Ian McCulloch wrote:
[...]
If optional<> is to be useful for references, then it should have the same semantics that a class acting as a proxy-reference has. ie, if I use optional<> with a class like
template <typename T> struct Proxy { explicit Proxy(T& obj) : my_obj(&obj) {}
Proxy& operator=(T const& obj) { *my_obj = obj; return *this; }
// ...
T* my_obj; };
then optional<T&> should act as close to optional<proxy<T> > as possible.
If optional< proxy<T> > is used, for whatever proxy<T>, you can't assign a value of type T to the optional. You can only assign of value of proxy<T>, so it is the proxy assignment from another proxy, or it's copy ctor, what counts. The proxy assignment from T just won't participate. So the proxy as you've shown here just won't compile in optional<> becasue it has no copy ctor nor assignment from another proxy. The assignment that you wrote there, albeit correct for the proxy<> itself, just won't ever be used if you wrap it into optional<>.
ie., if optional<T&> does something different just because the template argument is a reference
It doesn't. optional<T&> is semantically equivalent to optional<reference_wrapper<T> >
I wanted an optional<T&> - but I do find the rebinding behaviour surprising.
Because you got confused with one additional level of indirection. Go through your example again and look what happens if a reference proxy is used as an argument for optional<>. Best Fernando Cacciola SciSoft http://fcacciola.50webs.com/