"""C05构造数据审计；仅标准库，默认同目录输出UTF-8、LF、确定性纯JSON。

时间均为显式构造截点；价格非真实行情。小型穷举解释算法，不是生产管道。
时区案例固定官方文档2020-11-01洛杉矶回拨，显式偏移量，无tzdata依赖。
"""
from collections import Counter, defaultdict
from datetime import datetime, timedelta, timezone
from hashlib import sha256
from pathlib import Path
import json
import math


def record(inputs, **tables):
    return {'inputs': inputs, 'tables': tables}


def ohlc():
    rows = [
        ['P01',10,12,9,11,100], ['P02',13,12,9,11,90],
        ['P03',10,12,11,11,80], ['P04',10,12,9,13,70],
        ['P05',10,12,11,10,60], ['P06',10,9,12,11,50],
        ['P07',None,12,9,11,40], ['P08',10,'Infinity',9,11,30],
        ['P09',10,12,9,11,-5], ['P10',10,10,10,10,0],
        ['P11',10,12,9,11,20], ['P12',-1,2,-2,1,10]]
    matrix=[]
    for key,o,h,l,c,v in rows:
        p=[o,h,l,c]; finite=lambda x: isinstance(x,(int,float)) and math.isfinite(x)
        ready=all(finite(x) for x in p)
        bits=[any(x is None for x in p+[v]), any(x is not None and not finite(x) for x in p+[v]),
              ready and any(x<=0 for x in p), finite(v) and v<0,
              ready and h<l, ready and h<o, ready and h<c, ready and l>o, ready and l>c]
        state='隔离' if any(bits) else ('待核零量' if v==0 else '放行')
        matrix.append({'行':key,'O':o,'H':h,'L':l,'C':c,'V':v,'失败位图':''.join(str(int(b)) for b in bits),'状态':state})
    ledger=[{'口径':s,'行数':sum(r['状态']==s for r in matrix),'成交量':sum(r['V'] for r in matrix if r['状态']==s and isinstance(r['V'],int))} for s in ['放行','待核零量','隔离']]
    contrast=[{'条件':'严格放行','可用行':sum(r['状态']=='放行' for r in matrix)},
              {'条件':'取得零量来源确认后仅用于价格观察','可用行':sum(r['状态']!='隔离' for r in matrix)},
              {'条件':'错误仅查空值','可用行':sum(None not in r for r in rows)}]
    return record({'rows':rows,'columns':['行','O','H','L','C','V'],'bit_order':['缺失','非有限','非正价','负量','H<L','H<O','H<C','L>O','L>C']},matrix=matrix,ledger=ledger,contrast=contrast)


def cumulative():
    rows=[['M','09:30',100],['M','09:31',160],['M','11:30',250],['A','13:00',40],['A','13:01',90],['A','13:02',85],['A','15:00',180]]
    out=[]; prev={}
    for s,t,v in rows:
        first=s not in prev; d=v if first else v-prev[s]
        out.append({'会话':s,'时刻':t,'累计股':v,'重置':first,'原始差分':d,'状态':'负差待核' if d<0 else '暂可用'})
        prev[s]=v
    cases=[]
    for name,rv in [('基线待核',[r[2] for r in rows]),('源方确认13:02应为125',[100,160,250,40,90,125,180]),('源方确认下午沿用全日累计',[100,160,250,290,340,375,430])]:
        ds=[]; last={}
        for (s,t,_),v in zip(rows,rv):
            group='day' if name.endswith('全日累计') else s
            ds.append(v-last.get(group,0));last[group]=v
        cases.append({'条件':name,'直接和':sum(rv),'差分净和':sum(ds),'正差相加':sum(max(0,d) for d in ds),'负差数':sum(d<0 for d in ds),'可确认日量':sum(ds) if all(d>=0 for d in ds) else None})
    sessions=[{'会话':s,'末累计':prev[s],'差分净和':sum(r['原始差分'] for r in out if r['会话']==s)} for s in prev]
    return record({'rows':rows,'unit':'股','session_reset':True,'confirmed_correction':[100,160,250,40,90,125,180],'confirmed_day_cumulative':[100,160,250,290,340,375,430]},increments=out,sessions=sessions,contrast=cases)


