Hubble is a lightweight, extensible ruby based Pubsubhubbub (PuSH) hub.
In addition to conforming to the PuSH spec 0.3, a real life hub should have in its concerns how it is going to store topics, subscribers and meta-data about feeds.
Additionally it should decide how it would push notifications to subscribers, as well as diff new and old feeds.
Therefore, in order to allow more flexibility, in its core Hubble is agnostic to: storage, async execution and feed diffing algorithms.
But you don't get an empty box!. Hubble includes a default implementation of:
- Storage: redis
- Async execution: Resque
- Diffing: a custom simple feed diff (based on last seen entry id)
Hubble works with a redis profile out of the box. You will need a
redis-server running.
Running Hubble:
$ git clone https://github.com/jondot/hubble.git
$ cd hubble; bundle install
$ rackup
Running Hubble queue worker (for more, see Resque):
$ rake resque:work QUEUE=job_runner
You can now run an example loopback application. The loopback application can subscribe to the hub - as a subscriber, publish to the hub as a publisher, and show you what the hub pings it with.
$ cd examples; bundle install
$ rackup
Now try this:
assume hub runs at localhost:3000, loopback at localhost:9292, worker
is running as mentioned above
-
publish a topic - GET http://localhost:9292/publish?hub=http://localhost:3000
-
subscribe - GET http://localhost:9292/subscribe?hub=http://localhost:3000
(notice hub verifies in the background) -
now publish again - GET http://localhost:9292/publish?hub=http://localhost:3000
(notice we get a ping from hub with diff)
In order to support a notion of plug-and-play backends, Hubble introduces a 'profile'. We use a redis profile by default.
A profile is nothing more than a class which conforms to the interface
defined in the docs (see lib/hubble/profiles/redis_profile.rb) and
includes Hubble::Profile.
A profile also includes a configurable connection.
A diff algorithm is nothing more than a class which conforms to the following interface:
def diff(old_meta, new_content, content_type=nil)
where old_meta is any old content form that can help you extract the
new content from new_content.
content_type is there to help you toggle algorithms upon the content's type - if you decide you want to be aware of
that (i.e. parsing rss vs atom).
For your convenience,
- I've set up a watchr script for continous testing:
$ bundle exec watchr specs.watchr - Coverage by simplecov, run on demand:
$ rake spec COVERAGE=true
Copyright (c) 2011 Dotan Nahum @jondot. See LICENSE.txt for further details.