__ _          _
 / _(_)        | |
| |_ _ _ __ ___| |__   ___  ___  ___
|  _| | '__/ _ \ '_ \ / _ \/ __|/ _ \
| | | | | |  __/ | | | (_) \__ \  __/
|_| |_|_|  \___|_| |_|\___/|___/\___|

Build realtime web applications in Ruby and JS

Build Status Code Climate Test Coverage

What is Firehose?

Firehose is both a Rack application and JavaScript library that makes building real-time web applications possible.

Getting Started

First, you'll need to install and run Redis 2.6. Version 2.6 is required because Firehose uses Lua/EVAL for its transactions, which is not available in earlier versions of Redis.

Then install the gem.

$ gem install firehose

The Server

Now fire up the server.

$ firehose server
>> Thin web server (v1.3.1 codename Triple Espresso)
>> Maximum connections set to 1024
>> Listening on 127.0.0.1:7474, CTRL+C to stop

In case you're wondering, the Firehose application server runs the Rack app Firehose::Rack::App.new inside of Thin or Rainbows! Firehose::Rack::App consists of a bunch of smaller apps and a middleware, which is useful for hacking.

Publish a message to a bunch of subscribers

Lets test it out! Open two terminal windows. In one window, curl:

$ curl "http://localhost:7474/hello"

Then run the following in the other terminal:

$ curl -X PUT -d "Greetings fellow human being..." "http://localhost:7474/hello"

and you should see the message in the other terminal.

Greetings fellow human being...

Yeah, so?

You have a dirt simple HTTP pub-sub feed. You could setup an after_commit hook on ActiveRecord to push JSON to an end-point. On the other side, you could have a Backbone.js application that picks up the changes and updates the client-side UI.

Holy mackerel! Its a nice, clean, RESTful way to build real-time web applications.

The JavaScript Consumer

Firehose doesn't just stop at curl; it has a full-featured JavaScript client that lets you subscribe to channels for live updates.

Still have the server running? Copy and paste the code below into Firebug or the WebKit console.


new Firehose.Consumer({
  message: function(msg){
    console.log(msg);
  },
  connected: function(){
    console.log("Great Scotts!! We're connected!");
  },
  disconnected: function(){
    console.log("Well shucks, we're not connected anymore");
  },
  error: function(){
    console.log("Well then, something went horribly wrong.");
  },
  // Note that we do NOT specify a protocol here because we don't
  // know that yet.
  uri: '//localhost:7474/hello'
}).connect();

There's also a Consumer that uses channel multiplexing. The multiplexed consumer is useful for scenarios where you want to subscribe to messages from many channels at once, without having to use one connection per channel. You can specify a list of channels to subscribe to, including a handler function per channel that gets called with all messages coming from that channel.

Example:

new Firehose.MultiplexedConsumer({
  connected: function(){
    console.log("Great Scotts!! We're connected!");
  },
  disconnected: function(){
    console.log("Well shucks, we're not connected anymore");
  },
  error: function(){
    console.log("Well then, something went horribly wrong.");
  },
  // Note that we don't specify a general message handler function
  // but instead define one per channel below

  // Note that we do NOT specify a protocol here because we don't
  // know that yet. We also don't specify a specific channel name as part of
  // the URI but instead pass in a list of subscriptions below
  uri: '//localhost:7474/',

  // List of channel subscriptions:
  channels: {
    "/my/channel/1": {
      last_sequence: 10,        // defaults to 0 and can be ommitted
      message: function(msg) {
        console.log("got message on channel 1:");
        console.log(msg);
      }
    },
    "/my/channel/2": {
      message: function(msg) {
        console.log("got message on channel 2:");
        console.log(msg);
      }
    }
  }
}).connect();

Then publish another message.

$ curl -X PUT -d "\"This is almost magical\"" "http://localhost:7474/hello"

How is it different from socket.io?

socket.io attempts to store connection state per node instance. Firehose makes no attempt to store connection state.

Also, socket.io attempts to abstract a low-latency full-duplex port. Firehose assumes that its impossible to simulate this in older web browsers that don't support WebSockets. As such, Firehose focuses on low-latency server-to-client connections and encourages the use of existing HTTP transports, like POST and PUT, for client-to-server communications.

