"""Phase 1 数据同步 CLI（Tushare 首选 → SQLite，新浪校验兜底）。

用法（cd backend）：
    uv run python -m app.cli.sync basic
    uv run python -m app.cli.sync calendar --start 20240101 --end 20241231
    uv run python -m app.cli.sync daily --symbols 600519.SH,000001.SZ --start 20240101
    uv run python -m app.cli.sync daily --all --start 20240101            # 全市场
    uv run python -m app.cli.sync financial --all                        # 财务指标（增量）
    uv run python -m app.cli.sync financial --all --full                 # 财务指标（强制全量重拉）
    uv run python -m app.cli.sync verify --symbol 600519.SH               # 新浪交叉验证

增量与兜底：
- daily --resume：从本地最新交易日续传（已有）；Tushare 失败时走新浪校验兜底，
  只有「两源重叠历史一致」才用新浪补本地缺失交易日（source=sina/前复权）。
- financial：默认增量——本地已含最新应披露报告期则跳过；Tushare 失败时新浪
  数据须通过「两边一致」校验（重叠报告期 eps/销售毛利率逐期一致）才允许补入
  本地缺失键（source=sina）。失败股票留待下轮重跑补齐，不会静默导入未核验数据。
- 每次拉取写入 sync_log 审计（来源 / 成功与否 / 行数 / 区间），禁止静默切源。

本模块是组装层（composition root）：在此装配 Provider / Repository / Session，
业务逻辑在 application.services.data_sync，业务层仍只依赖抽象。
"""

from __future__ import annotations

import argparse
import sys
import time
from datetime import date, datetime, timedelta

from sqlalchemy import select

from app.application.services.data_sync import (
    DailySymbolResult,
    FinancialSymbolResult,
    VerifiedDailySyncer,
    VerifiedFinancialSyncer,
)
from app.core.config import get_settings
from app.infrastructure.data_sources.errors import DataSourceError
from app.infrastructure.data_sources.sina import SinaProvider
from app.infrastructure.data_sources.tushare import TushareProvider
from app.infrastructure.persistence.sqlalchemy.models.market import StockModel
from app.infrastructure.persistence.sqlalchemy.repositories.index_impl import (
    SqlAlchemyIndexConstituentRepository,
)
from app.infrastructure.persistence.sqlalchemy.repositories.market_impl import (
    SqlAlchemyAdjustFactorRepository,
    SqlAlchemyDailyBarRepository,
    SqlAlchemyFinancialRepository,
    SqlAlchemyStockRepository,
    SqlAlchemySyncLogRepository,
    SqlAlchemyTradingCalendarRepository,
)
from app.infrastructure.persistence.sqlalchemy.session import SessionLocal

_DATE_FMT = "%Y%m%d"


def _parse_day(text: str) -> date:
    return datetime.strptime(text, _DATE_FMT).date()


def _failover_provider(session):
    """Tushare 首选 + 新浪兜底（basic/calendar 用；daily/financial 走校验兜底服务）。

    FailoverProvider 每次尝试写 sync_log（AGENT.md §7）。能力矩阵：新浪仅提供
    日线/财务，basic/calendar 新浪不支持 → 抛错保留单源语义，日志可见。
    """
    from app.infrastructure.data_sources.failover import FailoverProvider
    from app.infrastructure.data_sources.sina import SinaProvider

    audit_repo = SqlAlchemySyncLogRepository(session)
    primary = TushareProvider(token=get_settings().tushare_token)
    return FailoverProvider(primary, fallback=SinaProvider(), audit=audit_repo.add)


def _session_ctx():
    return SessionLocal()


def cmd_basic(args) -> int:
    with _session_ctx() as session:
        provider = _failover_provider(session)
        stocks = provider.get_stock_basic()
        repo = SqlAlchemyStockRepository(session)
        touched = repo.upsert_many(stocks)
        session.commit()
    print(f"[basic] 拉取 {len(stocks)} 只，落库 {touched} 条")
    return 0


def cmd_calendar(args) -> int:
    start = _parse_day(args.start)
    end = _parse_day(args.end)
    with _session_ctx() as session:
        provider = _failover_provider(session)
        days = provider.get_trade_cal(start, end)
        repo = SqlAlchemyTradingCalendarRepository(session)
        touched = repo.upsert_many(days)
        session.commit()
    open_days = sum(1 for d in days if d.is_open)
    print(f"[calendar] {start}~{end} 共 {len(days)} 条（交易日 {open_days}），落库 {touched} 条")
    return 0


def _symbols_of(args) -> list[str]:
    if getattr(args, "all", False):
        with _session_ctx() as session:
            symbols = list(session.scalars(select(StockModel.symbol).order_by(StockModel.symbol)))
        if not symbols:
            print("[error] stock 表为空，请先运行：python -m app.cli.sync basic")
            sys.exit(2)
        return symbols
    return [s.strip() for s in args.symbols.split(",") if s.strip()]


