[WIP] add structured binding forward declaration

[sthalik]

This commit adds std::tuple_element and std::tuple_size forward declarations to avoid slurping in thousands of lines worth of STL headers. For unsupported STL versions, <array> is included instead.

[codecov[bot]]

Codecov Report

Base: 97.96% // Head: 97.96% // Increases project coverage by +0.00% :tada:

Coverage data is based on head (b771fe5) compared to base (6f115b2). Patch coverage: 100.00% of modified lines in pull request are covered.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #127   +/-   ##
=======================================
  Coverage   97.96%   97.96%           
=======================================
  Files         135      135           
  Lines       10940    10943    +3     
=======================================
+ Hits        10717    10720    +3     
  Misses        223      223           

Impacted Files	Coverage Δ
src/Corrade/Containers/OptionalStl.h	100.00% <100.00%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

:umbrella: View full report at Codecov.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.

[mosra]

Since I'm deep in the reviews already, let's unblock this too, finally (it's from February? wow :sweat_smile:).

As the change now boils down to #include <utility> to get declarations of std::tuple_element and std::tuple_size, I don't think it's needed to invent any special forward-declarating header for these two -- the size of <utility> is rather small, while maintenance cost of the forward-declaring header and potential for strange breakages after compiler upgrades would be far higher than that. So just drop the TupleStl.h altogether.

What's left is that I assume you'd want to use C++17 auto [a, b] = someType; for as many Corrade/Magnum types as possible, right? In case of Corrade that'd be mainly these I suppose:

• Containers::Pair, for which the std::tuple_whatever specializations could go into the already-existing Containers/PairStl.h, and a corresponding new C++17-only PairStlCpp17Test.cpp
• Containers::Triple, for which the std::tuple_whatever specializations could go into the already-existing Containers/TripleStl.h, and a corresponding new C++17-only TripleStlCpp17Test.cpp
• Containers::StaticArrayView, for which it could go into the already-existing Containers/ArrayViewStl.h, and a corresponding new ArrayViewStlCpp17Test.cpp

For completeness there's also Containers::StaticArray but it doesn't have a corresponding StaticArrayStl.h header yet, that one you'd have to create, and a new StaticArrayStlCpp17Test.cpp as well. (It's also fine if you implement just the Pair and Triple, those are the most important ones.)

It seems like a lot of new test files, but my assumption is that there will be more and more C++14/17/20-specific features (such as your constexpr additions), and the new test files will be reused for those as well.

The std::tuple_element / std::tuple_size is a C++11 feature already so I wouldn't hide them in any #if compiling_as_cpp17. Having them exposed unconditionally would mean std::get<N>() working with Corrade/Magnum types even under C++11, for example.

In case of Magnum I assume you'd want Math::Vector exposed this way, so that'd be a new VectorStl.h, and a corresponding new C++17-only VectorStlCpp17Test.cpp.

[sthalik]

The std::tuple_element / std::tuple_size is a C++11 feature already so I wouldn't hide them in any #if compiling_as_cpp17. Having them exposed unconditionally would mean std::get<N>() working with Corrade/Magnum types even under C++11, for example.

For types without an already-existing get<N> member function, this makes the implementation uglier since if constexpr is from C++17 and functions can't be partially specialized.

[mosra]

So, not going to do the specializations for other containers either? I'll convert that to an issue, then, in case it eventually becomes useful for someone else.

[sthalik]

I still plan on doing it for Pair and Triple at least, but have more pressing issues at this time.

[mosra]

For anyone still subscribed to this PR, a variant of the above together with structured bindings for StridedDimensions was added in 087ead32ce60655676ab0cb9f001e182aba1d4af and 5f4fc52ad2f07c4e022e8ba8e770144becb69b17.