path: root/guides/source/engines.md

**DO NOT READ THIS FILE ON GITHUB, GUIDES ARE PUBLISHED ON http://guides.rubyonrails.org.**

Getting Started with Engines
============================

In this guide you will learn about engines and how they can be used to provide
additional functionality to their host applications through a clean and very
easy-to-use interface.

After reading this guide, you will know:

* What makes an engine.
* How to generate an engine.
* How to build features for the engine.
* How to hook the engine into an application.
* How to override engine functionality in the application.
* Avoid loading Rails frameworks with Load and Configuration Hooks

--------------------------------------------------------------------------------

What are engines?
-----------------

Engines can be considered miniature applications that provide functionality to
their host applications. A Rails application is actually just a "supercharged"
engine, with the `Rails::Application` class inheriting a lot of its behavior
from `Rails::Engine`.

Therefore, engines and applications can be thought of as almost the same thing,
just with subtle differences, as you'll see throughout this guide. Engines and
applications also share a common structure.

Engines are also closely related to plugins. The two share a common `lib`
directory structure, and are both generated using the `rails plugin new`
generator. The difference is that an engine is considered a "full plugin" by
Rails (as indicated by the `--full` option that's passed to the generator
command). We'll actually be using the `--mountable` option here, which includes
all the features of `--full`, and then some. This guide will refer to these
"full plugins" simply as "engines" throughout. An engine **can** be a plugin,
and a plugin **can** be an engine.

The engine that will be created in this guide will be called "blorgh". This
engine will provide blogging functionality to its host applications, allowing
for new articles and comments to be created. At the beginning of this guide, you
will be working solely within the engine itself, but in later sections you'll
see how to hook it into an application.

Engines can also be isolated from their host applications. This means that an
application is able to have a path provided by a routing helper such as
`articles_path` and use an engine that also provides a path also called
`articles_path`, and the two would not clash. Along with this, controllers, models
and table names are also namespaced. You'll see how to do this later in this
guide.

It's important to keep in mind at all times that the application should
**always** take precedence over its engines. An application is the object that
has final say in what goes on in its environment. The engine should
only be enhancing it, rather than changing it drastically.

To see demonstrations of other engines, check out
[Devise](https://github.com/plataformatec/devise), an engine that provides
authentication for its parent applications, or
[Thredded](https://github.com/thredded/thredded), an engine that provides forum
functionality. There's also [Spree](https://github.com/spree/spree) which
provides an e-commerce platform, and
[RefineryCMS](https://github.com/refinery/refinerycms), a CMS engine.

Finally, engines would not have been possible without the work of James Adam,
Piotr Sarnacki, the Rails Core Team, and a number of other people. If you ever
meet them, don't forget to say thanks!

Generating an engine
--------------------

To generate an engine, you will need to run the plugin generator and pass it
options as appropriate to the need. For the "blorgh" example, you will need to
create a "mountable" engine, running this command in a terminal:

```bash
$ rails plugin new blorgh --mountable
```

The full list of options for the plugin generator may be seen by typing:

```bash
$ rails plugin --help
```

The `--mountable` option tells the generator that you want to create a
"mountable" and namespace-isolated engine. This generator will provide the same
skeleton structure as would the `--full` option. The `--full` option tells the
generator that you want to create an engine, including a skeleton structure
that provides the following:

  * An `app` directory tree
  * A `config/routes.rb` file:

    ```ruby
    Rails.application.routes.draw do
    end
    ```

  * A file at `lib/blorgh/engine.rb`, which is identical in function to a
    standard Rails application's `config/application.rb` file:

    ```ruby
    module Blorgh
      class Engine < ::Rails::Engine
      end
    end
    ```

The `--mountable` option will add to the `--full` option:

  * Asset manifest files (`application.js` and `application.css`)
  * A namespaced `ApplicationController` stub
  * A namespaced `ApplicationHelper` stub
  * A layout view template for the engine
  * Namespace isolation to `config/routes.rb`:

    ```ruby
    Blorgh::Engine.routes.draw do
    end
    ```

  * Namespace isolation to `lib/blorgh/engine.rb`:

    ```ruby
    module Blorgh
      class Engine < ::Rails::Engine
        isolate_namespace Blorgh
      end
    end
    ```

Additionally, the `--mountable` option tells the generator to mount the engine
inside the dummy testing application located at `test/dummy` by adding the
following to the dummy application's routes file at
`test/dummy/config/routes.rb`:

```ruby
mount Blorgh::Engine => "/blorgh"
```

### Inside an Engine

#### Critical Files

At the root of this brand new engine's directory lives a `blorgh.gemspec` file.
When you include the engine into an application later on, you will do so with
this line in the Rails application's `Gemfile`:

```ruby
gem 'blorgh', path: 'engines/blorgh'
```

Don't forget to run `bundle install` as usual. By specifying it as a gem within
the `Gemfile`, Bundler will load it as such, parsing this `blorgh.gemspec` file
and requiring a file within the `lib` directory called `lib/blorgh.rb`. This
file requires the `blorgh/engine.rb` file (located at `lib/blorgh/engine.rb`)
and defines a base module called `Blorgh`.

```ruby
require "blorgh/engine"

module Blorgh
end
```

TIP: Some engines choose to use this file to put global configuration options
for their engine. It's a relatively good idea, so if you want to offer
configuration options, the file where your engine's `module` is defined is
perfect for that. Place the methods inside the module and you'll be good to go.

