Please ignore the previous code. Kindly consider the following:
template<class T> struct type {};
// static_pointer_cast_impl overload for raw pointers template<class T, class U> inline T* static_pointer_cast_impl(U *ptr, type<T>) { return static_cast<T*>(ptr); }
// static_pointer_cast ADL dispatcher function template<class T, class U> inline T static_pointer_cast(U u) { return static_pointer_cast_impl(u, type<T>()); // ADL kicks in here }
// ...
template<class T> class smart_ptr { /* ... */ };
// static_pointer_cast_impl overload for raw smart_ptr template<class T, class U> inline smart_ptr<T> static_pointer_cast_impl(smart_ptr<U> const& p, type<T>) { return smart_ptr<T>(p.get()); }
Seems ok, although "type<T>" could be something like "cast_to_pointer<T>". Whatever we choose, I think we should add it to "shared_ptr" and "intrusive_ptr", and define well-known boost ADL casting functions. Current "xxx_pointer_cast" is in TR1, so I don't know if we have arrived late for an standard C++ ADL friendly cast functions. Peter? Regards, Ion