def bars():
    # 秒数从09:30:00算起，覆盖区间完整保留两端外侧桶。
    rows=[['T1',0,10,1],['T2',299,11,2],['T3',300,12,3],['T4',301,9,4],['T5',600,13,5]]
    clock=lambda sec:(datetime(2026,10,20,9,30)+timedelta(seconds=sec)).strftime('%H:%M:%S')
    allocations=[]; summary=[]; output=[]
    for closed in ['left','right']:
        for label in ['left','right']:
            groups=defaultdict(list)
            for key,sec,p,v in rows:
                k=sec//300 if closed=='left' else (sec-1)//300
                groups[k].append((key,sec,p,v)); a=k*300;b=a+300
                allocations.append({'closed':closed,'label':label,'事件':key,'事件时刻':clock(sec),'桶左':clock(a),'桶右':clock(b),'显示标签':clock(a if label=='left' else b)})
            for k,items in sorted(groups.items()):
                p=[r[2] for r in items];a=k*300;b=a+300
                output.append({'closed':closed,'label':label,'标签':clock(a if label=='left' else b),'覆盖':('[' if closed=='left' else '(')+clock(a)+','+clock(b)+(')' if closed=='left' else ']'),'O':p[0],'H':max(p),'L':min(p),'C':p[-1],'V':sum(r[3] for r in items),'最早完整时刻':clock(b+2)})
            summary.append({'closed':closed,'label':label,'事件数':sum(len(v) for v in groups.values()),'总量':sum(r[3] for r in rows),'桶数':len(groups)})
    return record({'rows':rows,'columns':['事件','距09:30秒','价格元','增量股'],'width_seconds':300,'delivery_delay_seconds':2},allocation=allocations,bars=output,conservation=summary)


def asof():
    quotes=[['Q1','X',0,99,101],['Q2','X',1000,101,103],['Q3','X',2100,103,105],['Q4','Y',1400,49,51]]
    orders=[['O1','X',900,'买',102],['O2','X',2000,'卖',103],['O3','X',4000,'买',105],['O4','Y',1500,'卖',50],['O5','Z',1500,'买',20]]
    candidates=[];chosen=[]
    for oid,s,t,side,fill in orders:
        qs=[q for q in quotes if q[1]==s]
        for qid,_,qt,bid,ask in qs:
            candidates.append({'订单':oid,'报价':qid,'订单减报价毫秒':t-qt,'过去或相等':qt<=t,'绝对时差':abs(t-qt)})
        for direction in ['backward','nearest']:
            for tol in [1200,2000]:
                pool=[q for q in qs if direction=='nearest' or q[2]<=t]
                q=min(pool,key=lambda q:(abs(t-q[2]),q[2])) if pool else None
                accepted=q is not None and abs(t-q[2])<=tol
                mid=(q[3]+q[4])/2 if accepted else None
                chosen.append({'订单':oid,'方向':direction,'容差毫秒':tol,'候选':q[0] if q else None,'时差毫秒':t-q[2] if q else None,'放行':accepted,'拒因':None if accepted else ('无同证券候选' if q is None else '过期'),'中价':mid,'成交减中价':fill-mid if accepted else None,'方向成本元':((fill-mid) if side=='买' else (mid-fill)) if accepted else None})
    return record({'quotes':quotes,'quote_columns':['报价','证券','毫秒','bid','ask'],'orders':orders,'order_columns':['订单','证券','毫秒','方向','成交价']},candidates=candidates,selection=chosen)


def identities():
    aliases=[['A1','AAA','S1',1,4],['A2','BBB','S1',4,8],['A3','AAA','S2',4,8],['A4','CCC','S3',1,3],['A5','CCC','S3',4,8],['A6','BBB','S2',6,7]]
    quotes=[['E1','AAA',2,10],['E2','BBB',4,11],['E3','AAA',5,20],['E4','CCC',3,30],['E5','BBB',6,12],['E6','CCC',5,31],['E7','BBB',7,13]]
    joined=[]
    for eid,code,day,p in quotes:
        hits=[a for a in aliases if a[1]==code and a[3]<=day<a[4]]
        joined.append({'事件':eid,'别名':code,'日':day,'价':p,'候选映射':','.join(a[0] for a in hits) or '无','证券ID':hits[0][2] if len(hits)==1 else None,'状态':'放行' if len(hits)==1 else ('空档' if not hits else '重叠冲突')})
    checks=[]
    for code in sorted({a[1] for a in aliases}):
        for day in range(1,8):
            hits=[a for a in aliases if a[1]==code and a[3]<=day<a[4]]
            if len(hits)!=1:checks.append({'别名':code,'日':day,'匹配数':len(hits),'状态':'空档' if not hits else '重叠冲突'})
    continuity=[]
    for sid in ['S1','S2','S3']:
        rs=[r for r in joined if r['证券ID']==sid]
        continuity.append({'证券ID':sid,'事件':','.join(r['事件'] for r in rs),'行数':len(rs),'别名种数':len({r['别名'] for r in rs})})
    contrast=[]
    for eid,code,day,p in quotes:
        first=next((a for a in aliases if a[1]==code),None)
        contrast.append({'事件':eid,'忽略有效期首条证券ID':first[2] if first else None,'正确证券ID':next(r['证券ID'] for r in joined if r['事件']==eid)})
    return record({'aliases':aliases,'alias_columns':['映射','别名','证券ID','含起日','不含止日'],'quotes':quotes,'quote_columns':['事件','别名','日','价'],'calendar':'构造2026-10-01至07'},joined=joined,interval_audit=checks,continuity=continuity,contrast=contrast)