Within `lib/blorgh/engine.rb` is the base class for the engine:

```ruby
module Blorgh
  class Engine < ::Rails::Engine
    isolate_namespace Blorgh
  end
end
```

By inheriting from the `Rails::Engine` class, this gem notifies Rails that
there's an engine at the specified path, and will correctly mount the engine
inside the application, performing tasks such as adding the `app` directory of
the engine to the load path for models, mailers, controllers, and views.

The `isolate_namespace` method here deserves special notice. This call is
responsible for isolating the controllers, models, routes and other things into
their own namespace, away from similar components inside the application.
Without this, there is a possibility that the engine's components could "leak"
into the application, causing unwanted disruption, or that important engine
components could be overridden by similarly named things within the application.
One of the examples of such conflicts is helpers. Without calling
`isolate_namespace`, the engine's helpers would be included in an application's
controllers.

NOTE: It is **highly** recommended that the `isolate_namespace` line be left
within the `Engine` class definition. Without it, classes generated in an engine
**may** conflict with an application.

What this isolation of the namespace means is that a model generated by a call
to `bin/rails g model`, such as `bin/rails g model article`, won't be called `Article`, but
instead be namespaced and called `Blorgh::Article`. In addition, the table for the
model is namespaced, becoming `blorgh_articles`, rather than simply `articles`.
Similar to the model namespacing, a controller called `ArticlesController` becomes
`Blorgh::ArticlesController` and the views for that controller will not be at
`app/views/articles`, but `app/views/blorgh/articles` instead. Mailers are namespaced
as well.

Finally, routes will also be isolated within the engine. This is one of the most
important parts about namespacing, and is discussed later in the
[Routes](#routes) section of this guide.

#### `app` Directory

Inside the `app` directory are the standard `assets`, `controllers`, `helpers`,
`mailers`, `models` and `views` directories that you should be familiar with
from an application. The `helpers`, `mailers` and `models` directories are
empty, so they aren't described in this section. We'll look more into models in
a future section, when we're writing the engine.

Within the `app/assets` directory, there are the `images`, `javascripts` and
`stylesheets` directories which, again, you should be familiar with due to their
similarity to an application. One difference here, however, is that each
directory contains a sub-directory with the engine name. Because this engine is
going to be namespaced, its assets should be too.

Within the `app/controllers` directory there is a `blorgh` directory that
contains a file called `application_controller.rb`. This file will provide any
common functionality for the controllers of the engine. The `blorgh` directory
is where the other controllers for the engine will go. By placing them within
this namespaced directory, you prevent them from possibly clashing with
identically-named controllers within other engines or even within the
application.

NOTE: The `ApplicationController` class inside an engine is named just like a
Rails application in order to make it easier for you to convert your
applications into engines.

NOTE: Because of the way that Ruby does constant lookup you may run into a situation
where your engine controller is inheriting from the main application controller and
not your engine's application controller. Ruby is able to resolve the `ApplicationController` constant, and therefore the autoloading mechanism is not triggered. See the section [When Constants Aren't Missed](autoloading_and_reloading_constants.html#when-constants-aren-t-missed) of the [Autoloading and Reloading Constants](autoloading_and_reloading_constants.html) guide for further details. The best way to prevent this from
happening is to use `require_dependency` to ensure that the engine's application
controller is loaded. For example:

``` ruby
# app/controllers/blorgh/articles_controller.rb:
require_dependency "blorgh/application_controller"

module Blorgh
  class ArticlesController < ApplicationController
    ...
  end
end
```

WARNING: Don't use `require` because it will break the automatic reloading of classes
in the development environment - using `require_dependency` ensures that classes are
loaded and unloaded in the correct manner.

Lastly, the `app/views` directory contains a `layouts` folder, which contains a
file at `blorgh/application.html.erb`. This file allows you to specify a layout
for the engine. If this engine is to be used as a stand-alone engine, then you
would add any customization to its layout in this file, rather than the
application's `app/views/layouts/application.html.erb` file.

If you don't want to force a layout on to users of the engine, then you can
delete this file and reference a different layout in the controllers of your
engine.

#### `bin` Directory

This directory contains one file, `bin/rails`, which enables you to use the
`rails` sub-commands and generators just like you would within an application.
This means that you will be able to generate new controllers and models for this
engine very easily by running commands like this:

```bash
$ bin/rails g model
```

Keep in mind, of course, that anything generated with these commands inside of
an engine that has `isolate_namespace` in the `Engine` class will be namespaced.

