codex_jxs_code/strategy/data_loader.py

"""
数据加载模块 - 从 SQLite 加载多周期K线数据为 DataFrame
"""
import numpy as np
import pandas as pd
from peewee import SqliteDatabase
from pathlib import Path

DB_PATH = Path(__file__).parent.parent / 'models' / 'database.db'

# 周期 -> 表名 (bitmart / binance)
PERIOD_MAP = {
    '1s': 'bitmart_eth_1s',
    '1m': 'bitmart_eth_1m',
    '3m': 'bitmart_eth_3m',
    '5m': 'bitmart_eth_5m',
    '15m': 'bitmart_eth_15m',
    '30m': 'bitmart_eth_30m',
    '1h': 'bitmart_eth_1h',
}

BINANCE_PERIOD_MAP = {
    '1s': 'binance_eth_1s',
    '1m': 'binance_eth_1m',
    '3m': 'binance_eth_3m',
    '5m': 'binance_eth_5m',
    '15m': 'binance_eth_15m',
    '30m': 'binance_eth_30m',
    '1h': 'binance_eth_1h',
}


def load_klines(period: str, start_date: str, end_date: str, tz: str | None = None,
                source: str = "bitmart") -> pd.DataFrame:
    """
    加载指定周期、指定日期范围的K线数据
    :param period: '1s','1m','3m','5m','15m','30m','1h'
    :param start_date: 'YYYY-MM-DD'
    :param end_date: 'YYYY-MM-DD' (不包含该日)
    :param tz: 日期解释的时区，如 'Asia/Shanghai' 表示按北京时间；None 则用本地时区
    :return: DataFrame with columns: datetime, open, high, low, close
    """
    period_map = BINANCE_PERIOD_MAP if source == "binance" else PERIOD_MAP
    table = period_map.get(period)
    if not table:
        raise ValueError(f"不支持的周期: {period}, 可选: {list(PERIOD_MAP.keys())}")

    if tz:
        start_ts = int(pd.Timestamp(start_date, tz=tz).timestamp() * 1000)
        end_ts = int(pd.Timestamp(end_date, tz=tz).timestamp() * 1000)
    else:
        start_ts = int(pd.Timestamp(start_date).timestamp() * 1000)
        end_ts = int(pd.Timestamp(end_date).timestamp() * 1000)

    db = SqliteDatabase(str(DB_PATH))
    db.connect()
    cursor = db.execute_sql(
        f'SELECT id, open, high, low, close FROM [{table}] '
        f'WHERE id >= ? AND id < ? ORDER BY id',
        (start_ts, end_ts)
    )
    rows = cursor.fetchall()
    db.close()

    df = pd.DataFrame(rows, columns=['timestamp_ms', 'open', 'high', 'low', 'close'])
    df['datetime'] = pd.to_datetime(df['timestamp_ms'], unit='ms')
    df.set_index('datetime', inplace=True)
    df.drop(columns=['timestamp_ms'], inplace=True)
    df = df.astype(float)
    return df


def load_multi_period(periods: list, start_date: str, end_date: str, source: str = "bitmart") -> dict:
    """
    加载多个周期的数据
    :return: {period: DataFrame}
    """
    result = {}
    for p in periods:
        result[p] = load_klines(p, start_date, end_date, source=source)
        print(f"  加载 {p}: {len(result[p])} 条 ({start_date} ~ {end_date})")
    return result


def get_1m_touch_direction(df_5m: pd.DataFrame, df_1m: pd.DataFrame,
                           arr_mid: np.ndarray, kline_step_min: int = 5) -> np.ndarray:
    """
    根据 1 分钟线判断每根 5m K 线「先涨碰到均线」还是「先跌碰到均线」。
    返回: 1=先涨碰到(可开多), -1=先跌碰到(可开空), 0=未碰到或无法判断
    """
    df_1m = df_1m.copy()
    df_1m["_bucket"] = df_1m.index.floor(f"{kline_step_min}min")

    # 5m 索引与 mid 对齐
    mid_sr = pd.Series(arr_mid, index=df_5m.index)

    touch_map: dict[pd.Timestamp, int] = {}
    for bucket, grp in df_1m.groupby("_bucket", sort=True):
        mid = mid_sr.get(bucket, np.nan)
        if pd.isna(mid):
            touch_map[bucket] = 0
            continue

        o = grp["open"].to_numpy(dtype=float)
        h = grp["high"].to_numpy(dtype=float)
        l_ = grp["low"].to_numpy(dtype=float)

        touch = 0
        for j in range(len(grp)):
            if l_[j] <= mid <= h[j]:
                touch = 1 if o[j] < mid else -1
                break
        touch_map[bucket] = touch

    # 对齐到主周期 index
    out = np.zeros(len(df_5m), dtype=np.int32)
    for i, t in enumerate(df_5m.index):
        out[i] = touch_map.get(t, 0)
    return out


if __name__ == '__main__':
    data = load_multi_period(['5m', '15m', '1h'], '2020-01-01', '2024-01-01')
    for k, v in data.items():
        print(f"{k}: {v.shape}, {v.index[0]} ~ {v.index[-1]}")