import pandas as pd,numpy as np,glob,os
pd.set_option('display.width',250);pd.set_option('display.max_columns',40);pd.set_option('display.max_rows',400)
D='../data/1m'
def L(n):
    d=pd.read_csv(f'{D}/{n}_1m.csv',parse_dates=['ts']); d['ts']=d.ts.dt.tz_localize(None); return d.set_index('ts')
def P(f): return pd.read_parquet(f).set_index('ts')
out=open('results/intraday_report.txt','w')
def pr(*a):
    s=' '.join(str(x) for x in a); print(s); out.write(s+'\n')
LOT={'NIFTY':65,'BANKNIFTY':30}
for day,phases in [('2026-10-23',[('13:20','13:37'),('13:37','13:41'),('13:41','13:47'),('13:47','13:54'),('13:54','14:10')]),
                   ('2026-10-13',[('09:15','09:30'),('09:30','10:00'),('13:45','14:15'),('14:15','14:30'),('14:30','14:45'),('14:45','15:00'),('15:00','15:14')])]:
    pr('\n=================',day)
    spot=L('NIFTY').close[day]; bns=L('BANKNIFTY').close[day]
    fut={}
    for u in ['NIFTY','BANKNIFTY']:
        fs=sorted(glob.glob(f'data/{u}_FUT_*_{day}.parquet'))
        fut[u]={os.path.basename(f).split('_FUT_')[1][:9]:P(f) for f in fs}
    # minute OI sanity: |dOI| <= minute volume
    for u in fut:
        for k,v in fut[u].items():
            bad=(v.oi.diff().abs()>v.volume).sum(); pr(f'{u} {k}: minutes where |dOI|>volume (OI feed glitch/lag) = {bad} of {len(v)}')
    rows=[]
    for a,b in phases:
        ta,tb=pd.Timestamp(f'{day} {a}'),pd.Timestamp(f'{day} {b}')
        r=dict(phase=f'{a}->{b}',nifty_spot=round(spot.get(tb,np.nan)-spot.get(ta,np.nan),1),bn_spot=round(bns.get(tb,np.nan)-bns.get(ta,np.nan),1))
        for u in fut:
            near=list(fut[u].values())[0] if False else fut[u]
            oct_=[v for k,v in fut[u].items() if 'OCT' in k][0]
            r[f'{u[:2]}fut_pts']=round(oct_.close[tb]-oct_.close[ta],1)
            vol=sum(v.volume[(v.index>ta)&(v.index<=tb)].sum() for v in fut[u].values())
            doi=sum(v.oi[tb]-v.oi[ta] for v in fut[u].values())
            r[f'{u[:2]}_vol_ctr']=int(vol/LOT[u]); r[f'{u[:2]}_dOI_ctr_allexp']=int(doi/LOT[u])
            r[f'{u[:2]}_dOI_ctr_oct']=int((oct_.oi[tb]-oct_.oi[ta])/LOT[u])
        rows.append(r)
    t=pd.DataFrame(rows); pr(t.to_string(index=False)); t.to_csv(f'results/fut_phases_{day}.csv',index=False)
    # minute table NIFTY all expiries combined around burst
    o=fut['NIFTY']; comb=sum(v.oi for v in o.values()); vv=sum(v.volume for v in o.values())
    oc=[v for k,v in o.items() if 'OCT' in k][0]
    m=pd.DataFrame({'spot':spot,'fut':oc.close,'basis':oc.close-spot,'vol_ctr':(vv/65).round(),'dOI_ctr':(comb.diff()/65).round(),'bn_spot':bns,'bnfut_vol_ctr':(sum(v.volume for v in fut['BANKNIFTY'].values())/30).round(),'bn_dOI_ctr':(sum(v.oi for v in fut['BANKNIFTY'].values()).diff()/30).round()})
    w=('13:30','13:58') if day=='2026-10-23' else ('14:15','14:50')
    mm=m.between_time(*w); pr(mm.to_string()); mm.to_csv(f'results/fut_minutes_{day}.csv')
out.close()