"""Independent public-input recomputation of candidate rules and 24h event outcomes."""
import csv,gzip,json,statistics,sys,math
from collections import Counter
from pathlib import Path
base=Path(sys.argv[1]) if len(sys.argv)>1 else Path(__file__).resolve().parent
report=json.loads((base/'research.json').read_text())
with gzip.open(base/'inputs.csv.gz','rt',newline='') as f: raw=list(csv.DictReader(f))
with (base/'events.csv').open(newline='') as f: events=list(csv.DictReader(f))
with gzip.open(base/'candidate-ledger.csv.gz','rt',newline='') as f: candidates=list(csv.DictReader(f))
with gzip.open(base/'controls.csv.gz','rt',newline='') as f: controls=list(csv.DictReader(f))
series={s:{} for s in report['symbols']}
for row in raw:
    series[row['symbol']][int(float(row['opened']))]=row
assert len(events)==report['retained_events']
assert Counter(r['status'] for r in candidates)==Counter(report['candidate_status_counts'])
assert {r['id'] for r in events} == {r['id'] for r in candidates if r['status']=='retained_event'}
checked=0
for row in events:
    s=series[row['symbol']];t=int(float(row['event_open']))
    prev=[float(s[t-i*3600]['qv']) for i in range(1,25)]
    r=s[t];ratio=float(r['qv'])/statistics.median(prev)
    assert math.isclose(ratio,float(row['relative_volume']),rel_tol=0,abs_tol=1e-9)
    move=100*(float(r['c'])/float(r['o'])-1)
    assert abs(move)<=.25 and ratio>=2
    assert math.isclose(move,float(row['candle_return_pct']),rel_tol=0,abs_tol=1e-9)
    for h in (1,4,24):
        key=f'return_{h}h_pct'
        if row[key]:
            assert all(t+i*3600 in s for i in range(1,h+1))
            result=100*(float(s[t+h*3600]['c'])/float(r['c'])-1)
            assert math.isclose(result,float(row[key]),rel_tol=0,abs_tol=1e-9)
            checked+=1
    if row['mfe_24h_pct']:
        sub=[s[t+i*3600] for i in range(1,25)]
        mfe=100*(max(float(x['h']) for x in sub)/float(r['c'])-1)
        mae=100*(min(float(x['l']) for x in sub)/float(r['c'])-1)
        assert math.isclose(mfe,float(row['mfe_24h_pct']),abs_tol=1e-9)
        assert math.isclose(mae,float(row['mae_24h_pct']),abs_tol=1e-9)
        checked+=2
assert checked>0
for published in report['groups']:
    group=published['group']; horizon=published['horizon_hours']
    members=(controls if group=='flat_nonspike_unmatched' else events if group=='all_events' else
             [r for r in events if r['symbol']==group] if group in report['symbols'] else
             [r for r in events if r['flow_group']==group])
    values=[float(r[f'return_{horizon}h_pct']) for r in members if r[f'return_{horizon}h_pct']]
    assert len(members)==published['total_rows']
    assert len(values)==published['n']
    assert sum(v>0 for v in values)==published['positive_n']
    for key,actual in [('median_return_pct',statistics.median(values) if values else None),
                       ('positive_pct',100*sum(v>0 for v in values)/len(values) if values else None),
                       ('median_mfe_pct',statistics.median(float(r['mfe_24h_pct']) for r in members if r['mfe_24h_pct']) if horizon==24 and values else None),
                       ('median_mae_pct',statistics.median(float(r['mae_24h_pct']) for r in members if r['mae_24h_pct']) if horizon==24 and values else None)]:
        expected=published[key]
        assert (actual is None and expected is None) or (actual is not None and expected is not None and math.isclose(actual,expected,abs_tol=1e-9)),(group,horizon,key,actual,expected)
print(f'PASS: {len(events)} retained events, {checked} outcome/path values and {len(report["groups"])} summary cells checked from frozen inputs; independent source provenance and causal claims not proven')