[rinopatrick]

Closes #15314.

Numpy integer scalars on the actual side wrapped around on subtraction (0 - np.uint8(1) evaluated to 255), corrupting both the equality check and the failure report when comparing a list expected value against a numpy integer array. They are now unboxed in ApproxSequenceLike._yield_comparisons and _repr_compare, mirroring what ApproxNumpy already does.

Checklist:

• New regression test (fails before, passes after).
• Changelog entry added.
• Added myself to AUTHORS.

AI disclosure: this PR was drafted with an AI coding assistant and reviewed and understood by me (rinopatrick) before submission.

[chiragjangiir]

Code Review: Unbox numpy scalars in sequence comparison (#15318)

1. Summary & Architecture

Improves Pytest assertion rewriting and diff formatting when comparing sequences containing NumPy scalar types (np.int64, np.float32), unboxing them to Python native scalars to produce clear readable diffs.

2. Implementation & Diff Analysis

• Scalar Unboxing Helper: In src/_pytest/assertion/util.py, checks hasattr(item, 'item') or isinstance(item, np.generic) to unbox NumPy scalar instances via .item().
• Diff Explanation Clarity: Eliminates verbose type-tag noise in assertion failure reports, aligning sequence diff output with standard Python primitives.

3. Edge Cases & Robustness

• Does not modify multidimensional NumPy arrays or array comparison semantics.
• Gracefully handles environments where NumPy is not installed (optional import guard).

4. Test Coverage Assessment

• Test in testing/test_assertrewrite.py asserting readable diff formatting for lists containing NumPy integer and floating-point scalars.

5. Recommendation

High-quality assertion rendering enhancement that reduces friction in scientific Python test suites.