文章目录
1. 天勤量化不同版本区别
为了获取所有期货分钟数据和股票日线数据，需要提前申请专业版账号，值得高兴的是每个账号有15天的试用期，官网地址。但是股票和期货数据只能获取2018年之后的，期货合约代码最久的是20年8月到期。
2. 多进程获取期货分钟数据和股票日线数据
import logging
import os.path
from datetime import datetime, timedelta
from typing import Union
import ray
import pandas as pd
from pandas import Series, Timestamp, DataFrame
from tqsdk import TqApi, TqAuth
SETTING = {"user": "your_user", "password": "your_password",
"FUTURE": "future_symbols.csv", "STOCK": "stock_symbols.csv"}
@ray.remote
class TraceData:
def __init__(self, account, password):
self.symbol = None
self.klines_flag = False
self.exchange = None
self.logger = None
self.api = TqApi(auth=TqAuth(account, password))
self.root_dir = os.path.abspath(os.path.curdir)
self.init()
def init(self, exchange: str = None, symbol: str = None) -> None:
# 订阅数据需要的字段
self.exchange = exchange
self.symbol = symbol
# 检查klines和log目录是否创建
klines_dir = os.path.join(self.root_dir, "klines")
if not os.path.exists(klines_dir):
os.mkdir(klines_dir)
log_dir = os.path.join(self.root_dir, "log")
if not os.path.exists(log_dir):
os.mkdir(log_dir)
# 准备日志记录工具
self.logger = logging.getLogger("loging")
self.logger.setLevel(logging.DEBUG)
formatter = logging.Formatter("%(asctime)s [%(levelname)s] %(message)s")
logfile_path = os.path.join(os.path.join(self.root_dir, "log"),
(datetime.now().date().strftime('%Y%m%d')))
file_handler = logging.FileHandler(logfile_path, mode="a",
encoding="utf8")
file_handler.setLevel(logging.INFO)
file_handler.setFormatter(formatter)
self.logger.addHandler(file_handler)
# 获取所有交易标的合约
def get_all_symbols(self, ins_class, expired=False) -> list:
"""
ins_class (str): [可选] 合约类型
* FUTURE: 期货
* STOCK: 股票
"""
exchanges = []
all_symbols = []
if ins_class == "FUTURE":
exchanges = ["SHFE", "CFFEX", "DCE", "CZCE", "INE"]
elif ins_class == "STOCK":
exchanges = ["SSE", "SZSE"]
for exchange in exchanges:
symbol = self.api.query_quotes(ins_class=ins_class,
exchange_id=exchange,
expired=expired)
all_symbols.extend(symbol)
df: Series = pd.Series(all_symbols,
index=[i + 1 for i in range(len(all_symbols))])
filename = SETTING.get(ins_class)
filepath = os.path.join(self.root_dir, filename)
if not os.path.exists(filepath):
df.to_csv(filepath, index=True, header=False)
return all_symbols
def save_klines(self, symbols: list):
"""下载指定标的k线数据"""
klines_dir_path = os.path.join(os.path.join(self.root_dir, "klines"),
datetime.now().date().strftime('%Y%m%d'))
if not os.path.exists(klines_dir_path):
os.mkdir(klines_dir_path)
for symbol in symbols:
klines_file_path = os.path.join(klines_dir_path, f"{symbol}.csv")
if os.path.exists(klines_file_path):
continue
klines = pd.DataFrame()
try:
klines = self.api.get_kline_serial(symbol, 60, 600)
except Exception as e:
self.logger.log(logging.WARNING, f"{e}")
print(f"{datetime.now()}:{e}")
if not klines.empty:
klines_copy = klines.copy(deep=True)
klines_copy["new_datetime"]: datetime = klines_copy[
"datetime"].apply(
lambda x: Timestamp(x).to_pydatetime() + timedelta(hours=8))
local_time = datetime.now()
klines_copy = klines_copy[
(klines_copy.new_datetime >= datetime(local_time.year,
local_time.month,
local_time.day - 1,
15,
30)) & (
klines_copy.new_datetime < datetime(local_time.year,
local_time.month,
local_time.day,
15,
30))]