#### `test` Directory

The `test` directory is where tests for the engine will go. To test the engine,
there is a cut-down version of a Rails application embedded within it at
`test/dummy`. This application will mount the engine in the
`test/dummy/config/routes.rb` file:

```ruby
Rails.application.routes.draw do
  mount Blorgh::Engine => "/blorgh"
end
```

This line mounts the engine at the path `/blorgh`, which will make it accessible
through the application only at that path.

Inside the test directory there is the `test/integration` directory, where
integration tests for the engine should be placed. Other directories can be
created in the `test` directory as well. For example, you may wish to create a
`test/models` directory for your model tests.

Providing engine functionality
------------------------------

The engine that this guide covers provides submitting articles and commenting
functionality and follows a similar thread to the [Getting Started
Guide](getting_started.html), with some new twists.

### Generating an Article Resource

The first thing to generate for a blog engine is the `Article` model and related
controller. To quickly generate this, you can use the Rails scaffold generator.

```bash
$ bin/rails generate scaffold article title:string text:text
```

This command will output this information:

```
invoke  active_record
create    db/migrate/[timestamp]_create_blorgh_articles.rb
create    app/models/blorgh/article.rb
invoke    test_unit
create      test/models/blorgh/article_test.rb
create      test/fixtures/blorgh/articles.yml
invoke  resource_route
 route    resources :articles
invoke  scaffold_controller
create    app/controllers/blorgh/articles_controller.rb
invoke    erb
create      app/views/blorgh/articles
create      app/views/blorgh/articles/index.html.erb
create      app/views/blorgh/articles/edit.html.erb
create      app/views/blorgh/articles/show.html.erb
create      app/views/blorgh/articles/new.html.erb
create      app/views/blorgh/articles/_form.html.erb
invoke    test_unit
create      test/controllers/blorgh/articles_controller_test.rb
invoke    helper
create      app/helpers/blorgh/articles_helper.rb
invoke  assets
invoke    js
create      app/assets/javascripts/blorgh/articles.js
invoke    css
create      app/assets/stylesheets/blorgh/articles.css
invoke  css
create    app/assets/stylesheets/scaffold.css
```

The first thing that the scaffold generator does is invoke the `active_record`
generator, which generates a migration and a model for the resource. Note here,
however, that the migration is called `create_blorgh_articles` rather than the
usual `create_articles`. This is due to the `isolate_namespace` method called in
the `Blorgh::Engine` class's definition. The model here is also namespaced,
being placed at `app/models/blorgh/article.rb` rather than `app/models/article.rb` due
to the `isolate_namespace` call within the `Engine` class.

Next, the `test_unit` generator is invoked for this model, generating a model
test at `test/models/blorgh/article_test.rb` (rather than
`test/models/article_test.rb`) and a fixture at `test/fixtures/blorgh/articles.yml`
(rather than `test/fixtures/articles.yml`).

After that, a line for the resource is inserted into the `config/routes.rb` file
for the engine. This line is simply `resources :articles`, turning the
`config/routes.rb` file for the engine into this:

```ruby
Blorgh::Engine.routes.draw do
  resources :articles
end
```

Note here that the routes are drawn upon the `Blorgh::Engine` object rather than
the `YourApp::Application` class. This is so that the engine routes are confined
to the engine itself and can be mounted at a specific point as shown in the
[test directory](#test-directory) section. It also causes the engine's routes to
be isolated from those routes that are within the application. The
[Routes](#routes) section of this guide describes it in detail.

Next, the `scaffold_controller` generator is invoked, generating a controller
called `Blorgh::ArticlesController` (at
`app/controllers/blorgh/articles_controller.rb`) and its related views at
`app/views/blorgh/articles`. This generator also generates a test for the
controller (`test/controllers/blorgh/articles_controller_test.rb`) and a helper
(`app/helpers/blorgh/articles_helper.rb`).

Everything this generator has created is neatly namespaced. The controller's
class is defined within the `Blorgh` module:

```ruby
module Blorgh
  class ArticlesController < ApplicationController
    ...
  end
end
```

NOTE: The `ArticlesController` class inherits from
`Blorgh::ApplicationController`, not the application's `ApplicationController`.

The helper inside `app/helpers/blorgh/articles_helper.rb` is also namespaced:

```ruby
module Blorgh
  module ArticlesHelper
    ...
  end
end
```

This helps prevent conflicts with any other engine or application that may have
an article resource as well.

Finally, the assets for this resource are generated in two files:
`app/assets/javascripts/blorgh/articles.js` and
`app/assets/stylesheets/blorgh/articles.css`. You'll see how to use these a little
later.

You can see what the engine has so far by running `bin/rails db:migrate` at the root
of our engine to run the migration generated by the scaffold generator, and then
running `rails server` in `test/dummy`. When you open
`http://localhost:3000/blorgh/articles` you will see the default scaffold that has
been generated. Click around! You've just generated your first engine's first
functions.

