Re: [Boost-users] another metaprogramming question

Subject: Re: [Boost-users] another metaprogramming question
From: Steven Watanabe (watanabesj_at_[hidden])
Date: 2009-03-13 16:36:50

AMDG

Noah Roberts wrote:
> This seems to me like it should work.  Yet the check() call resolves, 
> without error, to the first one...which should cause an error and 
> trigger SFINAE.  Can anyone help me figure out what gives?
>
> Compiler is MSVC++ 8.  Normally the SFINAE stuff works so I'm betting 
> on my mistake this time.
>
> Basically I need to check it the metafunction meta<T> returns 
> something.  If it does then the object is a record.  If it doesn't 
> then it is not.

What do you mean by "returns something"?
meta<T>::type always exists, so SFINAE doesn't apply.

> template < typename RECORD >
> struct meta
> {
>   struct type;
> };
>
> <snip>
>
> struct test { typedef boost::mpl::vector< int, double > fields; };

Would this work:

BOOST_MPL_HAS_XXX_TRAITS_DEF(fields);
template<typename RECORD>
struct field_is_sequence : boost::mpl::is_sequence<typename 
RECORD::fields> {};
template<typename RECORD>
struct meta : boost::mpl::and_<has_fields<RECORD>, 
field_is_sequence<RECORD> > {};

> int main()
> {
>   // output is 1\n1\n.
>   std::cout << is_record_check<test>::value << std::endl;
>   std::cout << is_record_check<int>::value << std::endl;

meta<int>::type exists, and since only a pointer to
it is used it isn't instantiated.

>   // meta<int>::type x; - yet this will cause compilation error.

Now meta<int>::type is instantiated, so we get a compilation error.

In Christ,
Steven Watanabe