Design Patterns with Ruby on Rails part 2: Query Object

This post is the second part of a series of posts about design patterns with Ruby on Rails.
See other parts here:
Part 1

Query Object

The query object is a pattern that helps in decomposing your fat ActiveRecord models and keeping your code slim and readable by extracting complex SQL queries or scopes into the separated classes that are easy to reuse and test.

Naming Convention

Usually located in app/queries directory, and of course, can be organized into multiple namespaces.
Typical file name has _query suffix, class name has Query suffix.
e.g.:
Posts::ActivePostsQuery

module Posts
  class ActivePostsQuery
    def self.call
      Post.where(status: 'active', deleted_at: nil)
    end
  end
end

Class methods are more practical and easier to stub.

There is a way to divide this query into pieces and become even more reusable? Yes, chaining methods.

class PostsQuery
  def initialize(posts = Post.all)
    @posts = posts
  end

  def active
    @posts.where(active: true, pending: false)
  end

  def pending
    @posts.where(pending: true, active: false)
  end

  def deleted
    @posts.with_deleted
  end
end

This way:

query = PostsQuery.new
query.deleted.pending

Query objects and model scopes

I have the following scope defined in Post class:

class Post < ActiveRecord::Base
  scope :active, -> {
    where(section: ['web', 'mobile'], status: 'active', deleted_at: nil)
  }
end

and you want to extract this query and keep this behavior with Post.active, how can you do this?

The solution may seen obvious:

class Post < ActiveRecord::Base
  def self.active
    ActivePostsQuery.call
  end
end

but it adds more code and it is no longer visible in the scope definition.

Using a scope with query objects

Don't forget that your query object has to return a relation.

class Post < ActiveRecord::Base
  scope :active, ActivePostsQuery
end

Let's design our query object class:

class ActivePostsQuery
  class << self
    delegate :call, to: :new
  end

  def initialize(relation = Post)
    @relation = relation
  end

  def call
    @relation.where(status: 'active', deleted_at: nil)
  end
end

Now the query still available via Post.active scope.

Refactoring

Let's take a look at how to use query objects in practice when refactoring larger queries.

class PostsController < ApplicationController
  def index
    @posts = Post.where(active: true, deleted_at: nil)
      .joins(:authors).where(emails: { active: true })
  end
end

When is time to write tests for this action it won't be possible without consulting the database.
It gets easier after extracting this query to a separated class:

class ActivePostsWithAuthorQuery
  attr_reader :relation

  def self.call(relation = Post)
    new(relation).call
  end

  def new(relation = Post)
    @relation = relation
  end

  def call
    active.with_author
  end

  def active
    relation.where(active: true, deleted_at: nil)
  end

  def with_author
    relation.joins(:authors).where(emails: { active: true })
  end
end

and the controller can be now updated:

class PostsController < ApplicationController
  def index
    @posts = ActivePostsWithAuthorQuery.call
  end
end

Testing

The queries it's not a controller concern anymore, it simplifies your test, now you can simply call allow(ActivePostsWithAuthorQuery).to receive(:call).and_return(...) in your spec.

Conclusion

There is always a trade-off and moving to a separated class is not always a good idea, it adds a complexity layer to your code and has a maintainability cost, use this pattern with caution, and enjoy your extendable code.