def _stock_names(session, symbols: list[str]) -> dict[str, str]:
    """一次性取出股票名称（进度描述用）；批量查询避开 SQLite 变量上限。"""
    names: dict[str, str] = {}
    for i in range(0, len(symbols), 500):
        chunk = symbols[i : i + 500]
        rows = session.execute(
            select(StockModel.symbol, StockModel.name).where(StockModel.symbol.in_(chunk))
        )
        names.update({sym: nm for sym, nm in rows})
    return names


def _warn_notes(notes: list[str]) -> None:
    for note in notes:
        print(f"  [warn] {note}", file=sys.stderr)


def cmd_daily(args) -> int:
    from sqlalchemy import func

    from app.infrastructure.persistence.sqlalchemy.models.market import StockDailyModel

    symbols = _symbols_of(args)
    start = _parse_day(args.start) if args.start else date(2005, 1, 1)
    end = _parse_day(args.end) if args.end else date.today()
    started = time.monotonic()
    n_ok = n_sina = n_failed = n_skip = 0
    rows_tushare = rows_sina = 0
    with _session_ctx() as session:
        names = _stock_names(session, symbols)
        audit = SqlAlchemySyncLogRepository(session).add
        syncer = VerifiedDailySyncer(
            primary=TushareProvider(token=get_settings().tushare_token),
            fallback=SinaProvider(),
            bars=SqlAlchemyDailyBarRepository(session),
            factors=SqlAlchemyAdjustFactorRepository(session),
            audit=audit,
        )
        bar_repo = SqlAlchemyDailyBarRepository(session)
        # 增量基准：本地数据已到该日期即视为「已最新」，resume 时不再调 API
        global_latest = (
            session.scalar(select(func.max(StockDailyModel.trade_date))) if args.resume else None
        )
        for i, symbol in enumerate(symbols, start=1):
            begin = start
            if args.resume:
                latest = bar_repo.latest_date(symbol)
                if latest is not None:
                    if global_latest is not None and latest >= global_latest:
                        n_skip += 1  # 已同步到本地最新交易日，无需续拉
                        continue
                    begin = max(begin, latest + timedelta(days=1))
                if begin > end:
                    n_skip += 1  # 无待拉区间（如区间已含在本地）
                    continue
            if getattr(args, "sleep", 0) > 0:
                time.sleep(args.sleep)
            res: DailySymbolResult = syncer.sync_symbol(symbol, begin, end)
            session.commit()  # 逐只落库：中断/报错只丢当前一只，重跑增量续传
            if res.status == "ok":
                n_ok += 1
                rows_tushare += res.bars_written
            elif res.status == "sina":
                n_sina += 1
                rows_sina += res.bars_written
            elif res.status == "failed":
                n_failed += 1
            _warn_notes(res.notes)
            if i % 100 == 0:
                name = names.get(symbol, "")
                print(
                    f"  ... {i}/{len(symbols)} {symbol} {name}: "
                    f"累计 tushare {rows_tushare} 根 + 新浪补缺 {rows_sina} 根；"
                    f"成功 {n_ok} / 新浪 {n_sina} / 失败待重试 {n_failed}"
                )
    elapsed = time.monotonic() - started
    detail = (
        f"[daily] {len(symbols)} 只股票：成功 {n_ok} / 新浪校验补缺 {n_sina} / "
        f"已最新跳过 {n_skip} / 失败待重试 {n_failed}"
    )
    if args.resume:
        detail += f"（本地最新 {global_latest}）"
    detail += f"；写入 {rows_tushare} 根（tushare 不复权）+ {rows_sina} 根（sina 前复权），耗时 {elapsed:.0f}s"
    print(detail)
    return 0


def _fin_progress_line(i: int, n: int, symbol: str, name: str, res: FinancialSymbolResult) -> str:
    """financial 逐只进度行：结果 + 导入内容简单描述（报告期/公告区间、来源）。"""
    head = f"[financial {i}/{n}] {symbol} {name or ''}".rstrip()
    if res.status == "skip":
        return f"{head}：已最新，跳过（增量）"
    if res.status == "failed":
        return f"{head}：失败待重试（tushare 失败；新浪源 {'未通过校验' if res.source == 'sina' else '不可用'}）"
    if res.status == "sina":
        return (
            f"{head}：tushare 失败 → 新浪校验通过，补入 {res.written} 行（source=sina）"
            + _fin_span(res)
        )
    # status == ok（tushare 成功）
    if res.written:
        updated = f"，覆盖更新 {res.updated} 行" if res.updated else ""
        return f"{head}：tushare 返回 {res.fetched} 行 → 新增 {res.written} 行{updated}" + _fin_span(res)
    return f"{head}：tushare 返回 {res.fetched} 行，均已在库，无新增"


