I have a class that will have an aiohttp.ClientSession object in it.
Normally when you use
async with aiohttp.ClientSession() as session:
# some code
The session will close since the session's __aexit__ method is called.
I cant use a context manager since I want to keep the session persistent for the entire lifetime of the object.
This works:
import asyncio
import aiohttp
class MyAPI:
def __init__(self):
self.session = aiohttp.ClientSession()
def __del__(self):
# Close connection when this object is destroyed
print('In __del__ now')
asyncio.shield(self.session.__aexit__(None, None, None))
async def main():
api = MyAPI()
asyncio.run(main())
However if in some place an exception is raised, the event loop is closed before the __aexit__ method is finished. How can I overcome this?
stacktrace:
Traceback (most recent call last):
File "/home/ron/.PyCharm2018.3/config/scratches/async.py", line 19, in <module>
asyncio.run(main())
File "/usr/local/lib/python3.7/asyncio/runners.py", line 43, in run
return loop.run_until_complete(main)
File "/usr/local/lib/python3.7/asyncio/base_events.py", line 568, in run_until_complete
return future.result()
File "/home/ron/.PyCharm2018.3/config/scratches/async.py", line 17, in main
raise ValueError
ValueError
In __del__ now
Exception ignored in: <function MyAPI.__del__ at 0x7f49982c0e18>
Traceback (most recent call last):
File "/home/ron/.PyCharm2018.3/config/scratches/async.py", line 11, in __del__
File "/usr/local/lib/python3.7/asyncio/tasks.py", line 765, in shield
File "/usr/local/lib/python3.7/asyncio/tasks.py", line 576, in ensure_future
File "/usr/local/lib/python3.7/asyncio/events.py", line 644, in get_event_loop
RuntimeError: There is no current event loop in thread 'MainThread'.
sys:1: RuntimeWarning: coroutine 'ClientSession.__aexit__' was never awaited
Unclosed client session
client_session: <aiohttp.client.ClientSession object at 0x7f49982c2e10>
Don't use a __del__
hook to clean up asynchronous resources. You can't count it being called at all, let alone control when it'll be used or if the async loop is still available at that time. You really want to handle this explicitly.
Either make the API an async context manager, or otherwise explicitly clean up resources at exit, with a finally
handler, say; the with
and async with
statements are basically designed to encapsulate resource cleanup traditionally handled in finally
blocks.
I'd make the API
instance a context manager here:
class MyAPI:
def __init__(self):
self.session = aiohttp.ClientSession()
async def __aenter__(self):
return self
async def __aexit__(self, *excinfo):
await self.session.close()
Note that all that ClientSession.__aexit__()
really does is await on self.close()
, so the above goes straight to that coroutine.
Then use this in your main loop with:
async def main():
async with MyAPI() as api:
pass
Another option is to supply your own session object to the MyAPI
instance and take responsibility yourself for closing it when you are done:
class MyAPI:
def __init__(self, session):
self.session = session
async def main():
session = aiohttp.ClientSession()
try:
api = MyAPI(session)
# do things with the API
finally:
await session.close()