def dst():
    rows=[['Z1','2020-11-01T00:30:00',-7,0],['Z2','2020-11-01T01:30:00',-7,0],['Z3','2020-11-01T01:30:00',-8,1],['Z4','2020-11-01T02:30:00',-8,0]]
    table=[]
    for key,wall,offset,fold in rows:
        local=datetime.fromisoformat(wall).replace(tzinfo=timezone(timedelta(hours=offset)),fold=fold)
        utc=local.astimezone(timezone.utc)
        wrong=datetime.fromisoformat(wall).replace(tzinfo=timezone(timedelta(hours=-8))).astimezone(timezone.utc)
        table.append({'事件':key,'本地':wall,'offset小时':offset,'fold':fold,'UTC':utc.isoformat(),'错误固定负8UTC':wrong.isoformat(),'错误减正确秒':int((wrong-utc).total_seconds())})
    checks=[{'映射':'完整offset','唯一时刻数':len({r['UTC'] for r in table}),'严格递增':all(table[i]['UTC']<table[i+1]['UTC'] for i in range(len(table)-1))},
            {'映射':'错误固定负8','唯一时刻数':len({r['错误固定负8UTC'] for r in table}),'严格递增':all(table[i]['错误固定负8UTC']<table[i+1]['错误固定负8UTC'] for i in range(len(table)-1))},
            {'映射':'仅墙上时间','唯一时刻数':len({r['本地'] for r in table}),'严格递增':all(table[i]['本地']<table[i+1]['本地'] for i in range(len(table)-1))}]
    return record({'rows':rows,'zone_context':'America/Los_Angeles，仅2020-11-01四个指定时刻','offsets_are_explicit':True},mapping=table,assertions=checks)


def missing():
    # 每名候选在形成截点已入池；未来标签只作评价，不参与选池。
    values=[4,3,-2,1,0,-1,5,-3,2,1,2,3,None,None,None,None,None,None,None,None]
    reasons=['已知']*12+['窗口未满']*3+['停牌未取到退出价']*2+['源缺失']*2+['退市终局待核']
    rows=[{'证券':f'S{i+1:02}','入池':True,'形成时可知特征':20-i,'标签收益百分数':v,'标签状态':'可得' if v is not None else '缺失','原因':reasons[i]} for i,v in enumerate(values)]
    known=[v for v in values if v is not None];wins=sum(v>0 for v in known);n=len(values);m=n-len(known)
    summary=[{'口径':'完整案例','分子胜数':wins,'分母':len(known),'胜率百分数':wins/len(known)*100},
             {'口径':'缺失全部不胜的下界','分子胜数':wins,'分母':n,'胜率百分数':wins/n*100},
             {'口径':'缺失全部胜的上界','分子胜数':wins+m,'分母':n,'胜率百分数':(wins+m)/n*100}]
    sensitivity=[{'缺失中胜数':k,'总胜数':wins+k,'总分母':n,'胜率百分数':(wins+k)/n*100} for k in range(m+1)]
    reason_table=[{'缺失原因':r,'人数':reasons.count(r)} for r in dict.fromkeys(reasons) if r!='已知']
    return record({'values':values,'reasons':reasons,'win_rule':'收益严格大于0','formation_time':'构造2026-10-01 09:00Z','evaluation_time':'构造2026-10-15 09:00Z'},ledger=rows,summary=summary,sensitivity=sensitivity,reasons=reason_table)