If you'd rather play around in the console, `rails console` will also work just
like a Rails application. Remember: the `Article` model is namespaced, so to
reference it you must call it as `Blorgh::Article`.

```ruby
>> Blorgh::Article.find(1)
=> #<Blorgh::Article id: 1 ...>
```

One final thing is that the `articles` resource for this engine should be the root
of the engine. Whenever someone goes to the root path where the engine is
mounted, they should be shown a list of articles. This can be made to happen if
this line is inserted into the `config/routes.rb` file inside the engine:

```ruby
root to: "articles#index"
```

Now people will only need to go to the root of the engine to see all the articles,
rather than visiting `/articles`. This means that instead of
`http://localhost:3000/blorgh/articles`, you only need to go to
`http://localhost:3000/blorgh` now.

### Generating a Comments Resource

Now that the engine can create new articles, it only makes sense to add
commenting functionality as well. To do this, you'll need to generate a comment
model, a comment controller and then modify the articles scaffold to display
comments and allow people to create new ones.

From the application root, run the model generator. Tell it to generate a
`Comment` model, with the related table having two columns: an `article_id` integer
and `text` text column.

```bash
$ bin/rails generate model Comment article_id:integer text:text
```

This will output the following:

```
invoke  active_record
create    db/migrate/[timestamp]_create_blorgh_comments.rb
create    app/models/blorgh/comment.rb
invoke    test_unit
create      test/models/blorgh/comment_test.rb
create      test/fixtures/blorgh/comments.yml
```

This generator call will generate just the necessary model files it needs,
namespacing the files under a `blorgh` directory and creating a model class
called `Blorgh::Comment`. Now run the migration to create our blorgh_comments
table:

```bash
$ bin/rails db:migrate
```

To show the comments on an article, edit `app/views/blorgh/articles/show.html.erb` and
add this line before the "Edit" link:

```html+erb
<h3>Comments</h3>
<%= render @article.comments %>
```

This line will require there to be a `has_many` association for comments defined
on the `Blorgh::Article` model, which there isn't right now. To define one, open
`app/models/blorgh/article.rb` and add this line into the model:

```ruby
has_many :comments
```

Turning the model into this:

```ruby
module Blorgh
  class Article < ApplicationRecord
    has_many :comments
  end
end
```

NOTE: Because the `has_many` is defined inside a class that is inside the
`Blorgh` module, Rails will know that you want to use the `Blorgh::Comment`
model for these objects, so there's no need to specify that using the
`:class_name` option here.

Next, there needs to be a form so that comments can be created on an article. To
add this, put this line underneath the call to `render @article.comments` in
`app/views/blorgh/articles/show.html.erb`:

```erb
<%= render "blorgh/comments/form" %>
```

Next, the partial that this line will render needs to exist. Create a new
directory at `app/views/blorgh/comments` and in it a new file called
`_form.html.erb` which has this content to create the required partial:

```html+erb
<h3>New comment</h3>
<%= form_for [@article, @article.comments.build] do |f| %>
  <p>
    <%= f.label :text %><br>
    <%= f.text_area :text %>
  </p>
  <%= f.submit %>
<% end %>
```

When this form is submitted, it is going to attempt to perform a `POST` request
to a route of `/articles/:article_id/comments` within the engine. This route doesn't
exist at the moment, but can be created by changing the `resources :articles` line
inside `config/routes.rb` into these lines:

```ruby
resources :articles do
  resources :comments
end
```

This creates a nested route for the comments, which is what the form requires.

The route now exists, but the controller that this route goes to does not. To
create it, run this command from the application root:

```bash
$ bin/rails g controller comments
```

This will generate the following things:

```
create  app/controllers/blorgh/comments_controller.rb
invoke  erb
 exist    app/views/blorgh/comments
invoke  test_unit
create    test/controllers/blorgh/comments_controller_test.rb
invoke  helper
create    app/helpers/blorgh/comments_helper.rb
invoke  assets
invoke    js
create      app/assets/javascripts/blorgh/comments.js
invoke    css
create      app/assets/stylesheets/blorgh/comments.css
```

The form will be making a `POST` request to `/articles/:article_id/comments`, which
will correspond with the `create` action in `Blorgh::CommentsController`. This
action needs to be created, which can be done by putting the following lines
inside the class definition in `app/controllers/blorgh/comments_controller.rb`:

```ruby
def create
  @article = Article.find(params[:article_id])
  @comment = @article.comments.create(comment_params)
  flash[:notice] = "Comment has been created!"
  redirect_to articles_path
end

private
  def comment_params
    params.require(:comment).permit(:text)
  end
```

This is the final step required to get the new comment form working. Displaying
the comments, however, is not quite right yet. If you were to create a comment
right now, you would see this error:

```
Missing partial blorgh/comments/_comment with {:handlers=>[:erb, :builder],
:formats=>[:html], :locale=>[:en, :en]}. Searched in:   *
"/Users/ryan/Sites/side_projects/blorgh/test/dummy/app/views"   *
"/Users/ryan/Sites/side_projects/blorgh/app/views"
```

The engine is unable to find the partial required for rendering the comments.
Rails looks first in the application's (`test/dummy`) `app/views` directory and
then in the engine's `app/views` directory. When it can't find it, it will throw
this error. The engine knows to look for `blorgh/comments/_comment` because the
model object it is receiving is from the `Blorgh::Comment` class.

