[tuple.apply] Redundant/inconsistent std:: qualification for std::get

[dangelog]

The definition of std::apply uses std::get, although spelling std:: shouldn't be necessary (due to the [contents]). In the very same section, std::make_from_tuple calls get without qualification. It would be nice to unify the two one way or another.

[jensmaurer]

The problem is that neither apply-impl nor make-from-tuple-impl are shown to be in namespace std, so that unqualified lookup doesn't actually find std::get.

@jwakely, what's your preference:

• put apply-impl and make-from-tuple-impl inside namespace std and get rid of std:: in std::get or
• leave the namespaces as-is and qualify get and index_sequence in make-from-tuple-impl and apply-impl.

[dangelog]

Hi,
That's a good point. I guess qualifying is a good idea, given also the presence of forward in there?

[jensmaurer]

No, std::forward and std::move are special (by convention in the standard; nothing special about them per se).

[CaseyCarter]

I'd prefer wrapping our exposition-only functions with namespace std {} to avoid confusion. I could live with fully-qualifying all names in such functions as if they were in the global namespace. I would rather not have some names qualified (std::get) and others not (size_t, index_sequence) without good reason. ("We always qualify move and forward" isn't a good reason, but that's really a separate issue.)

[jwakely]

I agree with Casey.