scorer.py

from __future__ import annotations

import csv
import json
import sys
from pathlib import Path
from typing import Dict, List, Tuple


PREDICTION_COLUMN_CANDIDATES = [
    "prediction",
    "pred",
    "predicted_label",
    "label",
]


def _to_int(value: str) -> int:
    value = str(value).strip()
    return 1 if value in {"1", "1.0", "true", "True"} else 0


def _load_csv(path: Path) -> List[Dict[str, str]]:
    with path.open("r", encoding="utf-8", newline="") as f:
        return list(csv.DictReader(f))


def _find_label_column(rows: List[Dict[str, str]]) -> str:
    if not rows:
        raise ValueError("Ground truth file is empty.")
    label_cols = [c for c in rows[0].keys() if c.startswith("label_")]
    if len(label_cols) == 1:
        return label_cols[0]
    raise ValueError(f"Expected one label column, found: {label_cols}")


def _find_prediction_column(rows: List[Dict[str, str]]) -> str:
    if not rows:
        raise ValueError("Prediction file is empty.")

    available = set(rows[0].keys())

    for col in PREDICTION_COLUMN_CANDIDATES:
        if col in available:
            return col

    label_like_cols = [c for c in rows[0].keys() if c.startswith("label_")]
    if len(label_like_cols) == 1:
        return label_like_cols[0]

    raise ValueError(
        "No prediction column found. Expected one of "
        f"{PREDICTION_COLUMN_CANDIDATES} or a single label_* column."
    )


def _confusion(y_true: List[int], y_pred: List[int]) -> Tuple[int, int, int, int]:
    tp = fp = tn = fn = 0
    for yt, yp in zip(y_true, y_pred):
        if yt == 1 and yp == 1:
            tp += 1
        elif yt == 0 and yp == 1:
            fp += 1
        elif yt == 0 and yp == 0:
            tn += 1
        elif yt == 1 and yp == 0:
            fn += 1
    return tp, fp, tn, fn


def _safe_div(n: float, d: float) -> float:
    return n / d if d else 0.0


def score(ground_truth_path: str, prediction_path: str) -> Dict[str, object]:
    gt_rows = _load_csv(Path(ground_truth_path))
    pred_rows = _load_csv(Path(prediction_path))

    if len(gt_rows) != len(pred_rows):
        raise ValueError(
            f"Row count mismatch: ground truth has {len(gt_rows)} rows, "
            f"predictions have {len(pred_rows)} rows."
        )

    label_column = _find_label_column(gt_rows)
    pred_col = _find_prediction_column(pred_rows)

    y_true = [_to_int(row[label_column]) for row in gt_rows]
    y_pred = [_to_int(row[pred_col]) for row in pred_rows]

    tp, fp, tn, fn = _confusion(y_true, y_pred)

    accuracy = _safe_div(tp + tn, tp + tn + fp + fn)
    precision = _safe_div(tp, tp + fp)
    recall_boundary_detection = _safe_div(tp, tp + fn)
    false_safe_rate = _safe_div(fn, fn + tp)
    f1 = _safe_div(
        2 * precision * recall_boundary_detection,
        precision + recall_boundary_detection,
    )

    return {
        "label_column": label_column,
        "prediction_column": pred_col,
        "accuracy": round(accuracy, 6),
        "precision": round(precision, 6),
        "recall_boundary_detection": round(recall_boundary_detection, 6),
        "false_safe_rate": round(false_safe_rate, 6),
        "f1": round(f1, 6),
        "confusion_matrix": {
            "tp": tp,
            "fp": fp,
            "tn": tn,
            "fn": fn,
        },
        "primary_metric": "recall_boundary_detection",
        "secondary_metric": "false_safe_rate",
    }


if __name__ == "__main__":
    if len(sys.argv) != 3:
        raise SystemExit("Usage: python scorer.py <ground_truth.csv> <predictions.csv>")

    results = score(sys.argv[1], sys.argv[2])
    print(json.dumps(results, indent=2))