Add range_slice

[nmm0]

Fixes #472

[mhoemmen]

Should calculate_extents::value always return IndexType or cw of IndexType, instead of deducing the return type?

[crtrott]

Besides one place where I think we use the previous code, we also should name canonical_range_slice to canonical-slice-range to make it clearer that that thing doesn't return a range_slice. Note that the C++ draft has a typo where it uses both spellings ....

[crtrott]

This seems excessive, the previous code was perfectly fine or? I mean this thing does return extent_slice in the end doesn't it?

[crtrott]

The wording is not always the best way to implement something. We are making sometimes reuse decisions etc. to make the wording easier, more consistent, easier to understand. That doesn't mean one should implement it that way. So in this case I really don't see a reason why we would not use the previous code?

[nmm0]

I rewrote this part because I found it easier to understand and less lines of code.

I.e. the pair case is just the range_slice with a stride of 1

I can always revert it if you would like, in the end it works out to the same thing

[crtrott]

I think I'd rather have it reverted. It makes more sense to me to directly return the extent_slice.

[mhoemmen]

Just an optional suggestion -- sometimes return type deduction can slow down builds.

Suggested change

	// Some compilers do not do ctad properly with canonical_range_slice
	template<class OffsetType, class ExtentType, class StrideType>
	constexpr auto make_extent_slice(OffsetType offset, ExtentType extent, StrideType stride)
	{
	return extent_slice<OffsetType, ExtentType, StrideType>{offset, extent, stride};
	}
	// Some compilers do not do ctad properly with canonical_range_slice
	template<class OffsetType, class ExtentType, class StrideType>
	constexpr extent_slice<OffsetType, ExtentType, StrideType>
	make_extent_slice(OffsetType offset, ExtentType extent, StrideType stride)
	{
	return {offset, extent, stride};
	}

[Copilot]

🟡 Changes recommended

Static reversed ranges can produce incorrect extents, and range_slice does not enforce required index-like component types.

3 open findings

• Reject reversed static ranges to prevent unsigned extent underflow · New
• Enforce index-like component types for range_slice parameters · New
• Update stale test comment referencing strided_slice · New

What changed in this PR

Adds C++26 range_slice support alongside extent_slice throughout submdspan processing.

Changes:

• Introduces and canonicalizes range_slice.
• Updates mapping and extent deduction.
• Expands slice and submdspan tests.

File	Description
include/​experimental/​__p2630_bits/​strided_slice.hpp	Defines range_slice.
include/​experimental/​__p2630_bits/​submdspan_canonicalize_slices.hpp	Canonicalizes ranges into extent slices.
include/​experimental/​__p2630_bits/​submdspan_extents.hpp	Canonicalizes slices during extent deduction.
include/​experimental/​__p2630_bits/​submdspan_mapping.hpp	Recognizes contiguous range slices.
tests/​test_strided_slice.cpp	Tests slice construction and members.
tests/​test_canonicalize_slices.cpp	Tests dynamic and static canonicalization.
tests/​test_submdspan.cpp	Exercises ranges across layouts.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.