qlib-quant-platform / update_data.py
Serena2026's picture
Fix: correct data dir path detection and clean up update_data.py
566df98 verified
Raw History Blame Contribute Delete
8.74 kB
"""
Qlib A股数据自动更新脚本
基于 Yahoo Finance 获取最新数据并更新 Qlib 数据库
"""
import os
import sys
import logging
from datetime import datetime, timedelta
from pathlib import Path
logging.basicConfig(
level=logging.INFO,
format="%(asctime)s [%(levelname)s] %(message)s",
handlers=[logging.StreamHandler()]
)
logger = logging.getLogger(__name__)
# 自动检测数据目录(兼容 Docker 容器和本地环境)
_CANDIDATES = [
"/root/.qlib/qlib_data/cn_data",
os.path.expanduser("~/.qlib/qlib_data/cn_data"),
]
QLIB_DATA_DIR = next((p for p in _CANDIDATES if os.path.exists(p)), _CANDIDATES[0])
def get_last_date_from_calendar():
"""从日历文件直接读取最新日期"""
calendar_file = Path(QLIB_DATA_DIR) / "calendars" / "day.txt"
if calendar_file.exists():
lines = [l.strip() for l in calendar_file.read_text().splitlines() if l.strip()]
if lines:
last = lines[-1]
logger.info(f"日历文件最新日期: {last}")
return last
return "2020-09-25"
def update_calendar_file(new_dates):
"""向日历文件追加新的交易日"""
calendar_file = Path(QLIB_DATA_DIR) / "calendars" / "day.txt"
if not calendar_file.exists():
logger.warning("日历文件不存在,跳过更新")
return
existing = set(l.strip() for l in calendar_file.read_text().splitlines() if l.strip())
to_add = [d for d in new_dates if d not in existing]
if to_add:
with open(calendar_file, 'a') as f:
for d in sorted(to_add):
f.write(d + '\n')
logger.info(f"日历文件新增 {len(to_add)} 个交易日(最新: {sorted(to_add)[-1]})")
else:
logger.info("日历文件无需更新")
def update_instruments_end_date(end_date):
"""更新 instruments 文件中每只股票的结束日期"""
instruments_dir = Path(QLIB_DATA_DIR) / "instruments"
for fname in ["all.txt", "csi300.txt", "csi500.txt"]:
fpath = instruments_dir / fname
if not fpath.exists():
continue
lines = fpath.read_text().splitlines()
new_lines = []
for line in lines:
parts = line.split("\t")
if len(parts) >= 3:
parts[2] = end_date
new_lines.append("\t".join(parts))
else:
new_lines.append(line)
fpath.write_text("\n".join(new_lines) + "\n")
logger.info(f"instruments 文件结束日期已更新为 {end_date}")
def get_stock_list():
"""读取全部股票代码"""
all_file = Path(QLIB_DATA_DIR) / "instruments" / "all.txt"
stocks = []
if all_file.exists():
for line in all_file.read_text().splitlines():
parts = line.strip().split("\t")
if parts and parts[0]:
stocks.append(parts[0])
if not stocks:
stocks = [
"SH600519", "SH600036", "SH601318", "SH600276", "SH600000",
"SH600887", "SH600030", "SH601012", "SH600031", "SH600900",
"SZ000001", "SZ000858", "SZ000333", "SZ300750", "SZ002415",
"SZ000002", "SZ000651", "SZ002594", "SZ300059", "SZ000568",
]
return stocks
def to_yahoo_code(qlib_code):
if qlib_code.startswith("SH"):
return qlib_code[2:] + ".SS"
elif qlib_code.startswith("SZ"):
return qlib_code[2:] + ".SZ"
return qlib_code
def download_and_write(stocks, start_date, end_date):
"""下载 Yahoo Finance 数据并写入 Qlib 二进制格式"""
try:
import yfinance as yf
import pandas as pd
import numpy as np
except ImportError:
os.system("pip install yfinance pandas numpy -q")
import yfinance as yf
import pandas as pd
import numpy as np
features_dir = Path(QLIB_DATA_DIR) / "features"
features_dir.mkdir(parents=True, exist_ok=True)
all_trading_dates = set()
success_count = 0
batch_size = 30
# 只更新已有数据的股票
existing_stocks = [s for s in stocks if (features_dir / s.lower()).exists()]
if not existing_stocks:
existing_stocks = stocks[:100]
target_stocks = existing_stocks[:300]
logger.info(f"准备更新 {len(target_stocks)} 只股票的数据...")
