On 4/7/2017 7:17 PM, Richard Smith wrote:
> On 7 April 2017 at 15:00, Agustín Bergé <agustinberge_at_[hidden]
> <mailto:agustinberge_at_[hidden]>> wrote:
>
> On 4/7/2017 5:19 PM, Richard Smith wrote:
> > On 30 March 2017 at 19:14, Agustín Bergé <agustinberge_at_[hidden] <mailto:agustinberge_at_[hidden]>
> > <mailto:agustinberge_at_[hidden] <mailto:agustinberge_at_[hidden]>>> wrote:
> >
> > LWG2911 gave us the new `is_aggregate` trait, but no feature-test macro.
> > I would like to suggest `__cpp_lib_is_aggregate`. What follows is a
> > motivational use case, courtesy of Jonathan Wakely:
> >
> > #include <vector>
> > template<typename T, typename... Args>
> > T make(Args&&... args)
> > {
> > #if __cpp_lib_is_aggregate
> > if constexpr (std::is_aggregate_v<T>)
> > return { std::forward<Args>(args)... };
> > else
> > #endif
> > return T(std::forward<Args>(args)...);
> > }
> > struct Agg { int i; };
> > int main()
> > {
> > auto v = make<std::vector<int>>(1, 2);
> > #if __cpp_lib_is_aggregate
> > // make<> only supports aggregates if std::is_aggregate is
> > available
> > auto a = make<Agg>(1);
> >
> >
> > What does the #else look like here? I'm not yet seeing how this feature
> > test macro is useful.
> >
> > #endif
> > }
>
> There is no #else form here, the example is depicting a program with
> decreased functionality. A user of `make` that depends on the feature
> simply uses it, and it results in a compilation error when the feature
> is not present:
>
> int main()
> {
> auto a = make<Agg>(1);
> }
>
> Whereas a user of `make` that does not depend on the feature simply does
> not use it, and lack of support for the feature does not result in a
> compilation error:
>
> int main()
> {
> auto v = make<std::vector<int>>(1, 2);
> }
>
> The feature test macro is needed for the implementation of `make`, in
> order to support both use cases.
>
>
> This 'make' seems like a terrible idea, but you've convinced me that
> people might make use of the feature test macro.

Perhaps we could improve the motivational example then, if needed?

The end goal is not to replace initialization with library machinery,
but to support aggregate initialization in `make_xxx`/`emplace`
functions in which initialization is performed by the library on behalf
of the user, without trapping on `std::initializer_list`.

Regards,

-- 
Agustín K-ballo Bergé
http://talesofcpp.fusionfenix.com