Goliath and Redis

I’ve been looking into using non-blocking web frameworks lately, and I started out with the poster-child for this movement; Node.js. While Node is very nice (especially when combined with CoffeeScript), it has some rough edges.

I currently prefer to use Ruby on the server side, and to keep my CoffeeScripts more towards the front end. (For example with Backbone.js)

Background

There are quite a few non-blocking Ruby web frameworks out there. I’ve looked at three of them; Sinatra::Synchrony, Goliath and Cramp.

I REALLY like the idea behind Sinatra::Synchrony, but it doesn’t seem to be as “battle hardened” as Goliath. It also has some issues with Ruby 1.9.3.

I’ve been thinking about using Goliath ever since I saw the presentation 0-60 with Goliath: Building High Performance Ruby Web-Services (video) at Øredev 2011.

Dependencies

Goliath, obviously.

Version 1.0 has been released, and I have taken the time to update this article with the relevant changes. One of the more notable changes are the removal of the built in router. I think that was a good decision. Gone are the days when you needed to clone Goliath from GitHub just to get a sane version.

I am also going to use the Ruby client library for Redis, which has built in support for EM-Synchrony.

Gemfile

source :rubygems

gem "goliath", "~> 1.0"
gem "hiredis", "~> 0.4"
gem "redis",   "~> 3.0",
    :require => ["redis/connection/synchrony", "redis"]

group :test do
  gem "em-http-request", "~> 1.0"
  gem "mock_redis", "~> 0.4"
end

Testing

You didn’t think I would skip on the testing, right?

My personal favorite among the myriad of Ruby test frameworks is minitest (The default test framework in Ruby 1.9) and that is what I’m going to use in this article.

spec/spec_helper.rb

require 'bundler'
Bundler.require

require 'minitest/spec'
require 'minitest/pride'
require 'minitest/autorun'
require 'goliath/test_helper'
require 'mock_redis'

$redis = MockRedis.new

class Goliath::Server
  def load_config(file = nil)
    config['redis'] = $redis
  end
end

spec/api_spec.rb

I hope that writing specs for Goliath will become less verbose in the future, currently it seems like I need to wrap each request in a with_api block.

require_relative 'spec_helper'
require_relative '../api'

describe Api do
  include Goliath::TestHelper

  it "responds to heartbeat" do
    with_api Api do
      get_request path: '/' do |api|
        api.response.must_equal 'OK'
      end
    end
  end

  it "can set and retrieve data" do
    with_api Api do
      get_request path: '/bar' do |api|
        api.response.must_equal ''
      end
    end

    with_api Api do
      put_request path: '/bar?value=foo' do |api|
        api.response.must_equal 'OK'
      end
    end

    with_api Api do
      get_request path: '/bar' do |api|
        api.response.must_equal 'foo'
      end
    end
  end
end

The application

This is a very simple API, it can only do three things:

Respond with “OK” and status code 200 GET /
Add new data PUT /foo?value=bar
Retrieve a value GET /foo

api.rb

class Api < Goliath::API
  use Goliath::Rack::Params
  use Goliath::Rack::Heartbeat, path: '/'

  def response(env)
    key = "example_api:#{env['REQUEST_PATH']}"
    val = params['value'].to_a

    res = case env['REQUEST_METHOD']
      when 'GET' then redis.get key
      when 'PUT' then redis.set key, val
    end

    [200, {}, res]
  end
end

Starting the application

First you need to have a Redis server running.

The final two pieces of the puzzle is a config.rb that contains the connection to Redis via a EM::Synchrony::ConnectionPool and a Goliath::Runner that takes care of running the application.

config.rb

config['redis'] = EM::Synchrony::ConnectionPool.new size: 2 do
  Redis.new
end

runner.rb

require "bundler"
Bundler.require

require 'goliath/runner'
require_relative 'api'

runner = Goliath::Runner.new(ARGV, nil)
runner.api = Api.new
runner.app = Goliath::Rack::Builder.build(Api, runner.api)
runner.run

Run ruby runner.rb -s -e prod -c config.rb and a server should spin up in production mode on port 9000.

Using the application

I like to use cURL when manually testing API’s:

Adding a new key/value pair

curl -X PUT http://0.0.0.0:9000/foo -d "value=bar"

Retrieve a stored value

curl -X GET http://0.0.0.0:9000/foo

code7 interactive