Consider (where queue is an asyncio.Queue, and I'm using the asyncio websockets library which uses the same server interface as asyncio in the standard library):
async def handler(websocket: websockets.WebSocketServerProtocol, path):
while True:
data = await queue.get()
await websocket.send(data)
What I'm finding is that the ConnectionClosed exception (triggered either when a client disconnects, and is also used for server shutdown) is only raised, obviously, within the loop when we are awaiting on send(). This can happen long after a websocket client has disconnected, because we will spend most of the time awaiting on queue.get().
Furthermore, I'm trying to implement graceful shutdown of the server. That code looks very similar to the standard
server.close()
loop.run_until_complete(server.wait_closed())
The issue is that my websocket handler, i.e., handler(), nevertheless still continues to wait on the queue for more data before raising the exceptions that wait_closed() placed on the event loop.
My question: is it correct to introduce the use of wait_for() to make the handler() function more responsive to these management events:
async def handler(websocket: websockets.WebSocketServerProtocol, path):
while True:
try:
data = await loop.wait_for(queue.get(), 5.0) # 5 seconds
await websocket.send(data)
except asyncio.TimeoutError:
logging.debug('Refreshing the loop')
Most of the asyncio examples I've found online have been quite simplistic and don't seem to cover these kinds of issues.
I could also put junk on the queue during my shutdown sequence, which would allow the exception to be raised, but that wouldn't help catching client disconnections earlier.
Any good references with examples would be much appreciated. I'm finding the official asyncio docs somewhat terse.
