#!/usr/bin/env python3
"""
Phase C dry-run 종료 후 모의 수익률(참고) 집계.

- 입력: data/ops/paper_fires.jsonl (06 dry-run 발화 로그)
- 출력: docs/05_ops/phase_c_paper_report.json + 콘솔 요약

주의: 실계좌 수익이 아님. 발화가 N봉 후 가격으로 계산한 forward % 합산(참고).
      hybrid 복리 PnL은 simulation_report.html 과 다릅니다.
"""
from __future__ import annotations

import json
import runpy
from datetime import datetime
from pathlib import Path

import numpy as np
import pandas as pd

runpy.run_path(str(Path(__file__).resolve().parent / "_bootstrap.py"))

from config import (  # noqa: E402
    MATCH_FORWARD_BARS,
    MATCH_PRIMARY_INTERVAL,
    SYMBOL,
    TRADING_FEE_RATE,
)
from deepcoin.matching.label_outcomes import _forward_ret_vectorized  # noqa: E402
from deepcoin.ops.monitor import Monitor  # noqa: E402
from deepcoin.paths import PAPER_FIRES_LOG, PAPER_WEEKLY_REPORT_JSON  # noqa: E402

_FEE_PCT = TRADING_FEE_RATE * 2 * 100


def load_paper_fires(path: Path) -> pd.DataFrame:
    """paper_fires.jsonl → DataFrame."""
    if not path.is_file():
        return pd.DataFrame()
    rows = []
    for line in path.read_text(encoding="utf-8").splitlines():
        line = line.strip()
        if line:
            rows.append(json.loads(line))
    if not rows:
        return pd.DataFrame()
    return pd.DataFrame(rows)


def attach_forward_returns(fires: pd.DataFrame, close_df: pd.DataFrame) -> pd.DataFrame:
    """
    would_trade=True 발화에 MATCH_FORWARD_BARS 기준 forward 수익률(%) 부여.

    Args:
        fires: paper 발화.
        close_df: 3분 종가 (datetime index).

    Returns:
        forward_ret_pct 컬럼 추가.
    """
    if fires.empty or close_df.empty:
        fires["forward_ret_pct"] = np.nan
        return fires

    close_df = close_df.sort_index()
    if not isinstance(close_df.index, pd.DatetimeIndex):
        close_df.index = pd.to_datetime(close_df.index)
    close_ts_ns = close_df.index.astype(np.int64).values
    close_px = close_df["Close"].astype(float).values

    sub = fires[fires["would_trade"] == True].copy()  # noqa: E712
    if sub.empty:
        fires["forward_ret_pct"] = np.nan
        return fires

    sig = pd.to_datetime(sub["signal_dt"])
    fire_ns = sig.astype(np.int64).values
    c0 = sub["close"].astype(float).values
    side = sub["side"].astype(str).values
    ret, valid = _forward_ret_vectorized(
        fire_ns, c0, close_ts_ns, close_px, side, MATCH_FORWARD_BARS, _FEE_PCT
    )
    fires = fires.copy()
    fires["forward_ret_pct"] = np.nan
    idx = sub.index
    fires.loc[idx, "forward_ret_pct"] = np.where(valid, ret, np.nan)
    return fires


def summarize(fires: pd.DataFrame) -> dict:
    """집계 dict."""
    traded = fires[fires["would_trade"] == True]  # noqa: E712
    with_ret = traded[traded["forward_ret_pct"].notna()]
    out: dict = {
        "generated_at": datetime.now().isoformat(timespec="seconds"),
        "symbol": SYMBOL,
        "forward_bars": MATCH_FORWARD_BARS,
        "fee_round_trip_pct": _FEE_PCT,
        "total_signals": int(len(fires)),
        "would_trade_count": int(len(traded)),
        "skipped_count": int(len(fires) - len(traded)),
        "labeled_count": int(len(with_ret)),
        "note": (
            "모의 forward 수익률. 실계좌·hybrid 복리 PnL 아님. "
            "매수·매도 leg 미결합 단순 합산."
        ),
    }
    if not with_ret.empty:
        out["mean_forward_ret_pct"] = round(float(with_ret["forward_ret_pct"].mean()), 4)
        out["sum_forward_ret_pct"] = round(float(with_ret["forward_ret_pct"].sum()), 4)
        by_side = (
            with_ret.groupby("side")["forward_ret_pct"]
            .agg(["count", "mean", "sum"])
            .round(4)
        )
        out["by_side"] = {k: v.to_dict() for k, v in by_side.iterrows()}
    by_rule = (
        traded.groupby("rule_id")
        .size()
        .to_dict()
        if not traded.empty
        else {}
    )
    out["fires_by_rule"] = by_rule
    return out


def main() -> None:
    """paper_fires 로드 → forward % → 리포트 저장."""
    fires = load_paper_fires(PAPER_FIRES_LOG)
    if fires.empty:
        print(f"[07] 발화 로그 없음: {PAPER_FIRES_LOG}")
        print("  Phase C 기간 06_execute_live.py (LIVE=0) 상시 실행 후 재시도")
        return

    mon = Monitor(cooldown_file=None)
    df = mon.read_candles_from_db(SYMBOL, MATCH_PRIMARY_INTERVAL, max_rows=50000)
    if df.empty:
        df = mon.get_coin_some_data(SYMBOL, MATCH_PRIMARY_INTERVAL)
    if not isinstance(df.index, pd.DatetimeIndex):
        df = df.set_index(pd.to_datetime(df["datetime"]))

    fires = attach_forward_returns(fires, df)
    report = summarize(fires)
    PAPER_WEEKLY_REPORT_JSON.parent.mkdir(parents=True, exist_ok=True)
    PAPER_WEEKLY_REPORT_JSON.write_text(
        json.dumps(report, ensure_ascii=False, indent=2),
        encoding="utf-8",
    )

    print(f"[07] 저장: {PAPER_WEEKLY_REPORT_JSON}")
    print(f"  기간 로그: {fires['ts'].min()} ~ {fires['ts'].max()}")
    print(f"  발화 {report['total_signals']} · 체결가정(would_trade) {report['would_trade_count']}")
    if "sum_forward_ret_pct" in report:
        print(
            f"  모의 forward 합산: {report['sum_forward_ret_pct']}% "
            f"(평균 {report['mean_forward_ret_pct']}%, "
            f"{MATCH_FORWARD_BARS}봉 후, 참고용)"
        )
    else:
        print("  forward 라벨 가능 건 없음 (봉 데이터 부족 또는 발화 없음)")


if __name__ == "__main__":
    main()