The Ruby Publisher

While you can certainly make your own PUT requests when publishing messages, Firehose includes a Ruby client for easy publishing.

require 'firehose'
require 'json'
json = {'hello'=> 'world'}.to_json
firehose = Firehose::Client::Producer::Http.new('//127.0.0.1:7474')
firehose.publish(json).to("/my/messages/path")

Configuration

Firehose can be configured via environmental variables. Take a look at the .env.sample file for more info.

Rack Configuration

There are two rack applications that are included with Firehose: Firehose::Rack::Producer which a client can PUT HTTP request with message payloads to publish information on Firehose and the Firehose::Rack::Consumer application which a client connects to via HTTP long polling or WebSockets to consume a message.

Consumer Configuration

# Kitchen-sink rack configuration file example
require 'firehose'

consumer = Firehose::Rack::Consumer.new do |app|
  # Configure how long the server should wait before send the client a 204
  # with a request to reconnect. Typically browsers time-out the client connection
  # after 30 seconds, so we set the `Firehose.Consumer` JS client to 25, and the
  # server to 20 to make sure latency or timing doesn't cause any problems.
  app.http_long_poll.timeout = 20
end

run consumer

Publisher Configuration

# Kitchen-sink rack configuration file example
require 'firehose'

# There's nothing to configure with the Publisher, but its possible that
# you might include rack middleware authorization mechanisms here to control
# who can publish to Firehose.

run Firehose::Rack::Publisher.new

Sprockets

Using Sprockets is the recommended method of including the included client-side assets in a web page.

  1. Add the firehose gem in your app's Gemfile.

  2. Append the firehose gem's assets to the sprockets path. In a Rails app, this is usually done in an initializer.

# Add firehose to a custom sprockets configuration.
my_sprockets_env = Sprockets::Environment.new
Firehose::Assets::Sprockets.configure my_sprockets_env
  1. Require your config file and the firehose gem. This would look something like this:
#= require some/other/js/file
#= require lib/firehose_config
#= require firehose
#= require some/more/js/files

It is important that your firehose config file comes first.

Not using sprockets?

If you don't intend to use the Firehose JavaScript client in a Ruby stack where Sprockets is available, you can grab the unminified source by running:

$ firehose javascript > firehose.js

Copy the firehose.js where needed in your project.

Web Server

Firehose currently supports Thin and Rainbows! (which is the default). Neither is listed as a dependency in the gemspec so that you don't need to install whichever one you aren't using. You can set which server to use via the .env file (recommended) or with the -s option to bin/firehose.

Exception Notification

If you'd like to be notified of exceptions, add something like this in your custom config.ru file.

# Use exceptional to handle anything missed by Rack::Exceptional
if exceptional_key = ENV['EXCEPTIONAL_KEY']
  require 'exceptional'
  EM.error_handler do |e|
    Firehose.logger.error "Unhandled exception: #{e.class} #{e.message}\n#{e.backtrace.join "\n"}"
    ::Exceptional.handle(e)
  end
end

Deployment

The recommended method ofGdeploying firehose is to deploy it separately from your main app.

  1. Create a new project with a Gemfile such as
gem "firehose"
gem "airbrake"
gem "rainbows",   :require => false
gem "rack", "~> 1.4.0" # if you're using Rainbows. See https://github.com/firehoseio/firehose/commit/dfe55fff
gem "foreman",    :require => false
gem "capistrano", :require => false

Of course, you could use exceptional instead of airbrake and thin instead of rainbows.

  1. Set up config/deploy.rb to your liking. You can follow most directions for using Capistrano and Foreman to deploy Rack apps, such as https://gist.github.com/1027117

  2. Set up config/rainbows.rb (if you are using Rainbows!). The gem includes example configurations scripts to get you started. There's also an example at https://gist.github.com/bradgessler/f2416efdbb1771e983b3.

New releases & version bump

For a new release of Firehose, bump the version number in lib/firehose/version.rb as well as package.json. Make sure, they have the same version number.

Support

Ruby version

Firehose will support the latest minor 2.x revisions of Ruby that are officially supported by the Ruby community. More details at https://www.ruby-lang.org/.