LWG-617: [sequences.general] <array> claimed to define a template meeting sequence container requirements

[hubert-reinterpretcast]

This is now LWG-617.

[sequences.general] paragraph 1 claims that <array> is one of various headers that define class templates that meet the requirements for sequence containers.

<array> does not define such a template.

[JohelEGP]

Why not? Reading [sequence.reqmts], it seems the intent is for it to be one, which means this is not editorial:

In addition, array is provided as a sequence container which provides limited sequence operations because it has a fixed number of elements.

[hubert-reinterpretcast]

Just before the part you quoted, we have:

The library provides four basic kinds of sequence containers: vector, forward_list, list, and deque.

Thus, in context, the sentence you quoted is saying that, except for specific reasons why it isn't, array is a sequence container.

[JohelEGP]

except for specific reasons why it isn't

This formulation seems wrong, because it is, according to the wording. Can you show a requirement it fails to meet?

[hubert-reinterpretcast]

This formulation seems wrong, because it is, according to the wording. Can you show a requirement it fails to meet?

Just read [array.overview] paragraph 3?

[JohelEGP]

An array meets some of the requirements of a sequence container.

So [array.overview]p3 disagrees with [sequences.general]p1 and [sequence.reqmts]p1.

[CaseyCarter]

Meeting some of the requirements does not preclude meeting all of the requirements.

[JohelEGP]

According to [sequence.reqmts]p5, array does not meet the sequence container requirements. It can't meet the initialization requirements just after it.

5
#
A type X meets the sequence container requirements if X meets the container requirements and the following statements and expressions are well-formed and have the specified semantics.

So only [sequences.general]p1 is wrong:

1
#
The headers , , <forward_­list>, , and define class templates that meet the requirements for sequence containers.

[CaseyCarter]

array<T, N> for N > 0 doesn't meet [container.reqmts]/10, so many specializations of array are neither containers nor contiguous containers, which means [array.overview]/1 and /3 are contradictory as well.

I don't think we can fix this all editorially, this probably deserves an LWG issue. (I think we probably want to make array be a sequence container by changing the requirements rather than yanking all the wording that suggests array is a sequence container.)

[CaseyCarter]

This was LWG-617, which I believe was incorrectly resolved NAD.

[JohelEGP]

This was LWG-617, which I believe was incorrectly resolved NAD.

And now it has been reopened.