Thierry GERAUD wrote:

Roland Levillain wrote:

...

There are still things to be improved, but at least we have (almost) all the desired functionality. The missing features that I can see ... +# define mlc_super_(T) \ + set_super_type<T>::ret

do we really need to distinguish between set_super and get_super?

that's an open question?

...

...

...

...

"Théo" == Thierry GERAUD <theo@lrde.epita.fr> writes:

Théo> Roland Levillain wrote:

...

There are still things to be improved, but at least we have (almost) all the desired functionality. The missing features that I can see ... +# define mlc_super_(T) \ + set_super_type<T>::ret

Théo> do we really need to distinguish between set_super and get_super? In fact, mlc_set_super may be absolutely useless, particularly if you consider template classes. In fact, I don't think it'd be worth to write shortcut macros like these: # define mlc_set_super_1(Type, Param, Super) \ template <typename Param> \ struct set_super_type<Type < Param> > \ { \ typedef Super<Param> ret; \ } # define mlc_set_super_2(Type, Param1, Param2, Super) \ template <typename Param1, typename Param2> \ struct set_super_type<Type < Param1, Param2 > > \ { \ typedef Super<Param1, Param2> ret; \ } (there are just too many cases!). In addition, mlc_super is just some sugar to shorten the code (avoid ::ret) when there is a single super class. I don't know if we can write an equivalent sugar for multiple inheritance. IMHO, we could get rid of mlc_set_super, but we should keep mlc_super/mlc_super_.