extend comparator support for np.clongdouble

[aseembits93]

No description provided.

[aseembits93]

@claude review this

[claude]

Claude finished @aseembits93's task in 59s —— View job

---

Code Review

Checklist

• Read repository guidelines (CLAUDE.md)
• Review modified files
• Analyze code changes
• Provide feedback

---

Summary

This PR extends comparator support for numpy extended precision types like np.clongdouble and np.longdouble by replacing specific complex type checks with a more general np.complexfloating type check, and adds proper NaN handling.

Changes Overview

codeflash/verification/comparator.py (lines 215-216):

• Before: isinstance(orig, (np.floating, np.complex64, np.complex128))
• After: isinstance(orig, (np.floating, np.complexfloating))
• Also added equal_nan=True parameter to np.isclose() call

tests/test_comparator.py (+58 lines):

• Added test_numpy_extended_precision_types() to test np.clongdouble and np.longdouble
• Added test_numpy_typing_types() to test numpy.typing type aliases

---

Review Feedback

✅ Strengths

1. Better type coverage: Using np.complexfloating is more general and covers all complex floating-point types (including np.complex64, np.complex128, np.complex256, np.clongdouble) rather than hardcoding specific types.

2. Proper NaN handling: Adding equal_nan=True is important for consistent behavior with other numeric comparisons in the codebase.

3. Comprehensive tests: The new tests cover both basic equality and NaN handling for extended precision types.

4. Defensive imports: Tests properly handle cases where numpy or specific types aren't available.

---

⚠️ Issues & Suggestions

1. Test scope mismatch with PR title (Minor)

The PR title says "extend comparator support for np.clongdouble", but the tests also include test_numpy_typing_types() which tests npt.NDArray and npt.NBitBase. These typing constructs are not scalar values like np.clongdouble, but rather generic type aliases used for type hints.

Concern: The test at line 2651 (assert comparator(arr_type1, arr_type2)) compares type aliases, not actual data. Type aliases like npt.NDArray[np.float64] are generic aliases used in type hints - comparing them doesn't really test the comparator's ability to handle actual numpy array data with those dtypes.

Suggestion: Consider removing test_numpy_typing_types() or clarifying in the PR description why these typing constructs need special comparator support. The comparator is primarily for comparing runtime values, not type hints.

Reference: tests/test_comparator.py:2639-2662

2. Missing edge case: Complex with NaN imaginary part (Minor)

The test at line 2629 tests np.clongdouble(complex(np.nan, 2)) (NaN in real part), but doesn't test NaN in the imaginary part or both parts.

Suggestion: Add test cases for:

# NaN in imaginary part only
nan_c3 = np.clongdouble(complex(1, np.nan))
nan_c4 = np.clongdouble(complex(1, np.nan))
assert comparator(nan_c3, nan_c4)

# NaN in both parts
nan_c5 = np.clongdouble(complex(np.nan, np.nan))
nan_c6 = np.clongdouble(complex(np.nan, np.nan))
assert comparator(nan_c5, nan_c6)

Reference: tests/test_comparator.py:2628-2631

3. Consider testing infinity for extended precision types (Minor)

Since the comparator handles infinity for regular floats (lines 277-280 in comparator.py), extended precision types should also handle infinity correctly.

Suggestion: Add tests for infinity:

# Infinity tests for longdouble
inf_l1 = np.longdouble(np.inf)
inf_l2 = np.longdouble(np.inf)
assert comparator(inf_l1, inf_l2)

# Infinity tests for clongdouble
inf_c1 = np.clongdouble(complex(np.inf, 1))
inf_c2 = np.clongdouble(complex(np.inf, 1))
assert comparator(inf_c1, inf_c2)

Reference: tests/test_comparator.py:2607-2636

---

Verdict

Approve with minor suggestions ✅

The core change is sound and improves type coverage. The main code change is minimal and correct. The test issues are minor and don't affect the correctness of the implementation. Consider addressing the suggestions above for better test coverage and clarity.

--- | Branch: comparator-numpy-types