#!/usr/bin/env python3
"""Aggregate T1 extended benchmark results.
Prints a Markdown-style table sorted by F1 desc."""
import os
import json
import glob
import numpy as np
from collections import defaultdict

ROOT = '${PULSE_ROOT}/results/t1_extended'


def collect(pattern):
    by_key = defaultdict(list)
    for f in sorted(glob.glob(pattern)):
        try:
            r = json.load(open(f))
        except Exception as e:
            print(f"  ERR reading {f}: {e}")
            continue
        key = r.get('method', os.path.basename(os.path.dirname(f)))
        # Distinguish ablations by tag
        tag = r.get('args', {}).get('tag', '')
        if tag:
            key = f"{key}_{tag}"
        by_key[key].append(r)
    return by_key


def main():
    groups = collect(f'{ROOT}/*/results.json')
    rows = []
    for key, rs in groups.items():
        f1s = [r['test_f1'] for r in rs]
        accs = [r['test_acc'] for r in rs]
        mods = ','.join(rs[0]['modalities'])
        rows.append({
            'method': key,
            'modalities': mods,
            'n_seeds': len(rs),
            'f1_mean': np.mean(f1s),
            'f1_std': np.std(f1s),
            'acc_mean': np.mean(accs),
            'acc_std': np.std(accs),
            'n_params': rs[0].get('n_params', 0),
        })
    rows.sort(key=lambda r: r['f1_mean'], reverse=True)

    print(f"\n{'Method':<28s} {'Modalities':<32s}  N  {'F1 mean±std':<14s}  "
          f"{'Acc mean±std':<14s}  Params")
    print('-' * 110)
    for r in rows:
        print(f"{r['method']:<28s} {r['modalities']:<32s}  {r['n_seeds']}  "
              f"{r['f1_mean']:.3f}±{r['f1_std']:.3f}   "
              f"{r['acc_mean']:.3f}±{r['acc_std']:.3f}   "
              f"{r['n_params']:,}")


if __name__ == '__main__':
    main()