This partial will be responsible for rendering just the comment text, for now.
Create a new file at `app/views/blorgh/comments/_comment.html.erb` and put this
line inside it:

```erb
<%= comment_counter + 1 %>. <%= comment.text %>
```

The `comment_counter` local variable is given to us by the `<%= render
@article.comments %>` call, which will define it automatically and increment the
counter as it iterates through each comment. It's used in this example to
display a small number next to each comment when it's created.

That completes the comment function of the blogging engine. Now it's time to use
it within an application.

Hooking Into an Application
---------------------------

Using an engine within an application is very easy. This section covers how to
mount the engine into an application and the initial setup required, as well as
linking the engine to a `User` class provided by the application to provide
ownership for articles and comments within the engine.

### Mounting the Engine

First, the engine needs to be specified inside the application's `Gemfile`. If
there isn't an application handy to test this out in, generate one using the
`rails new` command outside of the engine directory like this:

```bash
$ rails new unicorn
```

Usually, specifying the engine inside the Gemfile would be done by specifying it
as a normal, everyday gem.

```ruby
gem 'devise'
```

However, because you are developing the `blorgh` engine on your local machine,
you will need to specify the `:path` option in your `Gemfile`:

```ruby
gem 'blorgh', path: 'engines/blorgh'
```

Then run `bundle` to install the gem.

As described earlier, by placing the gem in the `Gemfile` it will be loaded when
Rails is loaded. It will first require `lib/blorgh.rb` from the engine, then
`lib/blorgh/engine.rb`, which is the file that defines the major pieces of
functionality for the engine.

To make the engine's functionality accessible from within an application, it
needs to be mounted in that application's `config/routes.rb` file:

```ruby
mount Blorgh::Engine, at: "/blog"
```

This line will mount the engine at `/blog` in the application. Making it
accessible at `http://localhost:3000/blog` when the application runs with `rails
server`.

NOTE: Other engines, such as Devise, handle this a little differently by making
you specify custom helpers (such as `devise_for`) in the routes. These helpers
do exactly the same thing, mounting pieces of the engines's functionality at a
pre-defined path which may be customizable.

### Engine setup

The engine contains migrations for the `blorgh_articles` and `blorgh_comments`
table which need to be created in the application's database so that the
engine's models can query them correctly. To copy these migrations into the
application run the following command from the `test/dummy` directory of your Rails engine:

```bash
$ bin/rails blorgh:install:migrations
```

If you have multiple engines that need migrations copied over, use
`railties:install:migrations` instead:

```bash
$ bin/rails railties:install:migrations
```

This command, when run for the first time, will copy over all the migrations
from the engine. When run the next time, it will only copy over migrations that
haven't been copied over already. The first run for this command will output
something such as this:

```bash
Copied migration [timestamp_1]_create_blorgh_articles.blorgh.rb from blorgh
Copied migration [timestamp_2]_create_blorgh_comments.blorgh.rb from blorgh
```

The first timestamp (`[timestamp_1]`) will be the current time, and the second
timestamp (`[timestamp_2]`) will be the current time plus a second. The reason
for this is so that the migrations for the engine are run after any existing
migrations in the application.

To run these migrations within the context of the application, simply run `bin/rails
db:migrate`. When accessing the engine through `http://localhost:3000/blog`, the
articles will be empty. This is because the table created inside the application is
different from the one created within the engine. Go ahead, play around with the
newly mounted engine. You'll find that it's the same as when it was only an
engine.

If you would like to run migrations only from one engine, you can do it by
specifying `SCOPE`:

```bash
bin/rails db:migrate SCOPE=blorgh
```

This may be useful if you want to revert engine's migrations before removing it.
To revert all migrations from blorgh engine you can run code such as:

```bash
bin/rails db:migrate SCOPE=blorgh VERSION=0
```

### Using a Class Provided by the Application

#### Using a Model Provided by the Application

When an engine is created, it may want to use specific classes from an
application to provide links between the pieces of the engine and the pieces of
the application. In the case of the `blorgh` engine, making articles and comments
have authors would make a lot of sense.

A typical application might have a `User` class that would be used to represent
authors for an article or a comment. But there could be a case where the
application calls this class something different, such as `Person`. For this
reason, the engine should not hardcode associations specifically for a `User`
class.

