"""생성된 페어 데이터셋으로 탐지 엔진 정밀도/재현율 평가.

성능지표 #4 (계획서 p.23): "표절 여부 판별 정밀도(precision)" 자체 평가.

사용:
  python scripts/evaluate_pairs.py --pairs data/training/pairs.jsonl
"""

from __future__ import annotations

import argparse
import json
import logging
import sys
from pathlib import Path

logging.basicConfig(level=logging.INFO, format="%(asctime)s %(levelname)s: %(message)s")
logger = logging.getLogger("eval-pairs")

ROOT = Path(__file__).resolve().parent.parent
sys.path.insert(0, str(ROOT))

from app.engine.detector import PlagiarismDetector  # noqa: E402


def main() -> int:
    parser = argparse.ArgumentParser()
    parser.add_argument("--pairs", default=str(ROOT / "data/training/pairs.jsonl"))
    parser.add_argument("--threshold", type=float, default=0.30)
    args = parser.parse_args()

    pairs_path = Path(args.pairs)
    if not pairs_path.exists():
        logger.error("Pairs file not found: %s (먼저 generate_plagiarism_pairs.py 실행)", pairs_path)
        return 1

    detector = PlagiarismDetector()
    logger.info("Engine ready (corpus_size=%d)", detector.corpus_size)

    tp = fp = tn = fn = 0
    by_transformation: dict[str, dict[str, int]] = {}

    with pairs_path.open("r", encoding="utf-8") as f:
        for i, line in enumerate(f, 1):
            row = json.loads(line)
            transformation = row.get("transformation", "unknown")
            expected = row["is_plagiarism"]
            result = detector.detect(
                doc_id=row["pair_id"],
                text=row["derived_text"],
            )
            predicted = result.is_infringement and result.confidence >= args.threshold

            if expected and predicted:
                tp += 1
            elif expected and not predicted:
                fn += 1
            elif not expected and predicted:
                fp += 1
            else:
                tn += 1

            bucket = by_transformation.setdefault(
                transformation, {"tp": 0, "fp": 0, "tn": 0, "fn": 0}
            )
            if expected and predicted:
                bucket["tp"] += 1
            elif expected and not predicted:
                bucket["fn"] += 1
            elif not expected and predicted:
                bucket["fp"] += 1
            else:
                bucket["tn"] += 1

            if i % 10 == 0:
                logger.info("[%d] processed...", i)

    precision = tp / (tp + fp) if (tp + fp) else 0.0
    recall = tp / (tp + fn) if (tp + fn) else 0.0
    f1 = 2 * precision * recall / (precision + recall) if (precision + recall) else 0.0

    print()
    print("=" * 60)
    print(f"전체 정밀도 (precision): {precision:.4f}   (목표 0.95)")
    print(f"재현율  (recall):        {recall:.4f}")
    print(f"F1:                       {f1:.4f}")
    print(f"TP={tp}  FP={fp}  TN={tn}  FN={fn}")
    print()
    print("[변형 유형별]")
    for t, b in by_transformation.items():
        total = sum(b.values())
        prec_t = b["tp"] / (b["tp"] + b["fp"]) if (b["tp"] + b["fp"]) else 0.0
        rec_t = b["tp"] / (b["tp"] + b["fn"]) if (b["tp"] + b["fn"]) else 0.0
        print(f"  {t:18s}  n={total:3d}  P={prec_t:.3f}  R={rec_t:.3f}  "
              f"TP={b['tp']} FP={b['fp']} TN={b['tn']} FN={b['fn']}")
    return 0


if __name__ == "__main__":
    raise SystemExit(main())