ezmobius / redactor Goto Github PK

And Agent/Actor based system written around the Redis key value store. Useful for coordination, command and control of distributed ruby processes

# simple getting started example

require 'redactor'

class Foo < RedActor::Actor

  mailbox :foo
  mailbox :bar

  def receive_foo(msg)
    p [:receive_foo, msg]
    send_msg 'bar', msg
  end

  def receive_bar(msg)
    p [:receive_bar, msg]
  end

end

RedActor.run

Currently for each mailbox you declare we spin up a thread that does a 10 second timeout BLPOP in a loop against a redis list named after the mailbox. For each message in the mailbox we will instantiate on of your Foo actor objects and pass in the redis connection in the initialize method so that you can access @redis to talk to redis directly or use send_msg(mailbox, message) to send a message along to another actor. When you call mailbox :foo you need to define an instance method Foo#receive_foo(msg). Using threads like this is fairly cheap since they are green threads and they are spending most of their time either asleep blocking for data from redis or they are running your code processing messages in a thread safe way. That said threads will be removed once redis gets a var-args blpop that can block on multiple keys.

TODO:

create actor registry where actors can register their capabilities to
allow for discovery of actors you want to subscribe to or send messages 
to.

create a scoring system for allowing a pool of actors working off the same 
mailbox and using the score to select where to route work

create a convention for identifying and passing work and data. use a  
system where the publisher places JSON encoded data in a key:
  # sending a mesage passing data on redis
  token = rand_token()
  @redis[token] = data.to_json
  msg = "feed:#{token}"
  send_msg :foo, msg

  # on the recieving actor side:
  mailbox :foo
  def receive_foo(msg)
    type, data = msg.split(':')
    data = JSON.parse(@redis[data])
    case type
    when 'feed'
      process_feed(data)
    when 'resize'
      resize_image(data)
    end
  end