To keep it simple in this case, the application will have a class called `User`
that represents the users of the application (we'll get into making this
configurable further on). It can be generated using this command inside the
application:

```bash
rails g model user name:string
```

The `bin/rails db:migrate` command needs to be run here to ensure that our
application has the `users` table for future use.

Also, to keep it simple, the articles form will have a new text field called
`author_name`, where users can elect to put their name. The engine will then
take this name and either create a new `User` object from it, or find one that
already has that name. The engine will then associate the article with the found or
created `User` object.

First, the `author_name` text field needs to be added to the
`app/views/blorgh/articles/_form.html.erb` partial inside the engine. This can be
added above the `title` field with this code:

```html+erb
<div class="field">
  <%= f.label :author_name %><br>
  <%= f.text_field :author_name %>
</div>
```

Next, we need to update our `Blorgh::ArticleController#article_params` method to
permit the new form parameter:

```ruby
def article_params
  params.require(:article).permit(:title, :text, :author_name)
end
```

The `Blorgh::Article` model should then have some code to convert the `author_name`
field into an actual `User` object and associate it as that article's `author`
before the article is saved. It will also need to have an `attr_accessor` set up
for this field, so that the setter and getter methods are defined for it.

To do all this, you'll need to add the `attr_accessor` for `author_name`, the
association for the author and the `before_validation` call into
`app/models/blorgh/article.rb`. The `author` association will be hard-coded to the
`User` class for the time being.

```ruby
attr_accessor :author_name
belongs_to :author, class_name: "User"

before_validation :set_author

private
  def set_author
    self.author = User.find_or_create_by(name: author_name)
  end
```

By representing the `author` association's object with the `User` class, a link
is established between the engine and the application. There needs to be a way
of associating the records in the `blorgh_articles` table with the records in the
`users` table. Because the association is called `author`, there should be an
`author_id` column added to the `blorgh_articles` table.

To generate this new column, run this command within the engine:

```bash
$ bin/rails g migration add_author_id_to_blorgh_articles author_id:integer
```

NOTE: Due to the migration's name and the column specification after it, Rails
will automatically know that you want to add a column to a specific table and
write that into the migration for you. You don't need to tell it any more than
this.

This migration will need to be run on the application. To do that, it must first
be copied using this command:

```bash
$ bin/rails blorgh:install:migrations
```

Notice that only _one_ migration was copied over here. This is because the first
two migrations were copied over the first time this command was run.

```
NOTE Migration [timestamp]_create_blorgh_articles.blorgh.rb from blorgh has been skipped. Migration with the same name already exists.
NOTE Migration [timestamp]_create_blorgh_comments.blorgh.rb from blorgh has been skipped. Migration with the same name already exists.
Copied migration [timestamp]_add_author_id_to_blorgh_articles.blorgh.rb from blorgh
```

Run the migration using:

```bash
$ bin/rails db:migrate
```

Now with all the pieces in place, an action will take place that will associate
an author - represented by a record in the `users` table - with an article,
represented by the `blorgh_articles` table from the engine.

Finally, the author's name should be displayed on the article's page. Add this code
above the "Title" output inside `app/views/blorgh/articles/show.html.erb`:

```html+erb
<p>
  <b>Author:</b>
  <%= @article.author.name %>
</p>
```

#### Using a Controller Provided by the Application

Because Rails controllers generally share code for things like authentication
and accessing session variables, they inherit from `ApplicationController` by
default. Rails engines, however are scoped to run independently from the main
application, so each engine gets a scoped `ApplicationController`. This
namespace prevents code collisions, but often engine controllers need to access
methods in the main application's `ApplicationController`. An easy way to
provide this access is to change the engine's scoped `ApplicationController` to
inherit from the main application's `ApplicationController`. For our Blorgh
engine this would be done by changing
`app/controllers/blorgh/application_controller.rb` to look like:

```ruby
module Blorgh
  class ApplicationController < ::ApplicationController
  end
end
```

By default, the engine's controllers inherit from
`Blorgh::ApplicationController`. So, after making this change they will have
access to the main application's `ApplicationController`, as though they were
part of the main application.

This change does require that the engine is run from a Rails application that
has an `ApplicationController`.

### Configuring an Engine

This section covers how to make the `User` class configurable, followed by
general configuration tips for the engine.

#### Setting Configuration Settings in the Application

The next step is to make the class that represents a `User` in the application
customizable for the engine. This is because that class may not always be
`User`, as previously explained. To make this setting customizable, the engine
will have a configuration setting called `author_class` that will be used to
specify which class represents users inside the application.

To define this configuration setting, you should use a `mattr_accessor` inside
the `Blorgh` module for the engine. Add this line to `lib/blorgh.rb` inside the
engine:

```ruby
mattr_accessor :author_class
```

This method works like its brothers, `attr_accessor` and `cattr_accessor`, but
provides a setter and getter method on the module with the specified name. To
use it, it must be referenced using `Blorgh.author_class`.

