"""QE-RESEARCH-03/1.0: recompute summaries of anonymous paired groups.

Standard library only. No network, trading or model training. Rows contain
within-group medians, not individual returns. Comparisons are descriptive.
"""
import argparse,csv,json,math,statistics,sys
from pathlib import Path

FIELDS=['variant','group','control_pnl','control_dd_pct','control_models','variant_pnl','variant_dd_pct','variant_models']
class InputError(ValueError):pass

def read_pairs(path):
    with Path(path).open(encoding='utf-8-sig',newline='') as f:
        reader=csv.DictReader(f)
        if reader.fieldnames!=FIELDS:raise InputError('E_COLUMNS: expected '+','.join(FIELDS))
        rows=list(reader)
    if not rows:raise InputError('E_EMPTY: no paired rows')
    seen=set();controls={}
    for row in rows:
        if set(row)!=set(FIELDS) or any(v is None for v in row.values()):raise InputError('E_ROW: incomplete or extra fields')
        if row['variant'] not in ['A','B'] or not row['group'].strip():raise InputError('E_LABEL: expected variant A/B and a group label')
        key=(row['variant'],row['group'])
        if key in seen:raise InputError('E_DUPLICATE_PAIR: '+':'.join(key))
        seen.add(key)
        for key in FIELDS[2:]:
            try:x=float(row[key])
            except (TypeError,ValueError):raise InputError('E_NUMBER: '+key)
            if not math.isfinite(x):raise InputError('E_NUMBER: '+key)
            if key.endswith('models') and (x<1 or not x.is_integer()):raise InputError('E_COUNT: '+key)
            if key.endswith('dd_pct') and not 0<=x<=100:raise InputError('E_DRAWDOWN: '+key)
            row[key]=int(x) if key.endswith('models') else x
        control=tuple(row[k] for k in FIELDS[2:5])
        if row['group'] in controls and controls[row['group']]!=control:raise InputError('E_CONTROL_MISMATCH: '+row['group'])
        controls[row['group']]=control
    if {r['variant'] for r in rows}!={'A','B'}:raise InputError('E_VARIANTS: both A and B required')
    return rows

def summarize(rows):
    result={}
    for arm in ['A','B']:
        rr=[r for r in rows if r['variant']==arm]
        result[arm]={'pairs':len(rr),'higher_pnl':sum(r['variant_pnl']>r['control_pnl'] for r in rr),'positive_pnl':sum(r['variant_pnl']>0 for r in rr),'positive_control':sum(r['control_pnl']>0 for r in rr),'lower_dd':sum(r['variant_dd_pct']<r['control_dd_pct'] for r in rr),'median_pnl_difference_usd':statistics.median(r['variant_pnl']-r['control_pnl'] for r in rr)}
    arms={a:{r['group']:r for r in rows if r['variant']==a} for a in ['A','B']};shared=sorted(arms['A'].keys()&arms['B'].keys())
    return {'schema':'QE-RESEARCH-03/1.0','status':'CALCULATED','variants':result,'shared_groups':len(shared),'union_groups':len(arms['A'].keys()|arms['B'].keys()),'shared_B_minus_A_median_usd':statistics.median(arms['B'][g]['variant_pnl']-arms['A'][g]['variant_pnl'] for g in shared) if shared else None,'limits':['Group medians, not independent model trials.','Already-observed research period; not a fresh performance test.','Higher profit versus control can still be a loss.']}

def main():
    parser=argparse.ArgumentParser(description=__doc__);parser.add_argument('csv');parser.add_argument('--out',type=Path);args=parser.parse_args()
    try:
        result=summarize(read_pairs(args.csv));text=json.dumps(result,indent=2,allow_nan=False)+'\n'
        if args.out:
            with args.out.open('x',encoding='utf-8') as f:f.write(text)
        else:print(text,end='')
    except (InputError,OSError) as error:
        print(str(error),file=sys.stderr);return 2
    return 0
if __name__=='__main__':raise SystemExit(main())