def _fin_span(res: FinancialSymbolResult) -> str:
    if not res.written or res.report_first is None:
        return ""
    if res.announce_first is None or res.announce_last is None:
        return ""
    return (
        f"；报告期 {res.report_first.isoformat()}~{res.report_last.isoformat()}"
        f"（公告 {res.announce_first.isoformat()}~{res.announce_last.isoformat()}）"
    )


def cmd_financial(args) -> int:
    symbols = _symbols_of(args)
    started = time.monotonic()
    n_ok = n_sina = n_failed = n_skip = 0
    rows_tushare = rows_sina = 0
    with _session_ctx() as session:
        names = _stock_names(session, symbols)
        audit = SqlAlchemySyncLogRepository(session).add
        syncer = VerifiedFinancialSyncer(
            primary=TushareProvider(token=get_settings().tushare_token),
            fallback=SinaProvider(),
            repo=SqlAlchemyFinancialRepository(session),
            audit=audit,
        )
        for i, symbol in enumerate(symbols, start=1):
            if getattr(args, "sleep", 0) > 0:
                time.sleep(args.sleep)
            res: FinancialSymbolResult = syncer.sync_symbol(symbol, force_full=args.full)
            session.commit()  # 逐只落库：中断只丢当前一只，重跑增量续传
            print(_fin_progress_line(i, len(symbols), symbol, names.get(symbol, ""), res))
            _warn_notes(res.notes)
            if res.status == "ok":
                n_ok += 1
                rows_tushare += res.written
            elif res.status == "sina":
                n_sina += 1
                rows_sina += res.written
            elif res.status == "failed":
                n_failed += 1
            elif res.status == "skip":
                n_skip += 1
    elapsed = time.monotonic() - started
    mode = "全量重拉(--full)" if args.full else "增量"
    print(
        f"[financial] 共 {len(symbols)} 只（{mode}）：成功 {n_ok} / 新浪校验兜底 {n_sina} / "
        f"已最新跳过 {n_skip} / 失败待重试 {n_failed}；"
        f"合计写入 {rows_tushare + rows_sina} 行（tushare {rows_tushare} + sina {rows_sina}），"
        f"耗时 {elapsed:.0f}s"
    )
    if n_failed:
        print(
            "  [tip] 失败股票未写入未核验数据，重跑本命令即可续传补齐；"
            "若因频率超限，可用 --sleep 加大间隔（如 --sleep 60）分多次跑。",
            file=sys.stderr,
        )
    return 0


def cmd_index_weight(args) -> int:
    """同步指数历史成分（Tushare index_weight；新浪不支持 → failover 审计留痕）。"""

    code = args.code
    with _session_ctx() as session:
        provider = _failover_provider(session)
        try:
            rows = provider.get_index_weight(code)
        except DataSourceError as exc:
            print(f"[index_weight] {code} 失败：{exc}")
            return 1
        repo = SqlAlchemyIndexConstituentRepository(session)
        touched = repo.upsert_many(rows)
        latest = repo.latest_date(code)
        session.commit()
    print(
        f"[index_weight] {code} 拉取 {len(rows)} 期成分行，落库 {touched} 条"
        f"，最新快照 {latest}（as_of 查询见 Universe.index_code）"
    )
    return 0


def cmd_verify(args) -> int:
    """新浪交叉验证：取新浪最新前复权收盘，与本地最新交易日对照。

    注意：新浪为前复权口径，数值不直接等于本地不复权收盘，
    本命令仅用于确认新浪可用性 / 最新交易日，不把新浪数据并入主库。
    """
    from app.infrastructure.persistence.sqlalchemy.repositories.market_impl import (
        SqlAlchemyDailyBarRepository,
    )

    sina = SinaProvider()
    end = date.today()
    start = end - timedelta(days=20)
    try:
        bars = sina.get_daily(args.symbol, start, end)
    except DataSourceError as exc:
        print(f"[verify] 新浪不可用: {exc}", file=sys.stderr)
        return 1
    if not bars:
        print(f"[verify] 新浪最近无数据（{args.symbol}）")
        return 1
    latest = max(bars, key=lambda b: b.trade_date)
    with _session_ctx() as session:
        local = SqlAlchemyDailyBarRepository(session).latest_date(args.symbol)
    print(
        f"[verify] {args.symbol}: 新浪最新 {latest.trade_date} 收盘(前复权) {latest.close}；"
        f"本地最新交易日 {local}"
    )
    return 0


def cmd_export(args) -> int:
    """把 SQLite 日线按年导出为 Parquet（data/parquet/stock_daily/<year>.parquet）。"""
    from pathlib import Path

    import pandas as pd

    from app.infrastructure.persistence.sqlalchemy.models.market import StockDailyModel