klines_copy["date"] = klines_copy["new_datetime"].apply(
lambda x: x.date().strftime("%Y%m%d"))
klines_copy["time"] = klines_copy["new_datetime"].apply(
lambda x: x.time().strftime("%H:%M:%S"))
klines_copy = klines_copy.drop(["new_datetime", "datetime"],
axis=1)
klines_copy.to_csv(klines_file_path, index=False)
# 输出日志
self.logger.log(logging.INFO, f"{symbol}.csv文件创建完成！")
print(f"{datetime.now()},{symbol}.csv文件创建完成！")
else:
# 输出日志
self.logger.log(logging.WARNING, f"{symbol}.csv文件为空！")
print(f"{datetime.now()},{symbol}.csv文件为空！")
def save_bars(self, symbols: list, duration_seconds: int, start: datetime,
end: datetime,
adj_type: Union[str, None] = None):
"""下载指定标的k线数据
adj_type (str/None): [可选]指定复权类型，默认为 None。adj_type 参数只对股票和基金类型合约有效。\
"F" 表示前复权；"B" 表示后复权；None 表示不做处理。
"""
if adj_type == "F":
klines_dir_path = os.path.join(self.root_dir,
f"F_klines_{str(duration_seconds)}s")
elif adj_type == "B":
klines_dir_path = os.path.join(self.root_dir,
f"B_klines_{str(duration_seconds)}s")
else:
klines_dir_path = os.path.join(self.root_dir,
f"klines_{str(duration_seconds)}s")
if not os.path.exists(klines_dir_path):
os.mkdir(klines_dir_path)
klines = pd.DataFrame()
for symbol in symbols:
klines_file_path = os.path.join(klines_dir_path, f"{symbol}.csv")
if os.path.exists(klines_file_path):
continue
try:
klines = self.api.get_kline_data_series(symbol,
duration_seconds, start,
end, adj_type)
except Exception as e:
self.logger.log(logging.WARNING, f"{e}")
print(f"{datetime.now()}:{e}")
if not klines.empty:
klines_copy = klines.copy(deep=True)
klines_copy["new_datetime"]: datetime = klines_copy[
"datetime"].apply(
lambda x: Timestamp(x).to_pydatetime() + timedelta(hours=8))
klines_copy["date"] = klines_copy["new_datetime"].apply(
lambda x: x.date().strftime("%Y%m%d"))
klines_copy["time"] = klines_copy["new_datetime"].apply(
lambda x: x.time().strftime("%H:%M:%S"))
klines_copy = klines_copy.drop(["new_datetime", "datetime"],
axis=1)
klines_copy.to_csv(klines_file_path, index=False)
# 输出日志
self.logger.log(logging.INFO, f"{symbol}.csv文件创建完成！")
print(f"{datetime.now()},{symbol}.csv文件创建完成！")
else:
# 输出日志
self.logger.log(logging.WARNING, f"{symbol}.csv文件为空！")
print(f"{datetime.now()},{symbol}.csv文件为空！")
def download_today_klines(task_num, ins_class) -> None:
"""
task_num: 进程数
ins_class = FUTURE/STOCK
"""
symbols_filepath = SETTING.get(ins_class)
if not os.path.exists(symbols_filepath):
tq = TraceData.remote(SETTING.get("user"), SETTING.get("password"))
symbols = ray.get(tq.get_all_symbols.remote(ins_class=ins_class))
ray.shutdown()
else:
symbols = pd.read_csv(symbols_filepath)
symbols = list(symbols.iloc[:, 1].values)
start_time = datetime.now()
tqs = [TraceData.remote(SETTING.get("user"), SETTING.get("password")) for _
in range(task_num)]
length = len(symbols) // task_num
task_id = []
for i in range(task_num):
if i == task_num - 1:
symbols_part = symbols[i * length:]
else:
symbols_part = symbols[i * length:(i + 1) * length]
id_ = tqs[i].save_klines.remote(symbols_part)
task_id.append(id_)
ray.get(task_id)
end_time = datetime.now()
print(end_time - start_time)
def download_history_klines(task_num, ins_class, start, end,
adj_type: Union[str, None] = None) -> None:
"""
task_num: 进程数
ins_class = FUTURE/STOCK
start: 开始时间
end: 结束时间
"""
