Re: [tor-dev] Relay Dashboard Code Review

[meejah <meejah@xxxxxxxxx>]

Sorry for the noise, I found the code ("develop" branch of
https://github.com/leivaburto/rwsd) and add a couple more comments:


1. The "WEBSOCKETS" global seems like it wants to be a class instead:
you've basically got a bunch of static functions (add_websocket, etc)
that manipulate a global variable (WEBSOCKETS) which will be hard to
mock, etc. Making this its own class that encapsulates the websockets
and contains methods ("init_websockets" becomes the constructor,
".add_websocket()" becomes a ".add" method etc) will make it easier to
mock and use. IMO. So, e.g., instead of calling init_websockets() in
run_server, you'd instantiate a new WebSocketTracker() or whatever and
pass it to your Application where your handlers can then get at it.

2. I didn't look at any threading very closely, but in general if you're
interacting with a threaded/syncrhonous library like Stem you're going
to want to do that via deferToThread()/callFromThread() and so forth to
avoid blocking the reactor for "too long" and to avoid calling reactor
methods from any thread that isn't the main one. Practically speaking,
with a 1:1 relay-to-local-Tor/Stem instance this might not really matter
very much (i.e. if there's always 0 or 1 web browser clients it probably
doesn't matter how long you pause the reactor).

Cheers,
meejah

_______________________________________________
tor-dev mailing list
tor-dev@xxxxxxxxxxxxxxxxxxxx
https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-dev