Pull TAsyncProcessor interface out of TBaseAsyncProcessor

[m50d]

As briefly mentioned on IRC (I'm lmm). This makes it possible to use e.g. TNonblockingServer with a custom async processor that doesn't extend TBaseAsyncProcessor.

(Use case: I have an implementation in scrooge that generates such async processors, for use in scala).

Async processors still need to "implement" TProcessor so that they can be configured as a processor (the implementation can be simply return false as it is in TBaseAsyncProcessor); this is somewhat awkward but a change to make that unnecessary would be much more intrusive.

[hcorg]

please create issue on Jira and paste link to it here - that'll connect it to this requests

see: http://thrift.apache.org/docs/HowToContribute for details :)

BTW: couldn't TAsyncProcessor extend TProcessor?
public class TBaseAsyncProcessor implements TAsyncProcessor, TProcessor
Processor word seems redundant ;)

[m50d]

It could, but ultimately they should become independent (or rather, there should be a superinterface that represents "either of these" and I guess a visitor rather than the instanceof check in AbstractNonblockingServer). TBaseAsyncProcessor and similar should not really implement the synchronous form of process() and should not really implement TProcessor at all (neither directly nor by dint of TAsyncProcessor extending it), but as I said in the initial pull request, that would be a more intrusive change; that said I'm happy to do that if you like.

[m50d]

Jira issue created: https://issues.apache.org/jira/browse/THRIFT-2804

[hcorg]

yep, visitor would be best :) If you ever feel like doing such change don't hesitate ;)

I'll try to merge this request later today, thanks for Jira issue!

[bsideup]

👍