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 random access iterator template in the detail namespace, tracking a flat linear index 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
• iterator_category is std::random_access_iterator_tag; full random access interface: operator-- (pre/post), operator+=/-=, operator+/-, operator[], operator</>/<=/>=, and iterator difference operator-
• 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; end()/cend() pass size() so the end iterator has a well-defined flat index for arithmetic
• Added #include <iterator> for iterator support types

tests/test_nd_span.cpp

• Added file-scope static_assert checks verifying that nd_span<int>::iterator and nd_span<int>::const_iterator have iterator_category == std::random_access_iterator_tag
• 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
  • Random access operations on a contiguous span (operator[], operator+/-, operator+=/-=, decrement, iterator difference, ordering)
  • Random access arithmetic over a non-contiguous 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 */ }

// Random access also works correctly
auto it = sub.begin();
it += 3;          // jump 3 elements forward
auto dist = sub.end() - sub.begin();  // == sub.size()

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

❌ Patch coverage is 98.34254% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 97.97%. Comparing base (d3e9064) to head (c5bf0a1).
⚠️ Report is 6 commits behind head on main.

Files with missing lines	Patch %	Lines
include/nd_array/nd_array.hpp	96.55%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main       #3      +/-   ##
==========================================
+ Coverage   97.92%   97.97%   +0.04%     
==========================================
  Files           3        3              
  Lines        1206     1382     +176     
==========================================
+ Hits         1181     1354     +173     
- Misses         25       28       +3     

Flag	Coverage Δ
Windows-windows-debug	97.70% <98.07%> (+0.04%)	⬆️
Windows-windows-msvc-debug	97.96% <98.34%> (+0.04%)	⬆️

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 it would be best if the iterator were random access or at least bidirectional. Also, test that the iterator is correctly implemented through a static_assert.

[Copilot]

I think it would be best if the iterator were random access or at least bidirectional. Also, test that the iterator is correctly implemented through a static_assert.

Done in c5bf0a1. nd_iterator is now a full random access iterator (std::random_access_iterator_tag) with operator--, operator+=/−=, operator+/−, operator[], operator</>/<=/>=, and iterator difference operator−. The iterator tracks a flat linear index internally so all arithmetic is O(1). Two static_asserts at file scope in the test file verify the iterator category at compile time, and new runtime sections exercise decrement, arithmetic, difference, and ordering.

[pingelit]

Great, this address all my ideas.