for i in range(0, len(target_stocks), batch_size):
batch = target_stocks[i:i + batch_size]
yahoo_batch = [to_yahoo_code(c) for c in batch]
try:
if len(yahoo_batch) == 1:
raw = yf.download(yahoo_batch[0], start=start_date, end=end_date,
auto_adjust=True, progress=False)
if not raw.empty:
raw.columns = [c.lower() for c in raw.columns]
data_map = {batch[0]: raw}
else:
data_map = {}
else:
raw = yf.download(yahoo_batch, start=start_date, end=end_date,
group_by='ticker', auto_adjust=True,
progress=False, threads=True)
data_map = {}
if not raw.empty:
for qlib_code, ycode in zip(batch, yahoo_batch):
try:
if ycode in raw.columns.get_level_values(0):
df = raw[ycode].copy()
df.columns = [c.lower() for c in df.columns]
df = df.dropna(how='all')
if not df.empty:
data_map[qlib_code] = df
except Exception:
pass
for qlib_code, df in data_map.items():
if df.empty:
continue
df.index = pd.to_datetime(df.index).tz_localize(None)
trading_dates = [d.strftime('%Y-%m-%d') for d in df.index]
all_trading_dates.update(trading_dates)
stock_dir = features_dir / qlib_code.lower()
stock_dir.mkdir(parents=True, exist_ok=True)
field_map = {'close': 'close', 'open': 'open',
'high': 'high', 'low': 'low', 'volume': 'volume'}
for field, col in field_map.items():
if col not in df.columns:
continue
series = df[col].dropna().astype(np.float32)
if series.empty:
continue
bin_file = stock_dir / f"{field}.day.bin"
if bin_file.exists():
existing_data = np.fromfile(str(bin_file), dtype=np.float32)
combined = np.concatenate([existing_data, series.values])
combined.tofile(str(bin_file))
else:
series.values.tofile(str(bin_file))
success_count += 1
except Exception as e:
logger.debug(f"批次 {i//batch_size+1} 下载失败: {e}")
continue
if (i // batch_size + 1) % 5 == 0:
logger.info(f"进度: {min(i+batch_size, len(target_stocks))}/{len(target_stocks)} 只")
logger.info(f"数据写入完成: 成功 {success_count}/{len(target_stocks)} 只")
return success_count, sorted(all_trading_dates)
def main():
import pandas as pd
today = datetime.today().strftime("%Y-%m-%d")
logger.info("=" * 50)
logger.info("Qlib 数据增量更新开始")
logger.info("=" * 50)
if not Path(QLIB_DATA_DIR).exists():
logger.error(f"Qlib 数据目录不存在: {QLIB_DATA_DIR}")
return
last_date = get_last_date_from_calendar()
start_date = (datetime.strptime(last_date, "%Y-%m-%d") + timedelta(days=1)).strftime("%Y-%m-%d")
if start_date >= today:
logger.info("✅ 数据已是最新,无需更新")
return
logger.info(f"增量更新: {start_date} → {today}")
stocks = get_stock_list()
logger.info(f"股票总数: {len(stocks)}")
count, trading_dates = download_and_write(stocks, start_date, today)
if count > 0:
if trading_dates:
update_calendar_file(trading_dates)
else:
bdays = pd.bdate_range(start=start_date, end=today)
update_calendar_file([d.strftime('%Y-%m-%d') for d in bdays])
update_instruments_end_date(today)
logger.info(f"🎉 数据更新完成!已更新 {count} 只股票,数据覆盖至 {today}")
else:
logger.warning("⚠️ 未能更新任何股票数据,将使用现有数据继续运行")
if __name__ == "__main__":
main()