Curator

See Untangle Domain and Persistence Logic with Curator for the announcement blog post.

Curator is a model and repository framework for Ruby. It's an alternative to ActiveRecord-like libraries where models are tightly coupled to persistence. Curator allows you to write domain object that are persistence free, and then write repositories that persist these objects. These ideas are largely taken from the Repository section of Domain Driven Design.

Currently, curator supports Riak, MongoDB and an in-memory data store for persistence. If you are interested in enhancing curator to support other data stores, please let us know.

Domain objects should include the Curator::Model module:

class Note
  include Curator::Model
  attr_accessor :id, :title, :description, :user_id
end

These models can be intiatiated with hashes and used just like regular ruby objects:

note = Note.new(:title => "My Note", :description => "My description")
puts note.description

Repositories should include the Curator::Repository module:

class NoteRepository
  include Curator::Repository
  indexed_fields :user_id
end

Repositories have save, find_by_id, and find_by and find_first_by methods for indexed fields. find_by methods return an array of all matching records, while find_first_by only returns the first match (with no ordering):

note = Note.new(:user_id => "my_user")
NoteRepository.save(note)

note1 = NoteRepository.find_by_id(note.id)
note2 = NoteRepository.find_first_by_user_id("my_user")
my_notes = NoteRepository.find_by_user_id("my_user")

Fields included in indexed_fields are indexed in both Riak and MongoDB when persisted.

As persistence gets more complicated, repositories can implement their own serialize and deserialize methods to handle any case. For example, if our note contained a PDF, the repository might look like:

class NoteRepository
  include Curator::Repository

  indexed_fields :user_id

  def self.serialize(note)
    attributes = super(note)
    attributes[:pdf] = Base64.encode64(note.pdf) if note.pdf
    attributes
  end

  def self.deserialize(attributes)
    note = super(attributes)
    note.pdf = Base64.decode64(attributes[:pdf]) if attributes[:pdf]
    note
  end
end

See curator_rails_example for an example application using curator.

If you use curator within Rails, all you need is to add curator to your Gemfile and create a config/riak.yml with contents like:

development:
  :http_port: 8098
  :host: localhost
test:
  :http_port: 8098
  :host: localhost

We recommend putting your models in app/models and your repositories in app/repositories. If you do this, don't forget to add app/repositories to the list of autoload paths:

# config/application.rb

config.autoload_paths += %W(#{config.root}/app/repositories)

You can also use Rails form builder with curator models:

<%= form_for @note, :url => { :action => "create" } do |f| %>
  <dl>
    <dt><%= f.label :title %></dt>
    <dd><%= f.text_field :title %></dd>
    <dt><%= f.label :description %></dt>
    <dd><%= f.text_area :description, :size => "60x12" %></dd>
  </dl>
  <%= f.submit "Create" %>

If you are not using Rails, you can configure curator manually:

Curator.configure(:riak) do |config|
  config.bucket_prefix = "my_app"
  config.environment = "development"
  config.migrations_path = File.expand_path(File.dirname(__FILE__) + "/../db/migrate")
  config.riak_config_file = File.expand_path(File.dirname(__FILE__) + "/config/riak.yml")
end

Alternatively, instead of using a YAML file to configure the client you can set it manually.

Curator.configure(:riak) do |config|
  config.bucket_prefix = "my_app"
  config.environment = "development"
  config.migrations_path = File.expand_path(File.dirname(__FILE__) + "/../db/migrate")
  config.client = Riak::Client.new
end

If you are writing tests using curator, it's likely that you will want a way to clean up your data in Riak between tests. Riak does not provide an easy way to clear out all data, so curator takes care of it for you. You can use the following methods if you change your backend from :riak to :resettable_riak:

• remove_all_keys - remove everything in Riak under the current bucket_prefix and environment
• reset! - remove all keys since the last reset!

For example, our spec_helper.rb file looks like this for our rspec test suite:

Curator.configure(:resettable_riak) do |config|
  config.bucket_prefix = "curator"
  config.environment = "test"
  config.migrations_path = File.expand_path(File.dirname(__FILE__) + "/../db/migrate")
  config.riak_config_file = File.expand_path(File.dirname(__FILE__) + "/../config/riak.yml")
end

RSpec.configure do |config|
  config.before(:suite) do
    Curator.data_store.remove_all_keys
  end

  config.after(:each) do
    Curator.data_store.reset!
  end
end

This ensures that our tests start with an empty Riak, and the data gets removed in between tests.

