify my original proposal some time ago and you might want to
> consider the new implementation (tested using Comeau 4.3.0.1 and VC7.1+ with
> boost 1.30.2):

In the interest of education and spreading good idiomatic MPL practice:

> #include "boost/type_traits/is_class.hpp"
> #include "boost/mpl/if.hpp"
>
> template<typename T>
> struct is_abstract_class
> {
> private:
>     typedef typename boost::mpl::if_c<boost::is_class<T>::value, T,
> int>::type TT;

      typedef typename boost::mpl::if_<boost::is_class<T>, T, int>::type TT;

>     typedef char (&yes)[1];
>     typedef char (&no) [2];
>
>     //
>     // Deduction fails if U is void, function type, reference type
> (14.8.2/2)
>     // or an abstract class type and according to WP issue #337
>     // http://anubis.dkuug.dk/jtc1/sc22/wg21/docs/cwg_active.html#337
>     // supported by EDG3.0+ and VC7.1+
>     //
>     template<typename U>
>     static no  check(U (*)[1]);
>
>     template<typename U>
>     static yes check(...);
>
> public:
>     static const bool value = sizeof(check<TT>(0)) == sizeof(yes);
> };
>

> I didn't find the type_traits self contained alternative for mpl::if_c.

IMO, why bother?  Those type_traits ice_whatever idioms are cumbersome
and we should eventually replace them with clean MPL idioms.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com