Update , move connection out of the loop, so that it can be reused, fixing another memory leak issue.

main
Guofu Li 2 years ago
parent 68610470c4
commit 59b970532e

@ -119,12 +119,15 @@ class DDBHFTLoader(DDBLoader):
# 不能重复创建Pool对象因此需要在循环的最外侧创建好Pool对象然后传参进去
with Pool(self.num_workers if num_workers is None else num_workers) as pool:
# Always reuse the connection object, to reduce the memory consumption.
with self.mssql_engine.connect() as conn:
# Loop through the stock list.
for hft_type_name in self.hft_type_list:
print('Will work on hft type:', hft_type_name)
with tqdm(stock_list) as pbar:
for stock_id in pbar:
pbar.set_description(f"Working on stock {stock_id}")
self.dump_hft_to_ddb(hft_type_name, stock_id, pbar=pbar, pool=pool)
self.dump_hft_to_ddb(hft_type_name, stock_id, conn, pbar=pbar, pool=pool)
def _get_stock_date_list(self, cache=False):
@ -354,7 +357,7 @@ class DDBHFTLoader(DDBLoader):
print('-' * 80)
def dump_hft_to_ddb(self, type_name, stock_id, trade_date=None, pbar=None, pool=None):
def dump_hft_to_ddb(self, type_name, stock_id, conn, trade_date=None, pbar=None, pool=None):
if (type_name, stock_id, 'OK') in self.dump_journal_df.index:
message = f"Will skip ({type_name}, {stock_id}) as it appears in the dump journal."
if pbar is None:
@ -376,7 +379,6 @@ class DDBHFTLoader(DDBLoader):
# 经过尝试按个股来做batch查询效率还是可以接受的
# mssql中索引字段是(S_INFO_WINDCODE, TRADE_DT)
with self.mssql_engine.connect() as conn:
stat = """
select * from [Level2Bytes{mssql_type_name}].dbo.[{mssql_type_name}]
where S_INFO_WINDCODE='{stock_id}'

Loading…
Cancel
Save