See Data migrations for NoSQL with Curator for an overview of data migrations. They have also been implemented in the curator_rails_example.

Each model instance has an associated version that is persisted along with the object. By default, all instances start at version 0. You can change the default by specifying the current_version in the model class:

class Note
  current_version 1
end

note = Note.new
note.version #=> 1

When the repository reads from the data store, it compares the stored version number to all available migrations for that collection. If any migrations are found with a higher version number, the attributes for the instance are run through each migration in turn and then used to instantiate the object. This means that migrations are lazy, and objects will get migrated as they are used, rather than requiring downtime while all migrations run.

In order to write a migration, create a folder with the collection name under the migrations_path that was configured in the Curator.configure block.

mkdir db/migrate/notes/

Then, create a file with a filename that matches #{version}_#{class_name}.rb:

# db/migrate/notes/0001_update_description.rb

class UpdateDescription < Curator::Migration
  def migrate(attributes)
    attributes.merge(:description => attributes[:description].to_s + " -- Passed through migration 1")
  end
end

Now, all Note objects that are read with a version lower than 1 will have their description ammended. Migrations are free to do what they want with the attributes. They can add, edit or delete attributes in any combination desired. All that matters is that the resulting attributes hash will be used to instantiate the model.

Since migrations merely accept and return a hash, they are easy to unit test. They do not affect the data store directly (like ActiveRecord migrations), so there is no harm in calling them in tests:

require 'spec_helper'
require 'db/migrate/notes/0001_update_description'

describe UpdateDescription do
  describe "migrate" do
    it "appends to the description" do
      attributes = {:description => "blah"}
      UpdateDescription.new(1).migrate(attributes)[:description].should == "blah -- Passed through migration 1
    end
  end
end

Repositories allow their default collection properties to be configured through a syntax similar to Sinatra's configuration interface. Configurable properties are data store specific and currently Riak is the only supported default-overridable data store.

The methods used to configure the Repository are set, enable, and disable:

class NoteRepository
  include Curator::Repository

  set :n_val, 5
  enable :allow_mult
  disable :last_write_wins
end

For more information on the properties that can be configured for Riak buckets, see this documentation.

Repository configuration is not applied automatically. To apply the uncommitted properties, run the curator:repository:apply rake task:

$ rake curator:repository:apply
[Curator] Preparing to apply settings to all repositories...
 * Updating settings for NoteRepository... Done!
[Curator] Done!

To manually apply uncommitted properties to a specific repository, use NoteRepository.apply_settings! in an irb or Rails console session.

Curator stores objects in the data store using the id as the key. The value is a json representation of the instance_values of the object. Your repository can implement serialize/deserialize to get different behavior.

The bucket name in Riak is <bucket_prefix>:<environment>:<collection>. The bucket prefix is configurable. By default, it will either be curator or the name of the Rails application if you are using curator within Rails. The collection is derived from the name of the Repository class, and it can be overriden. For example, if you implement a NoteRepository, the riak bucket will be curator:development:notes in development mode, and curator:production:notes in production mode.

The collection name in MongoDB is derived from the name of the Repository class, and it can be overriden. For example, if you implement a NoteRepository, the collection name will be notes.

MongoDB will preserve the types of attributes. For example, if you set an attribute as a Time object, it will come back as a Time object. If you do not set a key for an object, MongoDB will generate one as a BSON::ObjectId, which will be returned. If you want to find_by_id, you will have to use the BSON::ObjectId class, not a String. On the other hand, if you specify keys as strings, you can look them back up as strings.

We appreciate contributions of any kind. If you have code to show us, open a pull request. If you found a bug, want a new feature, or just want to talk design before submitting a pull request, open an issue.

Please include tests with code contributions, and try to follow conventions that you find in the code.

Riak is required in order to run the curator specs. After installing Riak, change the backend to leveldb. For example, here is how to install on OS X using homebrew:

brew install riak
edit /usr/local/Cellar/riak/<riak version>/libexec/etc/riak.conf

  change
    storage_backend = bitcask
  to
    storage_backend = leveldb

riak start

Curator has a set of shared_examples for data store specs. Take a look at spec/curator/shared_data_store_specs.rb. These cover most of the data store functionality, so include these on your spec and make them pass:

require 'spec_helper'
require 'curator/shared_data_store_specs'

module Curator::SomeNewDB
  describe Curator::SomeNewDB::DataStore do
    include_examples "data_store", DataStore

    ... other specs specific to SomeNewDB ...
  end
end

Curator is released under the MIT license.