On Wed, May 29, 2024 at 2:58 PM Arthur O'Dwyer wrote:
>
> Frederick's proposed solution (which, as I said, IMO is very bad)
> is that template type deduction should never deduce a
> (possibly cvref-qualified) specialization of `elide`.


What if we were to make it more generic, and not make it specifically
about std::elide?

What I mean is, instead of naming the tag "tag_elide", we call it
"tag_no_sole_param_constructor", so that in the future it can be used
to implement other new ideas.

Or if that doesn't sound appetising. . . here's another idea:

Similar to how deduction guides come after a class definition, what if
we could specify "template instantiation failure guides" for a class?
So you would write your class:

    template<typename F, typename... Params>
    class elide { . . . };

And then after the class definition, you would write a "template
instantiation failure guide" something like as follows:

    template< typename C, typename T> requires
std::is_specialization_v< std::remove_cvref_t<T>, std::elide >
    C::C(T&&) = false;

The above guide disallows the instantiation of any constructor which
takes exactly one parameter which is a specialisation of std::elide.