Update to C++20

[tcbrindle]

Now that C++20 has been finalised, NanoRange should aim to match the spec as closely as possible.

Current task list:

• Remove forwarding_range, add borrowed_range (PR Use borrowed_range instead of forwarding_range #86)
• Remove boolean, add boolean_testable (PR Update core concept definitions to match C++20 specification #89)
• default_constructible -> default_initializable (PR Update core concept definitions to match C++20 specification #89)
• Support move-only views
• Remove converting constructors from various views (yay!) (PR Update constructors of views according to the latest C++ draft #99)
• Add ssize() (PR Add ssize() #87 #92)
• Update algorithm result types (again)
• Add the algorithms that weren't in P0896 (sample, for_each_n, clamp) (PR Implement clamp, for_each_n and sample algorithms #96)
• Liberally sprinkle [[nodiscard]] around the place

[bstaletic]

Add the algorithms that weren't in P0896 (sample, shift_left/right, any others?)

Those three don't seem to have been "rangified".

Update algorithm result types (again)

Here's a complete list:

• count and count_if for the range overload should return range_difference_t<R>, not iter_difference_t<iterator_t<R>>.

• max and min for the range overload should return range_value_t<R>, not iter_value_t<iterator_t<R>>.

• minmax for the range overload should return minmax_result<range_value_t<R>>, not minmax_result<iter_value_t<iterator_t<R>>>.

• minmax_element lacks an alias like the following

    template<class I>
    using minmax_element_result = minmax_result<I>;
  

• partial_sort_copy now has a partial_sort_copy_result return type, which is an alias for copy_result.

• partition, unique and stable_partition now returns a subrange<I> and borrowed_subrange_t<R>.

count, count_if, min/max "family" and partial_sort_copy were straight-forward to change.

unique, partition and stable_partition were more intrusive and even broke the tests.

Here is my incomplete patch. The tests are were patched in the dirtiest way possible to make them works again. However, I do get an error I can't make sense of. The gist contains the error too.

[tcbrindle]

Hi, thanks for the patch.

Add the algorithms that weren't in P0896 (sample, shift_left/right, any others?)

Those three don't seem to have been "rangified".

sample has (see here), but it seems the other two have not -- strange, I seem to recall there was a national body comment about this in Belfast, but never mind.

Regarding changing the result types, I was intending to add these types and making copy_result et al aliases for those. This can of course be done after your changes.

Would it be possible to submit your patch as a Github pull request rather than a diff, to allow a CI run? Different compilers might make it easier to diagnose the strange subrange errors.

Thanks again.

[bstaletic]

Regarding changing the result types, I was intending to add these types and making copy_result et al aliases for those. This can of course be done after your changes.

I was going after what's in cppreference, because I still don't know how to navigate eel.is efficiently. That's why I missed things.

Would it be possible to submit your patch as a Github pull request rather than a diff, to allow a CI run? Different compilers might make it easier to diagnose the strange subrange errors.

Absolutely, just trying to clean some stuff up first. I don't like submitting pull requests that don't compile.

[tcbrindle]

P1243 is the paper adding extra range algorithms for C++20, thanks @al-mission-2016 for the heads-up.

[bstaletic]

For clamp, for_each_n and sample, I have an implementation, but I have not written any tests.

https://gist.github.com/bstaletic/16f1118be9668584d9081ec9c25c5deb

[al-mission-2016]

Oh, bstaletic, go ahead, please. I'm glad the work is going to be done.

[bstaletic]

@al-mission-2016 Thanks for the heads up. I incorporated the changes that @tcbrindle asked for in #95, but I still have not written the tests.

Here's the branch: https://github.com/bstaletic/NanoRange/tree/new_algos

[tcbrindle]

@bstaletic Excellent!

For the tests, it's probably easiest to adapt CMCSTL2's sample.cpp (just overwrite the existing, outdated sample.cpp in NanoRange) and Range-V3's for_each_n.cpp, as I have done for most of the rest of NanoRange's tests. Unfortunately neither CMCSTL2 or Range-V3 seem to have clamp() yet, but libc++ has some tests here which you might want to use.

[bstaletic]

I've added the tests and submitted the pull request. #96

[bstaletic]

Regarding [algorithms.results], should the change be as minimal as possible (for example, only change copy_result to alias in_out_result) or more like this where every foo_result is an alias to something in [algorithms.results]?

[tcbrindle]

The second one. Feel free to submit that PR once #96 is merged.