Refactor build_tor_connection about txtorcon

[meejah]

One thing to bear in bind is APAF and OONI are using this method, AFAIK, so we should keep backwards compatibility ideally for their use-cases (the above looks fine for that).

Also instead of isinstance() it's probably more consistent with Twisted to check if the object implements the correct interface, like "if zope.interface.implements(connection, IStreamClientEndpoint): IStreamClientEndpoint(connection).connect(TorProtocolFactory())". I just made that syntax up for the idea :)

[meejah]

...but yes, I like this! Will make the examples more concise.

[lukaslueg]

I've created a new branch buildtorconnection that changes build_tor_connection() and adds a build_local_tor_connection(). I've changed examples/attach_stream_by_country to use the new function (most of the changes in the example are again pep8 and one bug :-)).

The documentation in torstate was not changed yet.

Should the APAF- and OONI-people have a brief look on that? All the unittests continue to work unchanged and as long as build_tor_connection is not used with keyword-arguments, they are fine.

[lukaslueg]

Yes, it's actually checking IStreamClientEndpoint.providedBy() which should be fine.

Also notice that build_local_tor_connection() can be supplied with a password-argument and such which pass through **kwargs.

[lukaslueg]

Should built_tor_connection and build_local_tor_connection return a reference to the endpoint that might have been created?

endpoint, d = build_local_tor_connection()
# and
_, d = build_tor_connection(endpoint)
endpoint, d = build_tor_connection(('localhost', 9051))

[meejah]

I'd say probably not; any code that wants this can get it from "protocol.transport.addr" (or state.protocol.transport.addr)

[lukaslueg]

closed?

[meejah]

Thanks for the patch, and closing these.