    settings = get_settings()
    out_root = settings.storage.get("parquet_dir") or Path("data/parquet")
    out_root.mkdir(parents=True, exist_ok=True)
    total = 0
    years = [int(y) for y in (args.years or "").split(",") if y.strip()] or None
    with _session_ctx() as session:
        all_bars = session.execute(
            select(StockDailyModel).order_by(StockDailyModel.trade_date)
        ).scalars()
        frame = pd.DataFrame(
            [
                {
                    "symbol": b.symbol,
                    "trade_date": b.trade_date,
                    "open": float(b.open) if b.open is not None else None,
                    "high": float(b.high) if b.high is not None else None,
                    "low": float(b.low) if b.low is not None else None,
                    "close": float(b.close) if b.close is not None else None,
                    "volume": float(b.volume) if b.volume is not None else None,
                    "amount": float(b.amount) if b.amount is not None else None,
                }
                for b in all_bars
            ]
        )
    if frame.empty:
        print("[export] 无日线数据，请先运行 sync daily")
        return 0
    frame["trade_date"] = pd.to_datetime(frame["trade_date"])
    out_dir = out_root / "stock_daily"
    out_dir.mkdir(parents=True, exist_ok=True)
    for year, group in frame.groupby(frame["trade_date"].dt.year):
        if years and int(year) not in years:
            continue
        path = out_dir / f"{year}.parquet"
        group.sort_values(["symbol", "trade_date"]).to_parquet(path, index=False)
        total += len(group)
        print(f"[export] {year} → {path}（{len(group)} 行）")
    print(f"[export] 合计 {total} 行 → {out_dir}")
    return 0


def build_parser() -> argparse.ArgumentParser:
    parser = argparse.ArgumentParser(prog="app.cli.sync", description="Tushare 数据同步 CLI")
    sub = parser.add_subparsers(dest="command", required=True)

    p_basic = sub.add_parser("basic", help="同步股票基础信息")
    p_basic.set_defaults(func=cmd_basic)

    p_cal = sub.add_parser("calendar", help="同步交易日历")
    p_cal.add_argument("--start", required=True, help="YYYYMMDD")
    p_cal.add_argument("--end", required=True, help="YYYYMMDD")
    p_cal.set_defaults(func=cmd_calendar)

    p_daily = sub.add_parser("daily", help="同步日线与复权因子（Tushare 失败 → 新浪校验兜底补缺）")
    p_daily.add_argument("--symbols", default="", help="600519.SH,000001.SZ")
    p_daily.add_argument("--all", action="store_true", help="遍历 stock 表全部股票")
    p_daily.add_argument("--start", default="", help="YYYYMMDD（默认 20050101）")
    p_daily.add_argument("--end", default="", help="YYYYMMDD（默认今天）")
    p_daily.add_argument("--resume", action="store_true", help="从本地最新交易日续传（增量）")
    p_daily.add_argument(
        "--sleep",
        type=float,
        default=0,
        help="每只股票请求间隔秒数（限速时加大，如 1 或 60）",
    )
    p_daily.set_defaults(func=cmd_daily)

    p_fin = sub.add_parser("financial", help="同步财务指标快照（默认增量；Tushare 失败 → 新浪校验兜底）")
    p_fin.add_argument("--symbols", default="")
    p_fin.add_argument("--all", action="store_true", help="遍历 stock 表全部股票")
    p_fin.add_argument(
        "--full",
        action="store_true",
        help="强制全量重拉并覆盖既有行（默认只补本地缺失/更新的报告期，已最新跳过）",
    )
    p_fin.add_argument(
        "--sleep",
        type=float,
        default=0,
        help="每只股票请求间隔秒数（限速时加大，如 1 或 60）",
    )
    p_fin.set_defaults(func=cmd_financial)

    p_idx = sub.add_parser("index_weight", help="同步指数历史成分（如沪深300 000300.SH）")
    p_idx.add_argument("--code", required=True, help="指数代码，如 000300.SH / 000905.SH")
    p_idx.set_defaults(func=cmd_index_weight)

    p_verify = sub.add_parser("verify", help="新浪交叉验证最新行情")
    p_verify.add_argument("--symbol", required=True)
    p_verify.set_defaults(func=cmd_verify)

    p_export = sub.add_parser("export", help="日线按年导出 Parquet（data/parquet）")
    p_export.add_argument("--years", default="", help="逗号分隔年份，留空导出全部")
    p_export.set_defaults(func=cmd_export)
    return parser


def main(argv: list[str] | None = None) -> int:
    args = build_parser().parse_args(argv)
    try:
        return args.func(args)
    except DataSourceError as exc:
        print(f"[error] {exc}", file=sys.stderr)
        return 1
    except KeyboardInterrupt:
        print("\n[interrupt] 已中止", file=sys.stderr)
        return 130


if __name__ == "__main__":
    raise SystemExit(main())