Le 6 août 2013 à 08:28, Glyph <gl...@twistedmatrix.com> a écrit :
> > On Aug 5, 2013, at 2:09 AM, Benjamin BERTRAND <bee...@gmail.com> wrote: > >> That's why the inlineCallbacks solves the problem. When I yield on the >> self.server.dataReceived, I wait for the deferred in the publish methods to >> fire and everything is clean when I close the connection. >> >> Is there another way to do that? > > Jonathan Lange wrote about this a long time ago: > > http://blackjml.livejournal.com/23029.html Thanks for the link. I added a deferred in my RedisPublishService class and fire it at the end of my publish method so that I can wait for it before to close the connection. > > but better than doing that would be to avoid setting up and tearing down real > TCP connections during your tests. Those are slower than an in-memory thing > (like you would get from twisted.test.proto_helpers, the one and only public > module in our 'test' package) and also, as you have discovered, more > unpredictable. > > Is there any way to hand txredis a synthetic transport for testing? If not, > file a bug on it :). I guess it's possible to use a synthetic transport, but that wouldn't help much without a "fake" redis server. That would indeed be nice to have for testing. I actually just found a dummy txredis client that implements everything internally for testing: https://github.com/williamsjj/dummy-txredis Might be worth looking at. But I wonder if implementing a basic redis server with twisted (for testing purpose) wouldn't be nicer? Benjamin > > -glyph > > _______________________________________________ > Twisted-Python mailing list > Twisted-Python@twistedmatrix.com > http://twistedmatrix.com/cgi-bin/mailman/listinfo/twisted-python
_______________________________________________ Twisted-Python mailing list Twisted-Python@twistedmatrix.com http://twistedmatrix.com/cgi-bin/mailman/listinfo/twisted-python
