How to test a Controller Concern in Rails 4

Question

What is the best way to handle testing of concerns when used in Rails 4 controllers? Say I have a trivial concern Citations.

module Citations
    extend ActiveSupport::Concern
    def citations ; end
end

The expected behavior under test is that any controller which includes this concern would get this citations endpoint.

class ConversationController < ActionController::Base
    include Citations
end

Simple.

ConversationController.new.respond_to? :yelling #=> true

But what is the right way to test this concern in isolation?

class CitationConcernController < ActionController::Base
    include Citations
end

describe CitationConcernController, type: :controller do
    it 'should add the citations endpoint' do
        get :citations
        expect(response).to be_successful
    end
end

Unfortunately, this fails.

CitationConcernController
  should add the citations endpoint (FAILED - 1)

Failures:

  1) CitationConcernController should add the citations endpoint
     Failure/Error: get :citations
     ActionController::UrlGenerationError:
       No route matches {:controller=>"citation_concern", :action=>"citations"}
     # ./controller_concern_spec.rb:14:in `block (2 levels) in <top (required)>'

This is a contrived example. In my app, I get a different error.

RuntimeError:
  @routes is nil: make sure you set it in your test's setup method.

Answer 1

You will find many advice telling you to use shared examples and run them in the scope of your included controllers.

I personally find it over-killing and prefer to perform unit testing in isolation, then use integration testing to confirm the behavior of my controllers.

Method 1: without routing or response testing

Create a fake controller and test its methods:

describe MyControllerConcern do
  before do
    class FakesController < ApplicationController
      include MyControllerConcern
    end
  end

  after do
    Object.send :remove_const, :FakesController 
  end

  let(:object) { FakesController.new }

  it 'my_method_to_test' do
    expect(object).to eq('expected result')
  end

end

Method 2: testing response

When your concern contains routing or you need to test for response, rendering etc... you need to run your test with an anonymous controller. This allow you to gain access to all controller-related rspec methods and helpers:

describe MyControllerConcern, type: :controller do
  controller(ApplicationController) do
    include MyControllerConcern

    def fake_action; redirect_to '/an_url'; end
  end

  before do
    routes.draw {
      get 'fake_action' => 'anonymous#fake_action'
    }
  end
    
  describe 'my_method_to_test' do
    before do
      get :fake_action 
    end

    it do
      expect(response).to redirect_to('/an_url') 
    end
  end
end

As you can see, we define the anonymous controller with controller(ApplicationController). If your test concerne another class than ApplicationController, you will need to adapt this.

Also for this to work properly you must configure the following in your spec_helper.rb file:

config.infer_base_class_for_anonymous_controllers = true

Note: keep testing that your concern is included

It is also important to test that your concern class is included in your target classes, one line suffice:

describe SomeTargetedController do
  it 'includes MyControllerConcern' do
    expect(SomeTargetedController.ancestors.include? MyControllerConcern).to be(true) 
  end
end

Answer 2

Simplifying on method 2 from the most voted answer.

I prefer the anonymous controller supported in rspec http://www.relishapp.com/rspec/rspec-rails/docs/controller-specs/anonymous-controller

You will do:

describe ApplicationController, type: :controller do
  controller do
    include MyControllerConcern

    def index; end
  end

  describe 'GET index' do
    it 'will work' do
      get :index
    end
  end
end

Note that you need to describe the ApplicationController and set the type in case this does not happen by default.

Answer 3

My answer may look bit more complicated than these by @Benj and @Calin, but it has its advantages.

describe Concerns::MyConcern, type: :controller do

  described_class.tap do |mod|
    controller(ActionController::Base) { include mod }
  end

  # your tests go here
end

First of all, I recommend the use of anonymous controller which is a subclass of ActionController::Base, not ApplicationController neither any other base controller defined in your application. This way you're able to test the concern in isolation from any of your controllers. If you expect some methods to be defined in a base controller, just stub them.

Furthermore, it is a good idea to avoid re-typing concern module name as it helps to avoid copy-paste errors. Unfortunately, described_class is not accessible in a block passed to controller(ActionController::Base), so I use #tap method to create another binding which stores described_class in a local variable. This is especially important when working with versioned APIs. In such case it is quite common to copy large volume of controllers when creating a new version, and it's terribly easy to make such a subtle copy-paste mistake then.