symbols_filepath = SETTING.get(ins_class)
expired = True if ins_class == "FUTURE" else False
if not os.path.exists(symbols_filepath):
tq = TraceData.remote(SETTING.get("user"), SETTING.get("password"))
symbols = ray.get(tq.get_all_symbols.remote(
ins_class=ins_class,
expired=expired))
ray.shutdown()
else:
symbols = pd.read_csv(symbols_filepath)
symbols = list(symbols.iloc[:, 1].values)
start_time = datetime.now()
tqs = [TraceData.remote(SETTING.get("user"), SETTING.get("password")) for _
in range(task_num)]
length = len(symbols) // task_num
task_id = []
for i in range(task_num):
if i == task_num - 1:
symbols_part = symbols[i * length:]
else:
symbols_part = symbols[i * length:(i + 1) * length]
duration_seconds = 60 if ins_class == "FUTURE" else 86400
id_ = tqs[i].save_bars.remote(symbols_part,
duration_seconds=duration_seconds,
start=start, end=end,
adj_type=adj_type) # 确保数据都可以下载到
task_id.append(id_)
ray.get(task_id)
end_time = datetime.now()
print(end_time - start_time)
if __name__ == '__main__':
download_history_klines(8, ins_class="FUTURE", start=datetime(2018, 1, 1),
end=datetime(2022, 10, 23))
download_history_klines(8, ins_class="STOCK", start=datetime(2018, 1, 1),
end=datetime(2022, 10, 23), adj_type="F")
download_history_klines(8, ins_class="STOCK", start=datetime(2018, 1, 1),
end=datetime(2022, 10, 23), adj_type="B")
download_history_klines(8, ins_class="STOCK", start=datetime(2018, 1, 1),
end=datetime(2022, 10, 23))
运行截图：
读者可以根据自己电脑cpu的数量选择合适的task_num,建议4核用户选择4个进程。
3. 多进程获取期货分钟数据并且每日实时更新历史数据
import logging
import os.path
from datetime import datetime, timedelta
import ray
import pandas as pd
from pandas import Series, Timestamp, DataFrame
from tqsdk import TqApi, TqAuth
SETTING = {"user": "your_user", "password": "your_passwrod",
"FUTURE": "future_symbols.csv"}
@ray.remote
class TraceData:
def __init__(self, account, password):
self.symbol = None
self.exchange = None
self.logger = None
self.api = TqApi(auth=TqAuth(account, password))
self.root_dir = r"D:\MarketData" # 可以更改主路径位置
self.init()
def init(self) -> None:
# 检查目录是否创建
klines_dir = os.path.join(self.root_dir, "FutureData")
if not os.path.exists(klines_dir):
os.makedirs(klines_dir)
log_dir = os.path.join(self.root_dir, "log")
if not os.path.exists(log_dir):
os.makedirs(log_dir)
# 准备日志记录工具
self.logger = logging.getLogger("loging")
self.logger.setLevel(logging.DEBUG)
formatter = logging.Formatter("%(asctime)s [%(levelname)s] %(message)s")
logfile_path = os.path.join(self.root_dir, "log", str(datetime.now().date().strftime('%Y%m%d')) + ".log")
file_handler = logging.FileHandler(logfile_path, mode="a",
encoding="utf8")
file_handler.setLevel(logging.INFO)
file_handler.setFormatter(formatter)
self.logger.addHandler(file_handler)
# 获取所有交易标的合约
def get_all_symbols(self, ins_class, expired=False) -> list:
"""
ins_class (str): [可选] 合约类型
* FUTURE: 期货
* STOCK: 股票
"""
exchanges = []
all_symbols = []
if ins_class == "FUTURE":
exchanges = ["SHFE", "CFFEX", "DCE", "CZCE", "INE"]
elif ins_class == "STOCK":
exchanges = ["SSE", "SZSE"]
for exchange in exchanges:
symbol = self.api.query_quotes(ins_class=ins_class,
exchange_id=exchange,
expired=expired)
all_symbols.extend(symbol)
df: Series = pd.Series(all_symbols, index=[i + 1 for i in range(len(all_symbols))])