The next step is to switch the `Blorgh::Article` model over to this new setting.
Change the `belongs_to` association inside this model
(`app/models/blorgh/article.rb`) to this:

```ruby
belongs_to :author, class_name: Blorgh.author_class
```

The `set_author` method in the `Blorgh::Article` model should also use this class:

```ruby
self.author = Blorgh.author_class.constantize.find_or_create_by(name: author_name)
```

To save having to call `constantize` on the `author_class` result all the time,
you could instead just override the `author_class` getter method inside the
`Blorgh` module in the `lib/blorgh.rb` file to always call `constantize` on the
saved value before returning the result:

```ruby
def self.author_class
  @@author_class.constantize
end
```

This would then turn the above code for `set_author` into this:

```ruby
self.author = Blorgh.author_class.find_or_create_by(name: author_name)
```

Resulting in something a little shorter, and more implicit in its behavior. The
`author_class` method should always return a `Class` object.

Since we changed the `author_class` method to return a `Class` instead of a
`String`, we must also modify our `belongs_to` definition in the `Blorgh::Article`
model:

```ruby
belongs_to :author, class_name: Blorgh.author_class.to_s
```

To set this configuration setting within the application, an initializer should
be used. By using an initializer, the configuration will be set up before the
application starts and calls the engine's models, which may depend on this
configuration setting existing.

Create a new initializer at `config/initializers/blorgh.rb` inside the
application where the `blorgh` engine is installed and put this content in it:

```ruby
Blorgh.author_class = "User"
```

WARNING: It's very important here to use the `String` version of the class,
rather than the class itself. If you were to use the class, Rails would attempt
to load that class and then reference the related table. This could lead to
problems if the table wasn't already existing. Therefore, a `String` should be
used and then converted to a class using `constantize` in the engine later on.

Go ahead and try to create a new article. You will see that it works exactly in the
same way as before, except this time the engine is using the configuration
setting in `config/initializers/blorgh.rb` to learn what the class is.

There are now no strict dependencies on what the class is, only what the API for
the class must be. The engine simply requires this class to define a
`find_or_create_by` method which returns an object of that class, to be
associated with an article when it's created. This object, of course, should have
some sort of identifier by which it can be referenced.

#### General Engine Configuration

Within an engine, there may come a time where you wish to use things such as
initializers, internationalization or other configuration options. The great
news is that these things are entirely possible, because a Rails engine shares
much the same functionality as a Rails application. In fact, a Rails
application's functionality is actually a superset of what is provided by
engines!

If you wish to use an initializer - code that should run before the engine is
loaded - the place for it is the `config/initializers` folder. This directory's
functionality is explained in the [Initializers
section](configuring.html#initializers) of the Configuring guide, and works
precisely the same way as the `config/initializers` directory inside an
application. The same thing goes if you want to use a standard initializer.

For locales, simply place the locale files in the `config/locales` directory,
just like you would in an application.

Testing an engine
-----------------

When an engine is generated, there is a smaller dummy application created inside
it at `test/dummy`. This application is used as a mounting point for the engine,
to make testing the engine extremely simple. You may extend this application by
generating controllers, models or views from within the directory, and then use
those to test your engine.

The `test` directory should be treated like a typical Rails testing environment,
allowing for unit, functional and integration tests.

### Functional Tests

A matter worth taking into consideration when writing functional tests is that
the tests are going to be running on an application - the `test/dummy`
application - rather than your engine. This is due to the setup of the testing
environment; an engine needs an application as a host for testing its main
functionality, especially controllers. This means that if you were to make a
typical `GET` to a controller in a controller's functional test like this:

```ruby
module Blorgh
  class FooControllerTest < ActionDispatch::IntegrationTest
    include Engine.routes.url_helpers

    def test_index
      get foos_url
      ...
    end
  end
end
```

It may not function correctly. This is because the application doesn't know how
to route these requests to the engine unless you explicitly tell it **how**. To
do this, you must set the `@routes` instance variable to the engine's route set
in your setup code:

```ruby
module Blorgh
  class FooControllerTest < ActionDispatch::IntegrationTest
    include Engine.routes.url_helpers

    setup do
      @routes = Engine.routes
    end

    def test_index
      get foos_url
      ...
    end
  end
end
```

This tells the application that you still want to perform a `GET` request to the
`index` action of this controller, but you want to use the engine's route to get
there, rather than the application's one.

This also ensures that the engine's URL helpers will work as expected in your
tests.

Improving engine functionality
------------------------------

This section explains how to add and/or override engine MVC functionality in the
main Rails application.

### Overriding Models and Controllers

Engine model and controller classes can be extended by open classing them in the
main Rails application (since model and controller classes are just Ruby classes
that inherit Rails specific functionality). Open classing an Engine class
redefines it for use in the main application. This is usually implemented by
using the decorator pattern.

For simple class modifications, use `Class#class_eval`. For complex class
modifications, consider using `ActiveSupport::Concern`.

