"Rozental, Gennadiy" <gennadiy.rozental@thomson.com> wrote [example snipped]
And give me single example where second design is preferable to first?
I agree absolutely. And with the input from contributions this thread under my belt I shall revisit numeric::converter. There are a large number of params there .. I guess you think too many?
Actually I think the best design would look like this:
[example snipped] Yes this is potentially cool. Unfortunately closest I could get was: struct multiply_t { template< typename A , typename B > struct sig : multiply_traits<A,B> {}; // fails because A and B cannot be deduced from a or b in the //function parameters(VC7.1, gcc 3.2) #if 0 template< typename A , typename B > typename sig<A,B>::result_type operator ()( typename sig<A,B>::first_argument_type a, typename sig<A,B>::second_argument_type b ) { return a * b; } #else // ok template< typename A , typename B > typename sig<A,B>::result_type operator ()( A const& a,B b ) //whatever { return a * b; } #endif }multiply; The unnammed namespace scheme I should think more about ... but off-topic. regards Andy Little