def pagination():
    # 冻结教学服务：时间索引0..9，整点逐分事件，每窗含双端，升序，静默上限4。
    events=[{'键':f'K{i:02}','分钟':i} for i in range(10)];cap=4
    request_windows=[['整窗',0,9,0],['分页1',0,9,0],['分页2',0,9,4],['断页3',0,9,8],['完整分页3',0,9,8],['分窗A',0,3,0],['分窗B',3,6,0],['分窗C',6,9,0]]
    logs=[];sets={}
    for name,a,b,cursor in request_windows:
        eligible=[r for r in events if a<=r['分钟']<=b]
        got=[] if name=='断页3' else eligible[cursor:cursor+cap]
        sets[name]={r['键'] for r in got}
        logs.append({'请求':name,'窗口含双端':f'{a}..{b}','游标':cursor,'返回数':len(got),'首分钟':got[0]['分钟'] if got else None,'末分钟':got[-1]['分钟'] if got else None,'next游标':cursor+cap if cursor+cap<len(eligible) else None,'返回键':','.join(r['键'] for r in got) or '空'})
    paths=[];expected={r['键'] for r in events}
    for name,parts in [('整窗',['整窗']),('错误遇空停止',['分页1','分页2','断页3']),('完整游标链',['分页1','分页2','完整分页3']),('闭区间重叠分窗',['分窗A','分窗B','分窗C'])]:
        union=set().union(*(sets[p] for p in parts));count=sum(len(sets[p]) for p in parts)
        observed=[r['分钟'] for r in events if r['键'] in union]
        runs=[]
        for minute in observed:
            if runs and minute==runs[-1][1]+1:runs[-1][1]=minute
            else:runs.append([minute,minute])
        paths.append({'路径':name,'原始条数':count,'唯一条数':len(union),'重叠条数':count-len(union),'缺失键':','.join(sorted(expected-union)) or '无','观测分钟覆盖':','.join(map(str,observed)),'离散覆盖区间并集':','.join(f'[{a},{b}]' for a,b in runs),'完整':union==expected})
    overlap=[{'相邻窗':a+' / '+b,'重叠键':','.join(sorted(sets[a]&sets[b]))} for a,b in [('分窗A','分窗B'),('分窗B','分窗C')]]
    return record({'events':events,'cap':cap,'inclusive_endpoints':True,'snapshot':'教学冻结版v1','request_windows':request_windows,'fault':'游标8返回空但仍缺K08/K09'},requests=logs,coverage=paths,overlap=overlap)


def dualtime():
    # 分钟从2026-10-20 09:00Z起；同一观察期，各源版本独立。
    rows=[['A',1,100,0,4],['B',1,200,1,2],['A',2,110,5,9],['B',2,190,7,8]]
    eligibility=[];snapshots=[]
    for d in [3,6,8,10]:
        for source,version,value,release,ingest in rows:
            eligibility.append({'决策分钟':d,'源':source,'版本':version,'值':value,'release':release,'ingest':ingest,'公开可用':release<=d,'系统可用':release<=d and ingest<=d})
        for policy in ['公开边界','实际系统','入库额外延迟2分']:
            chosen={}
            for s in ['A','B']:
                pool=[r for r in rows if r[0]==s and r[3]<=d and (policy=='公开边界' or r[4]+(2 if policy=='入库额外延迟2分' else 0)<=d)]
                chosen[s]=max(pool,key=lambda r:(r[3],r[1])) if pool else None
            snapshots.append({'决策分钟':d,'口径':policy,'A版本':chosen['A'][1] if chosen['A'] else None,'A值':chosen['A'][2] if chosen['A'] else None,'B版本':chosen['B'][1] if chosen['B'] else None,'B值':chosen['B'][2] if chosen['B'] else None,'两源和':sum(r[2] for r in chosen.values()) if all(chosen.values()) else None})
    return record({'rows':rows,'columns':['源','版本','值','release分钟','ingest分钟'],'observation':'构造2026-09期','origin':'2026-10-20T09:00:00+00:00','decisions':[3,6,8,10]},eligibility=eligibility,snapshots=snapshots)


def canonical(rows):
    return ''.join(json.dumps(r,ensure_ascii=False,separators=(',',':'),sort_keys=True)+'\n' for r in rows).encode('utf-8')


