= Rosetta Queue

Rosetta Queue is a messaging gateway API with adapters for many messaging systems available in Ruby. Messaging systems can be easily switched out with a small configuration change. Code for testing on the object and application level is also provided.

The adapters provided currently are for stomp, amqp, and beanstalk.  We would like to add adapters for other messaging gateways.  The stomp adapter has been used in production along side with Apache's ActiveMQ.  The amqp adapter currently works along side RabbitMQ and passes the acceptance tests but as of yet has not been used in production.


== Quick Tutorial

Note: The API will most likely change until we reach 1.0.  We will be moving to a more concise API (i.e. queue(:test_queue) << message, etc...)  We will also be changing how exceptions are handled.

When using Rosetta Queue in an application you will need to configure the queues, adapters, and filters (optional).  These configurations should be placed in a file that gets loaded once when your program starts.  If you are using Rails then a good place for this is config/initializers/rosetta_queue.rb.


To set up destinations for producing and consuming messages:

RosettaQueue::Destinations.define do |queue|
  queue.map :test_queue, '/queue/my_test_queue'
end


Defining your adapter:

  RosettaQueue::Adapter.define do |a|
    a.user = ""
    a.password = ""
    a.host = "localhost"
    a.port = 61613
    a.type = "stomp"
  end


Define a logger for Rosetta Queue to use.  The logger should be a standard ruby logger:

  RosettaQueue.logger = Logger.new('/my_project/rosetta_queue.log')


You can optionally set up filters that are applied to all messages that are sent and received.  For example, if you want to use hashes as messages and serialize them as JSON the following filters (along with ActiveSupport) would accomplish this:

  RosettaQueue::Filters.define do |filter_for|
    filter_for.receiving { |message| ActiveSupport::JSON.decode(message) }
    filter_for.sending { |hash| hash.to_json }
  end


To publish a message:

  message = {"hello" => "world!"}  # Assuming you have a filter setup
  RosettaQueue::Producer.publish(:test_queue, message)

When consuming messages from a queue you will generally want to create a consumer to handle the messages:

  class TestConsumer
    include RosettaQueue::MessageHandler

    subscribes_to :vendor_status
    options :persistent => true

    def on_message(message)
      puts "We consumed a message: #{message.inspect}"
    end

  end

To fire the consumers up you will want to run a separate process create a manager with all of your consumers.

  require 'rosetta_queue'
  require 'rosetta_queue/consumer_managers/threaded'
  require 'test_consumer'

  RosettaQueue::ThreadedManager.create do |m|
    m.add(TestConsumer.new)
    m.start
  end


It is recommended that you set your adapter to the 'null' adapter for your specs and then include RosettaQueue::Matchers in any example group you are needing to specify behaviour with RosettaQueue.  The matchers currently switch out the null adapter for the fake adapter to verify the behaviour.  All the matchers for the unit tests are lambda based, like so:

  lambda { model.save }.should publish("foo", :to => :test_queue).

Please look at the publishing matchers for more information.  For examples on how to write acceptance tests for your Rosetta Queue's code please see RosettaQueue's own Cucumber features and read this {blog post}[http://www.benmabey.com/2009/02/17/using-cucumber-to-integrate-distributed-systems-and-test-messaging/].


=== AMQP

The AMQP adapter is based on the synchronous AMQP ruby client "Bunny" by celldee [http://github.com/celldee/bunny].  The original AMQP adapter for RosettaQueue was the 'evented' client by Aman Gupta [http://github.com/tmm1/amqp], but the evented model posed some problems for our integration tests.  That, and the desire to keep things simple led us to use a synchronous AMQP client.

==== Configuration

If you are using the AMQP adapter, there are some important rules when defining the adapter and mapping your destinations.  Two of the basic building blocks of AMQP are queues and exchanges.  Queues bind to exchanges in several different ways.  A can queue bind to an exchange and requests messages that match a specific routing key, which is called 'direct exchange' and is analogous to the 'point-to-point' messaging pattern.  Alternately, queues can bind to an exchange to receive all messages sent to that exchange; this is called 'fanout exchange' and is analogous to the 'publish-subscribe' pattern.  So, when you want to define a simple 'direct-exchange' queue, your queue name must begin with the term "queue".

   RosettaQueue::Destinations.define do |queue|
     queue.map :test_queue, 'queue.my_test_queue'
   end

And, when you want to define a queue that will bind to a 'fanout-exchange', your queue name must begin with the term "fanout".

   RosettaQueue::Destinations.define do |queue|
     queue.map :test_queue, 'fanout.my_test_queue'
   end

Oh, and when defining your adapter, be sure to specify the AMQP adapter as :amqp_synch, not :amqp.  This way, if we do get around to adding Aman's evented AMQP client, we can distinguish between the synchronous and asynchronous clients (i.e., :amqp_sync vs. :amqp_asynch).


== How to contribute
----------------------------------------------------------------
Gems you will need:
cucumber, rspec, yaml, stomp, bunny, beanstalk-client

You should be able to run the rspec code examples (specs) without any brokers running with autospec or 'rake spec'.

To run the cucumber features you will need the a messaging system setup that can speak stomp and AMQP.  We have been using the following brokers in testing:

=== Apache ActiveMQ (for the stomp adapter)
Go to http://activemq.apache.org/download.html to download the latest version, and see http://activemq.apache.org/getting-started.html for installation and configuration instructions.
The stomp client and features should work with ApacheMQ out of the box.  If you are running any ApacheMQ servers in production on your network you will want to disable the multicast autodiscovery in conf/activemq.xml.  (Around lines 56 and 98.)

=== RabbitMQ (for the amqp adapter)
Download the right tar from here: http://www.rabbitmq.com/download.html and follow the installation directions that comes with it.
The features rely on a user of 'rosetta' being setup with the password of 'password' and added to the default virtualhost. You can set them up like so:

  rabbitmqctl add_user rosetta password
  rabbitmqctl map_user_vhost rosetta /

=== Beanstalk (for the beanstalk adapter) ===

You should set up and run a local instance of beanstalk for your
tests.

==== beanstalkd ====

Mac OSX users can install beanstalkd from ports. Linux users should
check their package manager of choice. This is probably the optimal
method.

Diehards can clone git://github.com/kr/beanstalkd and build it from
source. It is nontrivial.

Once installed, you can launch beanstalk by running

  beanstalkd -p 11300

This will make it listen on localhost on the standard port 11300,
which the specs use. You might also want to daemonize beanstalk with
the -d option but for testing this is not recommended, as the queue
can get stuck with old messages, which will break your specs.

==== beanstalk-client ====

You will also need the Ruby beanstalk-client gem.

  sudo gem install beanstalk-client