"""SSKIM LAB public aggregate checker. Python 3, standard library only.
Run: python recompute_policy_summary.py study1-policy-summary.csv
No network/API calls. This recomputes aggregate rates, not the original experiment.
"""
import argparse
import csv
import json
import re

FIELDS = ['policy', 'episodes', 'normalized_successes', 'strict_successes', 'recovery_calls']

def summarize(filename):
    with open(filename, encoding='utf-8-sig', newline='') as source:
        reader = csv.DictReader(source)
        if reader.fieldnames != FIELDS:
            raise ValueError('Unexpected CSV header or order')
        result, names = [], set()
        for row in reader:
            policy = row['policy']
            if not policy or policy in names or len(policy) > 40:
                raise ValueError('Empty, duplicate or overlong policy')
            names.add(policy)
            if None in row or any(not re.fullmatch(r'[0-9]+', row[k] or '') for k in FIELDS[1:]):
                raise ValueError('Counts must be non-negative integers')
            n, normalized, strict, calls = (int(row[k]) for k in FIELDS[1:])
            if n <= 0 or normalized > n or strict > n:
                raise ValueError('Invalid denominator or success count')
            result.append(dict(policy=policy, episodes=n,
                               normalized_rate_percent=normalized / n * 100,
                               strict_rate_percent=strict / n * 100,
                               normalized_failures=n - normalized,
                               strict_failures=n - strict,
                               recovery_calls=calls, calls_per_episode=calls / n))
        if not result:
            raise ValueError('No policy rows')
        return result

if __name__ == '__main__':
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument('csv_file')
    args = parser.parse_args()
    print(json.dumps(summarize(args.csv_file), ensure_ascii=False, indent=2))