Ripple on Rails

May 22, 2012

Introduction

Motivation?? Checkout the RailsCasts episode on Mongoid

Started with the Seven DB in Seven weeks book. Wanted to see how Riak would mix with Rails / Ruby.

What is Riak

Riak is a distributed key-value store based on the Amazon Dynamo papers. Written in Erlang by the people at Basho, it’s an exciting addition to the crowded NoSQL zoo.

• Map Reduce
• eventually consistent
• read / write balancing

Installing Riak

Installing Riak is pretty straight forward for development, if you’re on a Mac use the excellent Homebrew package and you’ll get the latest 1.2 release.

brew install riak

Similarly on Linux it should be a matter of invoking your local package manager and you should be set. I’m assuming you’re using a Mac here, cause that’s what I’ve got but Linux should work as well. And Windows, well use a VM :-)

Ruby Environment

You’ll need some form of Ruby installed, I’d recommend installing via rvm over using your local ruby version. See rvm.io for details, it should simply be a matter of running a command in your terminal.

Next we’ll create a gemset and install rails into that.

rvm --create use 1.9.3@ripple
gem install rails

Creating a new Rails Application with Ripple

Now that we have Riak installed we’ll create a new Rails 3 application that uses Riak. Continuing in the Rails tradition, lets create a blogging app.

$ rails new blog --skip-active-record --skip-bundle -T

Now the first thing to do is add the Ripple gem to the Gemfile. Currently the latest version doesn’t work with latest rails so lets use the latest from git.

source 'http://rubygems.org'
gem 'rails', '3.2.2'

gem 'ripple', '1.0.0.beta'

group :development, :test do
    gem 'rspec'
    gem 'rspec-rails'
end

Then install the gems in the usual way.

 $ bundle

Once the gems have installed we’ll need to generate the Ripple configuration.

$ rails g
...
Ripple:
  ripple
  ripple:configuration
  ripple:js
  ripple:model
  ripple:observer
  ripple:test

So we want the default ripple one.

$ rails g ripple
create  config/ripple.yml
create  app/mapreduce
create  app/mapreduce/contrib.js
create  app/mapreduce/iso8601.js
create  app/mapreduce/ripple.js
insert  spec/spec_helper.rb
insert  spec/spec_helper.rb

As you can see it creates a ripple configuration file, some mapreduce assets and updates my spec_helper. Lets look just at the config/ripple.yml file for now.

The default config file is shown below. You may have to modify or add the ‘source:’ property to point to where you’ve installed Riak. I’ve put mine in the same directory as the rails project so it looks like this

# TODO use the default homebrew path here.
# Configure Riak connections for the Ripple library.
development:
    http_port: 9000
    pb_port: 9001
    host: 127.0.0.1
    source: /usr/local/bin   # Default for Homebrew.

# The test environment has additional keys for configuring the
# Riak::TestServer for your test/spec suite:
#
# * bin_dir specifies the path to the "riak" script that you use to
#           start Riak (just the directory)
# * js_source_dir specifies where your custom Javascript functions for
#           MapReduce should be loaded from. Usually app/mapreduce.
test:
    http_port: 9000
    pb_port: 9002
    host: 127.0.0.1
    source: /usr/local/bin   # Default for Homebrew.
    js_source_dir: <%= Rails.root + "app/mapreduce" %>

production:
    http_port: 8098
    pb_port: 8087
    host: 127.0.0.1
    source: /usr/local/bin   # Default for Homebrew.

With everything in place we can begin to build our application. We’ll start by generating an Article model with name and content fields and use the Rails scaffolding to create all the dependent bits.

$ rails g scaffold article name:string content:text

Ripple provides generators for models so that Ripple’s model generator is invoked when a model is created and ActiveRecord isn’t used. When we open up the model file we’ll see that it’s a simple class that includes Ripple::Document.

class Article
    include Ripple::Document
    property :name, String
    property :content, String
end

The big difference to an ActiveModel model is that the class defines each property, along with the type.

Ripple also adds a number of helpful rake tasks for Riak.

$ rake -T
...
rake riak:create        # Creates a Riak cluster for the current environment in db/, e.g. d...
rake riak:create:all    # Creates Riak clusters for all environments defined in config/ripp...
rake riak:destroy       # Destroys the generated Riak cluster for the current environment.
rake riak:drop          # Drops data only from the Riak cluster for the current environment.
rake riak:drop:all      # Drops data Riak clusters for all environments defined in config/r...
rake riak:start         # Starts the Riak cluster for the current environment
rake riak:stop          # Stops the Riak cluster for the current environment

Load the controller and show some actions

Make the controller run some basic rspec things

Show some basic validations on the data.

Point to Part2

Running MapReduce queries against Riak

Pulling in Postgres

Writing queries in CoffeeScript

Using the rails console

Using the Riak Console

What else does RailsCasts Mongoid cover?