data2dsl

Participant: gemini (AI agent)

Understanding

Resolve audit findings P1.2–P1.6 from docs/AUDYT_KODU_2026-08-28.md:

• Comparability validation (side, unit, version, semantics)
• Adapter correctness (OQL buses, Code2Schema entities, error status, zero numeric handling)
• Exact matching (actor word boundaries, SUMD exact key)
• Batch deduplication and ambiguity detection
• Evidence ID sanitization

Execution plan

1. Fix DeterministicComparator._is_compatible to validate side, unit, version, and window semantics.
2. Fix data2dsl_adapters.py actor regex, OQL buses attribute, SUMD key matching, and evidence IDs.
3. Fix data2dsl_skill.py normalizers (Code2Schema entities, explicit error status, coalesce numeric).
4. Fix data2dsl_batch.py duplicate observation handling with ambiguity detection.
5. Fix data2dsl_doctor.py and data2dsl_remediation.py fallback defaults.
6. Add dedicated regression tests in tests/test_audit_p1_verification.py.

Actual changes

• src/data2dsl_comparator.py: extended compatibility checks for side, unit, version, semantics.
• src/data2dsl_adapters.py: exact actor word boundary regex, OQL buses attribute fix, SUMD exact key match, sanitized evidence IDs.
• src/data2dsl_skill.py: fixed Code2Schema entities keyword, explicit error status respect, numeric 0 handling, and execute_compare missing side handling.
• src/data2dsl_batch.py: added _AMBIGUOUS sentinel and duplicate detection.
• src/data2dsl_doctor.py & src/data2dsl_remediation.py: safe fallbacks for missing observations and loose dicts.
• src/data2dsl_discovery.py: type annotations for mypy compliance.
• tests/test_audit_p1_verification.py: 10 dedicated regression tests.

Blockers

• None. 158 tests passing cleanly.