filepath = os.path.join(self.root_dir, SETTING.get(ins_class))
if not os.path.exists(filepath):
df.to_csv(filepath, index=True, header=False)
return all_symbols
def save_klines(self, symbols: list):
"""下载指定标的k线数据"""
# 指定下载目录
klines_dir_path = os.path.join(self.root_dir, "FutureData", "1min", "tq")
if not os.path.exists(klines_dir_path):
os.makedirs(klines_dir_path)
for symbol in symbols:
klines_file_name: str = f"{symbol}.1min.csv"
klines_file_path = os.path.join(klines_dir_path, klines_file_name)
# 如果文件夹有文件，则更新
if not os.path.exists(klines_file_path):
continue
klines: DataFrame = pd.DataFrame()
try:
klines: DataFrame = self.api.get_kline_serial(symbol, 60, 600)
except Exception as e:
self.logger.log(logging.ERROR, f"{e}")
print(f"{datetime.now()}:{e}")
if not klines.empty:
# 合成指定格式的DataFrame
klines_copy = klines.copy(deep=True)
klines_copy["new_datetime"]: datetime = klines_copy[
"datetime"].apply(
lambda x: Timestamp(x).to_pydatetime() + timedelta(hours=8))
local_time = datetime.now()
klines_copy = klines_copy[
(klines_copy.new_datetime >= datetime(local_time.year,
local_time.month,
local_time.day - 1,
15,
30)) & (
klines_copy.new_datetime < datetime(local_time.year,
local_time.month,
local_time.day,
15,
30))]
klines_copy["date"] = klines_copy["new_datetime"].apply(
lambda x: x.date().strftime("%Y%m%d"))
klines_copy["time"] = klines_copy["new_datetime"].apply(
lambda x: x.time().strftime("%H:%M:%S"))
klines_copy: DataFrame = klines_copy.drop(["id", "new_datetime", "datetime", "duration"],
axis=1)
try:
# before_kines历史CSV文件 klines_copy 当前的数据 df合成后的数据
before_kines: DataFrame = pd.read_csv(klines_file_path)
df: DataFrame = pd.concat([before_kines, klines_copy])
# 根据date和time去重
df.drop_duplicates(
subset=['date', 'time'],
keep='first',
inplace=True)
df.to_csv(klines_file_path, index=False)
self.logger.log(logging.INFO, f"{klines_file_name}文件更新完成！")
print(f"{datetime.now()},{klines_file_name}文件更新完成！")
except Exception as e:
self.logger.log(logging.ERROR, e)
else:
# 输出日志
self.logger.log(logging.WARNING, f"{klines_file_name}下载当天数据为空")
print(f"{datetime.now()},{klines_file_name}下载当天数据为空")
def save_bars(self, symbols: list, duration_seconds: int, start: datetime,
end: datetime):
"""下载指定标的k线数据
adj_type (str/None): [可选]指定复权类型，默认为 None。adj_type 参数只对股票和基金类型合约有效。\
"F" 表示前复权；"B" 表示后复权；None 表示不做处理。
"""
klines_dir_path = os.path.join(self.root_dir, "FutureData", f"{duration_seconds // 60}min", "tq")
if not os.path.exists(klines_dir_path):
os.makedirs(klines_dir_path)
klines = pd.DataFrame()
for symbol in symbols:
klines_file_name: str = f"{symbol}.{duration_seconds // 60}min.csv"
klines_file_path = os.path.join(klines_dir_path, klines_file_name)
if os.path.exists(klines_file_path):
continue
try:
klines = self.api.get_kline_data_series(symbol,
duration_seconds, start,
end)
except Exception as e:
self.logger.log(logging.ERROR, f"{e}")
print(f"{datetime.now()}:{e}")
if not klines.empty:
klines_copy = klines.copy(deep=True)
klines_copy["new_datetime"]: datetime = klines_copy[
"datetime"].apply(
lambda x: Timestamp(x).to_pydatetime() + timedelta(hours=8))
klines_copy["date"] = klines_copy["new_datetime"].apply(
lambda x: x.date().strftime("%Y%m%d"))
klines_copy["time"] = klines_copy["new_datetime"].apply(
