In addition to the above requirements, I expect the compiler to recursively instantiate template types, as in the dummy-example below:
template<typename T,unsigned Index,unsigned Position> struct dummy_encode2 { sizer<DOUBLE_ID>::type encode_value(sizer<Index> const*,sizer<Position>*); };
template<typename T,unsigned Index,unsigned Position=1> struct dummy_encode1 { BOOST_STATIC_CONSTANT(unsigned,instantiate=sizeof(dummy_encode2<T,Index,Position+1>)); sizer<POINTER_ID>::type encode_value(sizer<Index> const*,sizer<Position>*); };
Then the expression sizeof(dummy_encode1<double*,1>) should install two encode_value functions.
Sorry, I don't see any recursion here.
Here is a simplified example from my implementation: template<> struct encode_impl<POINTER_ID> { template<typename T,unsigned Index> struct encoder { enum(instantiate=sizeof(encode_modifier<T,Index+1>::type)); <- Recursion } }; template<typename T,unsigned Index> encode_impl<POINTER_ID>::encoder<T,Index> encode(void(*)(mpl::int_<POINTER_ID>,T*,mpl::int_<Index>)); where template<typename T,unsigned Index> struct encode_modifier { enum(instantiate=sizeof(encode(select_modifier<T>::type,T,mpl::int_<Index>))); <-Recursion }; Here: encode_modifier is instantiated with a type T. After analyzing the type, type deduction is done through the encode function. The encode function (or the sizeof operation) instantiates an encode_impl<N>::encoder<...> struct, which in turn instantiate one or more encode_modifier<...> structs which in turn ... What I fear is that some compiler will want to optimalize away the instantiate enums, as it is never referred to in the source. -- Peder Holt
-- Dave Abrahams Boost Consulting http://www.boost-consulting.com
_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost