WebSocketClientFactory's _batched_timer SHOULD use the passed event loop

Hey,

here’s my second issue report in a week:

The WebSocketClientFactory can be passed an event loop:

https://github.com/crossbario/autobahn-python/blob/bba4ed8418e59a678012ecb466092bee250b906f/autobahn/asyncio/websocket.py#L288-L289

… and WebSocketClientFactory’s _batched_timer SHOULD use it, which it doesn’t now: https://github.com/crossbario/autobahn-python/blob/bba4ed8418e59a678012ecb466092bee250b906f/autobahn/websocket/protocol.py#L3798-L3801

Why is this a problem? Because if someone doesn’t have a main event loop, autoPings won’t work. My application doesn’t have a main event loop, it only has an event loop in a separate thread, which I initialize later, along with the websocket client.

As far as I’ve seen, txaio uses the main event loop, and the websocket client can’t be told to use other than the main initialized one.

Correct me if I’m wrong, but now I have to have an asynchronous main event loop, just to have autopings working?

I can override the event loop with my one from the thread I start later with:

from asyncio import BaseEventLoop, new_event_loop
from txaio.aio import config
from threading import Thread

class SomeThread(Thread):
    def run(self):
        self.loop = new_event_loop()  # type: BaseEventLoop
        config.loop = self.loop

But… really? I don’t think this would be a nice solution, it looks rather like a hack to me. Also it’s important in the case of this hack to override the loop earlier than initializing the websocket client, otherwise it will use the original event loop.

[meejah]

Oh, hrmmmmm, txaio master’s make_batched_timer doesn’t take a loop= kwarg…

If you’re reading this because you ran into it: advice is to set your custom loop via txaio.config.loop = my_loop before you do any connections or listens and NOT pass loop= to the websocket factory

[meejah]

(BTW, thanks for tracking this down and for a clear bug report!)