Structured bindings for `extents`

[bernhardmgruber]

The extents class could provide support for structured bindings to improve its usability. This way, we could transform your example from:

  stdex::mdspan m{d.data(), stdex::extents{3, 3}};
  for (std::size_t i = 0; i < m.extent(0); ++i)
    for (std::size_t j = 0; j < m.extent(1); ++j) // was m.extent(0)

to:

  stdex::mdspan m{d.data(), stdex::extents{3, 3}};
  const auto [rows, cols] = m.extents();
  for (std::size_t i = 0; i < rows; ++i)
    for (std::size_t j = 0; j < cols; ++j)

See your modified example: https://godbolt.org/z/enYcvbG8a

It's not a huge improvement, but with increasing rank, it can save some redundant mentioning of m.extents(...). Or prevent accidentially passing the wrong index ;) It also helps when I pass an instance of extents around without its surrounding mdspan. Then instead of:

const auto x = e.extent(0);
const auto y = e.extent(1);
const auto z = e.extent(2);

I can write:

const auto [x, y, z] = e;

[crtrott]

This seems like a good extension. We can't change the current proposal to add stuff like this in time for C++23 however since the design approval cutoff is basically here. But we should be able to do this as a separate paper on top of P0009 targeting C++26.

[crtrott]

Not sure how this is possible. Isn't structured bindings always giving back the member thingies? (updated link for your exmaple btw: https://godbolt.org/z/bzTPb7K5Y)

[SoilRos]

@crtrott It does, is just that you specialized for int whereas you example uses long unsigned int. This is the fix: https://godbolt.org/z/cW6W3aGMM

[bernhardmgruber]

Btw, I have a draft proposal for this feature, but my PhD has kept me too busy the last weeks to send it for a first review to the mailing list. Here is what I got so far: https://api.csswg.org/bikeshed/?url=https://gist.githubusercontent.com/bernhardmgruber/3d8c422e2cea2df542411623dd443ca4/raw/afe327ffe7e089912301cbfdf83acb64ca1c6853/std_extents_structured_bindings.bs

I have some time next week, so I could start circulating the draft.

[bernhardmgruber]

The paper will become available as P2906 in the July mailing.

[mhoemmen]

@bernhardmgruber Thanks! : - D

[klimentyev]

@bernhardmgruber thanks for the proposal! I've tried to compile your examples on godbolt, but I've had errors. Do you happen to know why they don't compile?

[bernhardmgruber]

@klimentyev Thx! It seems mdspan was moved out of the std::experimental namespace and is now available just in std. I will update the paper. Here is a working example (experimental removed): https://godbolt.org/z/aqWEojToq

[klimentyev]

@bernhardmgruber thank you very much for the quick fix