def revision():
    a=[{'day':d,'value':v} for d,v in [(1,10),(2,12),(3,14),(4,16),(5,18)]]
    b=[{'day':d,'value':v} for d,v in [(1,10),(2,15),(4,16),(5,18),(6,20)]]
    hashes=[];changes=[]
    for tag,rows in [('A',a),('B',b)]:
        for r in rows:hashes.append({'快照':tag,'日':r['day'],'行SHA256':sha256(canonical([r])).hexdigest()})
    ad={r['day']:r['value'] for r in a};bd={r['day']:r['value'] for r in b}
    for d in sorted(set(ad)|set(bd)):
        if ad.get(d)!=bd.get(d):changes.append({'日':d,'旧':ad.get(d),'新':bd.get(d),'类型':'新增' if d not in ad else ('删除' if d not in bd else '修订'),'依赖影响日':f'{d}..{d+1}'})
    feature=[]
    for d in range(1,7):
        av=(ad[d-1]+ad[d])/2 if d-1 in ad and d in ad else None
        bv=(bd[d-1]+bd[d])/2 if d-1 in bd and d in bd else None
        feature.append({'日':d,'A两日均值':av,'B两日均值':bv,'A指标贡献':av if av is not None else 0,'B指标贡献':bv if bv is not None else 0,'桥接差额':(bv if bv is not None else 0)-(av if av is not None else 0)})
    manifest=[{'快照':tag,'行数':len(rows),'字节数':len(canonical(rows)),'文件SHA256':sha256(canonical(rows)).hexdigest(),'指标合计':sum(r[f'{tag}指标贡献'] for r in feature)} for tag,rows in [('A',a),('B',b)]]
    contrast=[{'口径':'固定日历完整两点','A合计':sum(r['A指标贡献'] for r in feature),'B合计':sum(r['B指标贡献'] for r in feature)},
              {'口径':'错误删除后按相邻行滚动','A合计':sum((a[i-1]['value']+a[i]['value'])/2 for i in range(1,len(a))),'B合计':sum((b[i-1]['value']+b[i]['value'])/2 for i in range(1,len(b)))}]
    return record({'A':a,'B':b,'calendar_days':list(range(1,7)),'dependency':'F(d)=(x(d-1)+x(d))/2；缺任一点F为空；指标仅加可得F，不把空F称为0均值','serialization':'每行sort_keys紧凑JSON+LF，UTF-8无BOM'},row_hashes=hashes,diff=changes,features=feature,manifest=manifest,contrast=contrast)


def joins():
    left=[['L1','A',1],['L2','A',2],['L3','B',3],['L4','C',4],['L5',None,5]]
    right=[['R1','A','科技'],['R2','A','金融'],['R3','B','工业'],['R4',None,'未知占位']]
    lc=Counter(r[1] for r in left);rc=Counter(r[1] for r in right)
    counts=[{'键':k,'左计数':lc[k],'右计数':rc[k],'内连接积':lc[k]*rc[k]} for k in ['A','B','C',None]]
    expanded=[]
    for lid,k,v in left:
        hits=[r for r in right if r[1]==k]
        for r in hits or [[None,None,None]]:
            expanded.append({'左行':lid,'键':k,'特征':v,'右行':r[0],'行业':r[2],'指示':'both' if r[0] else 'left_only'})
    clean_right=[r for r in right if r[0] in ['R1','R3']]
    repaired=[]
    for lid,k,v in left:
        hits=[r for r in clean_right if k is not None and r[1]==k]
        repaired.append({'左行':lid,'键':k,'行业':hits[0][2] if hits else None,'指示':'both' if hits else 'left_only','状态':'空键隔离' if k is None else ('放行' if hits else '未匹配')})
    validations=[]
    for tag,l,r in [('原表',left,right),('已确认R1为本次分类截点，移除空右键',left,clean_right)]:
        ul=len({x[1] for x in l})==len(l);ur=len({x[1] for x in r})==len(r)
        validations.append({'输入':tag,'one_to_many':ul,'many_to_one':ur,'one_to_one':ul and ur})
    ledgers=[{'情景':'错误空键可匹配左连接','左输入':len(left),'输出':len(expanded),'both':sum(r['指示']=='both' for r in expanded),'left_only':sum(r['指示']=='left_only' for r in expanded),'特征和':sum(r['特征'] for r in expanded)},
             {'情景':'右键证据解决且空键拒配','左输入':len(left),'输出':len(repaired),'both':sum(r['指示']=='both' for r in repaired),'left_only':sum(r['指示']=='left_only' for r in repaired),'特征和':sum(r[2] for r in left)}]
    return record({'left':left,'left_columns':['左行','键','特征'],'right':right,'right_columns':['右行','键','行业'],'repair_evidence':'构造源方确认本次分类截点应选R1，R2为错误混入的另一版本'},key_counts=counts,expanded=expanded,validation=validations,repaired=repaired,ledger=ledgers)


def build():
    funcs=[ohlc,cumulative,bars,asof,identities,dst,missing,pagination,dualtime,revision,joins]
    return {f'DJ4-{45+i:03}':fn() for i,fn in enumerate(funcs)}


if __name__=='__main__':
    result=build()
    assert len(result)==11
    destination=Path(__file__).with_suffix('.json')
    with destination.open('w',encoding='utf-8',newline='\n') as stream:
        stream.write(json.dumps(result,ensure_ascii=False,indent=2,allow_nan=False,sort_keys=True)+'\n')
    print('已生成11篇构造输入与核算表：'+destination.name)