tests/test_normalization_benchmark.py

"""Versioned expectations for kink merge / normalization (see tests/fixtures/normalization_benchmark_v1.json).

The fixture includes two hand-authored clusters (Snuggles, FMF vs MFM) plus **20** reproducible
case-only duplicates sampled with ``random_seed`` from ``scripts/generate_normalization_benchmark_fixture.py``.
Metrics: each case is a cluster-level check (canonical id + duplicate set vs ``propose_merge_clusters``).
Regenerate the random block only when merge rules change intentionally: run the generator and review the diff.
"""
from __future__ import annotations

import json
from pathlib import Path

from backend.kink_merge import propose_merge_clusters

_FIXTURE = Path(__file__).resolve().parent / "fixtures" / "normalization_benchmark_v1.json"


def _fixture() -> dict:
    return json.loads(_FIXTURE.read_text(encoding="utf-8"))


def _cases() -> list[dict]:
    return _fixture()["merge_cluster_cases"]


def test_normalization_benchmark_fixture_has_twenty_seeded_random_cases() -> None:
    data = _fixture()
    assert data.get("random_case_only_count") == 20
    assert data.get("random_seed") == 42
    assert len(data["merge_cluster_cases"]) == 22


def test_normalization_benchmark_merge_clusters() -> None:
    """Each benchmark case must match propose_merge_clusters output (golden + ≥20 seeded samples)."""
    for case in _cases():
        rows = [tuple(r) for r in case["rows"]]
        clusters = propose_merge_clusters(rows)
        canon = case["expect_canonical_id"]
        matched = [c for c in clusters if c["canonical_id"] == canon]
        assert len(matched) == 1, (case, clusters)
        c0 = matched[0]
        assert set(c0["duplicate_ids"]) == set(case["expect_duplicate_ids"])
        for eid in case.get("expect_excluded_ids", []):
            assert eid not in c0["duplicate_ids"], c0
            assert eid != c0["canonical_id"], c0