lambda x: x.time().strftime("%H:%M:%S"))
klines_copy = klines_copy.drop(["new_datetime", "datetime", "id", "duration"],
axis=1)
klines_copy.to_csv(klines_file_path, index=False)
# 输出日志
self.logger.log(logging.INFO, f"{klines_file_name}文件创建完成！")
print(f"{datetime.now()},{klines_file_name}文件创建完成！")
else:
# 输出日志
self.logger.log(logging.WARNING, f"{klines_file_name}文件为空！")
print(f"{datetime.now()}{symbol}.{klines_file_name}文件为空！")
def download_today_klines(task_num, ins_class) -> None:
"""
task_num: 进程数
ins_class:FUTURE
"""
symbols_filepath = SETTING.get(ins_class)
if not os.path.exists(symbols_filepath):
tq = TraceData.remote(SETTING.get("user"), SETTING.get("password"))
symbols = ray.get(tq.get_all_symbols.remote(ins_class=ins_class))
ray.shutdown()
else:
symbols = pd.read_csv(symbols_filepath)
symbols = list(symbols.iloc[:, 1].values)
start_time = datetime.now()
tqs = [TraceData.remote(SETTING.get("user"), SETTING.get("password")) for _
in range(task_num)]
length = len(symbols) // task_num
task_id = []
for i in range(task_num):
if i == task_num - 1:
symbols_part = symbols[i * length:]
else:
symbols_part = symbols[i * length:(i + 1) * length]
id_ = tqs[i].save_klines.remote(symbols_part)
task_id.append(id_)
ray.get(task_id)
end_time = datetime.now()
print(end_time - start_time)
def download_history_klines(task_num, ins_class, start, end) -> None:
"""
task_num: 进程数
ins_class = FUTURE
start: 开始时间
end: 结束时间
注意： 只能下载2018年1月2日以后的数据
"""
symbols_filepath = SETTING.get(ins_class)
if not os.path.exists(symbols_filepath):
tq = TraceData.remote(SETTING.get("user"), SETTING.get("password"))
symbols = ray.get(tq.get_all_symbols.remote(
ins_class=ins_class,
expired=False))
ray.shutdown()
else:
symbols = pd.read_csv(symbols_filepath)
symbols = list(symbols.iloc[:, 1].values)
start_time = datetime.now()
tqs = [TraceData.remote(SETTING.get("user"), SETTING.get("password")) for _
in range(task_num)]
length = len(symbols) // task_num
task_id = []
for i in range(task_num):
if i == task_num - 1:
symbols_part = symbols[i * length:]
else:
symbols_part = symbols[i * length:(i + 1) * length]
duration_seconds = 60 if ins_class == "FUTURE" else 86400
id_ = tqs[i].save_bars.remote(symbols_part,
duration_seconds=duration_seconds,
start=start, end=end)
task_id.append(id_)
ray.get(task_id)
end_time = datetime.now()
print(end_time - start_time)
if __name__ == '__main__':
# 先使用download_history_klines()函数下载历史数据，如果试用期限过了，
# 可以每天3点半后运行download_today_klines()函数会自动拼接历史数据
# download_history_klines(8, ins_class="FUTURE", start=datetime(2018, 1, 2),
# end=datetime.now().date())
download_today_klines(8, "FUTURE")
4.常见问题
4.1 ray包无法安装
ray官方文档：https://docs.ray.io/en/latest/ray-overview/installation.html
进入官方文档
下载对应的wheel包
将包放到工程文件下，进入pycharm的terminal界面，运行如下命令
# Clean removal of previous install
pip uninstall -y ray
# Install Ray with support for the dashboard + cluster launcher
pip install -U "ray[default] @ LINK_TO_WHEEL.whl"
# Install Ray with minimal dependencies
# pip install -U LINK_TO_WHEEL.whl
4.2 免费账户能否获得当天股票行情？
不行，如果要获取免费的股票行情，可以尝试用tushare或者rqdata，借用veighna框架可以非常轻松地将数据保存到本地数据库，笔者后续会出一篇博文。
4.3 下载数据超时，卡顿严重怎么办？
下载超时等待时间长，修改tqsdk源代码
pycharm使用curl+鼠标左键进入源码
第716行，将25改成8，缩短等待时间。
923
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
