Fix nd_span iterators to respect strides of non-contiguous views

[Copilot]

nd_span::begin()/end() returned raw pointers, iterating linearly over raw memory regardless of strides. This made range-based for loops and STL algorithms produce wrong results for subspans and transposed views (any view with non-unit or non-contiguous strides).

Changes

include/nd_array/nd_array.hpp

• Added detail::nd_iterator<ElementType, MaxRank> stride-aware forward iterator template in the detail namespace, tracking multi-dimensional indices and computing element addresses as data + Σ(indices[i] × strides[i])
• Template parameter is the element type directly (Ty for mutable, const Ty for const) rather than a bool IsConst flag, making the iterator reusable by other classes in the same header
• Supports implicit conversion from iterator to const_iterator
• Handles edge cases: zero rank, zero-sized extents
• Added iterator / const_iterator type aliases in nd_span pointing to detail::nd_iterator
• Replaced raw-pointer begin()/end()/cbegin()/cend() with stride-aware iterator versions
• Added #include <iterator> for std::forward_iterator_tag

tests/test_nd_span.cpp

Added "nd_span - Stride-aware iteration" test case covering:

• Column subspan (non-unit stride in outer dimension)
• Transposed view (swapped strides)
• Write-through mutation only touches elements inside the subspan

Example

// 3×5 matrix, values 0..14
nd_span<int> span(data.data(), 3, 5);

// Subspan of columns 1–3: extents [3,3], strides [5,1]
auto sub = span.subspan(1, {1, 4});

// Previously: iterated over raw memory [data+0 .. data+9] — wrong values/elements
// Now: visits span(0,1), span(0,2), span(0,3), span(1,1), ... correctly
for (auto v : sub) { /* stride-aware */ }

Original prompt

---

This section details on the original issue you should resolve

<issue_title>nd_span iterators do not consider the strides of subspans</issue_title>
<issue_description>As the title states. When using begin or end in a nd_span, the stride of that span is not considered and it iterates linearly over the contained memory. This is not useful when creating subspans from nd_array of other nd_spans.</issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes nd_span iterators do not consider the strides of subspans #2

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.92%. Comparing base (d3e9064) to head (af0aa64).

Additional details and impacted files

@@           Coverage Diff           @@
##             main       #3   +/-   ##
=======================================
  Coverage   97.92%   97.92%           
=======================================
  Files           3        3           
  Lines        1206     1206           
=======================================
  Hits         1181     1181           
  Misses         25       25           

Flag	Coverage Δ
Windows-windows-debug	97.65% <0.00%> (ø)
Windows-windows-msvc-debug	97.91% <0.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[pingelit]

This looks good, some minor changes to the iterator class would be nice to encourage reuse.

[pingelit]

I think moving the iterator to the detail namespace would make sense here. That way, it could be reused more easily.

[Copilot]

Done in 29f04dd — the iterator is now detail::nd_iterator<ElementType, MaxRank> in the detail namespace, and nd_span exposes it via using iterator = detail::nd_iterator<Ty, MaxRank> and using const_iterator = detail::nd_iterator<const Ty, MaxRank>.

[pingelit]

I think I prefer if the tempate argument would just be the underlying type, either const or non-const, not a bool.

[Copilot]

Done in 29f04dd — the template parameter is now the element type directly (template<typename ElementType, size_t MaxRank>), so nd_iterator<Ty, MaxRank> is the mutable iterator and nd_iterator<const Ty, MaxRank> is the const iterator. The bool IsConst + std::conditional_t machinery is gone.