import importlib.util
import unittest
from pathlib import Path


def load_audit_runner():
    script_path = Path(__file__).resolve().parents[1] / "scripts" / "run_round05_1_selector_patch_audit.py"
    spec = importlib.util.spec_from_file_location("run_round05_1_selector_patch_audit", script_path)
    module = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(module)
    return module


class Round051SelectorPatchAuditTests(unittest.TestCase):
    def test_behavior_record_contains_diff_contract_fields(self):
        runner = load_audit_runner()

        record = runner.build_behavior_record(
            source="round04_blind_pool",
            case_id="R04-BI-010",
            input_text="谁负责这件事一直说不清楚。",
            selector_result={
                "selected_models": [
                    {
                        "model_id": "qpi",
                        "score": 0.57,
                        "reasons": ["QPI problem-definition gate matched"],
                        "penalties": [],
                    }
                ],
                "rejected_models": [
                    {
                        "model_id": "intellectual_archaeology",
                        "score": 0.12,
                        "reasons": [],
                        "penalties": ["no IA heavy-depth gate"],
                    }
                ],
                "no_call": False,
            },
            targeted=True,
        )

        self.assertEqual(record["case_id"], "R04-BI-010")
        self.assertEqual(record["source"], "round04_blind_pool")
        self.assertEqual(record["selected_model_ids"], ["qpi"])
        self.assertEqual(record["rejected_model_ids"], ["intellectual_archaeology"])
        self.assertEqual(record["scores"]["qpi"], 0.57)
        self.assertEqual(record["scores"]["intellectual_archaeology"], 0.12)
        self.assertEqual(record["reasons"]["qpi"], ["QPI problem-definition gate matched"])
        self.assertEqual(record["penalties"]["intellectual_archaeology"], ["no IA heavy-depth gate"])
        self.assertIs(record["no_call"], False)
        self.assertIs(record["targeted"], True)

    def test_diff_marks_target_and_non_target_changes_with_deferred_disposition(self):
        runner = load_audit_runner()

        before = [
            {
                "case_id": "R04-BI-010",
                "source": "round04_blind_pool",
                "input": "谁负责这件事一直说不清楚。",
                "targeted": True,
                "selected_model_ids": [],
                "rejected_model_ids": ["qpi", "intellectual_archaeology"],
                "scores": {"qpi": 0.31, "intellectual_archaeology": 0.0},
                "no_call": True,
                "reasons": {"qpi": [], "intellectual_archaeology": []},
                "penalties": {"qpi": ["QPI positive signal missing"], "intellectual_archaeology": []},
            },
            {
                "case_id": "R04-BI-028",
                "source": "round04_blind_pool",
                "input": "我又卡住了。",
                "targeted": False,
                "selected_model_ids": [],
                "rejected_model_ids": ["qpi", "intellectual_archaeology"],
                "scores": {"qpi": 0.2, "intellectual_archaeology": 0.0},
                "no_call": True,
                "reasons": {"qpi": [], "intellectual_archaeology": []},
                "penalties": {"qpi": ["QPI positive signal missing"], "intellectual_archaeology": []},
            },
        ]
        after = [
            {
                **before[0],
                "selected_model_ids": ["qpi"],
                "rejected_model_ids": ["intellectual_archaeology"],
                "scores": {"qpi": 0.56, "intellectual_archaeology": 0.0},
                "no_call": False,
                "reasons": {"qpi": ["QPI problem-definition gate matched"], "intellectual_archaeology": []},
                "penalties": {"qpi": [], "intellectual_archaeology": []},
            },
            {
                **before[1],
                "scores": {"qpi": 0.25, "intellectual_archaeology": 0.0},
                "penalties": {"qpi": [], "intellectual_archaeology": []},
            },
        ]

        diff = runner.diff_behavior_records(before, after, targeted_cases={"R04-BI-010"})

        self.assertEqual(diff["summary"]["targeted_changes"], 1)
        self.assertEqual(diff["summary"]["non_target_changes"], 1)
        target_change = diff["changed_cases"][0]
        self.assertEqual(target_change["case_id"], "R04-BI-010")
        self.assertIs(target_change["targeted"], True)
        self.assertIn("selected_model_ids", target_change["changed_fields"])
        self.assertIn("no_call", target_change["changed_fields"])

        non_target_change = diff["changed_cases"][1]
        self.assertEqual(non_target_change["case_id"], "R04-BI-028")
        self.assertIs(non_target_change["targeted"], False)
        self.assertEqual(non_target_change["owner_disposition"], "defer")
        self.assertEqual(non_target_change["ccra_disposition"], "defer")


if __name__ == "__main__":
    unittest.main()