#### A note on Decorators and Loading Code

Because these decorators are not referenced by your Rails application itself,
Rails' autoloading system will not kick in and load your decorators. This means
that you need to require them yourself.

Here is some sample code to do this:

```ruby
# lib/blorgh/engine.rb
module Blorgh
  class Engine < ::Rails::Engine
    isolate_namespace Blorgh

    config.to_prepare do
      Dir.glob(Rails.root + "app/decorators/**/*_decorator*.rb").each do |c|
        require_dependency(c)
      end
    end
  end
end
```

This doesn't apply to just Decorators, but anything that you add in an engine
that isn't referenced by your main application.

#### Implementing Decorator Pattern Using Class#class_eval

**Adding** `Article#time_since_created`:

```ruby
# MyApp/app/decorators/models/blorgh/article_decorator.rb

Blorgh::Article.class_eval do
  def time_since_created
    Time.current - created_at
  end
end
```

```ruby
# Blorgh/app/models/article.rb

class Article < ApplicationRecord
  has_many :comments
end
```


**Overriding** `Article#summary`:

```ruby
# MyApp/app/decorators/models/blorgh/article_decorator.rb

Blorgh::Article.class_eval do
  def summary
    "#{title} - #{truncate(text)}"
  end
end
```

```ruby
# Blorgh/app/models/article.rb

class Article < ApplicationRecord
  has_many :comments
  def summary
    "#{title}"
  end
end
```

#### Implementing Decorator Pattern Using ActiveSupport::Concern

Using `Class#class_eval` is great for simple adjustments, but for more complex
class modifications, you might want to consider using [`ActiveSupport::Concern`]
(http://api.rubyonrails.org/classes/ActiveSupport/Concern.html).
ActiveSupport::Concern manages load order of interlinked dependent modules and
classes at run time allowing you to significantly modularize your code.

**Adding** `Article#time_since_created` and **Overriding** `Article#summary`:

```ruby
# MyApp/app/models/blorgh/article.rb

class Blorgh::Article < ApplicationRecord
  include Blorgh::Concerns::Models::Article

  def time_since_created
    Time.current - created_at
  end

  def summary
    "#{title} - #{truncate(text)}"
  end
end
```

```ruby
# Blorgh/app/models/article.rb

class Article < ApplicationRecord
  include Blorgh::Concerns::Models::Article
end
```

```ruby
# Blorgh/lib/concerns/models/article.rb

module Blorgh::Concerns::Models::Article
  extend ActiveSupport::Concern

  # 'included do' causes the included code to be evaluated in the
  # context where it is included (article.rb), rather than being
  # executed in the module's context (blorgh/concerns/models/article).
  included do
    attr_accessor :author_name
    belongs_to :author, class_name: "User"

    before_validation :set_author

    private
      def set_author
        self.author = User.find_or_create_by(name: author_name)
      end
  end

  def summary
    "#{title}"
  end

  module ClassMethods
    def some_class_method
      'some class method string'
    end
  end
end
```

### Overriding Views

When Rails looks for a view to render, it will first look in the `app/views`
directory of the application. If it cannot find the view there, it will check in
the `app/views` directories of all engines that have this directory.

When the application is asked to render the view for `Blorgh::ArticlesController`'s
index action, it will first look for the path
`app/views/blorgh/articles/index.html.erb` within the application. If it cannot
find it, it will look inside the engine.

You can override this view in the application by simply creating a new file at
`app/views/blorgh/articles/index.html.erb`. Then you can completely change what
this view would normally output.

Try this now by creating a new file at `app/views/blorgh/articles/index.html.erb`
and put this content in it:

```html+erb
<h1>Articles</h1>
<%= link_to "New Article", new_article_path %>
<% @articles.each do |article| %>
  <h2><%= article.title %></h2>
  <small>By <%= article.author %></small>
  <%= simple_format(article.text) %>
  <hr>
<% end %>
```

### Routes

Routes inside an engine are isolated from the application by default. This is
done by the `isolate_namespace` call inside the `Engine` class. This essentially
means that the application and its engines can have identically named routes and
they will not clash.

Routes inside an engine are drawn on the `Engine` class within
`config/routes.rb`, like this:

```ruby
Blorgh::Engine.routes.draw do
  resources :articles
end
```

By having isolated routes such as this, if you wish to link to an area of an
engine from within an application, you will need to use the engine's routing
proxy method. Calls to normal routing methods such as `articles_path` may end up
going to undesired locations if both the application and the engine have such a
helper defined.

For instance, the following example would go to the application's `articles_path`
if that template was rendered from the application, or the engine's `articles_path`
if it was rendered from the engine:

```erb
<%= link_to "Blog articles", articles_path %>
```

To make this route always use the engine's `articles_path` routing helper method,
we must call the method on the routing proxy method that shares the same name as
the engine.

```erb
<%= link_to "Blog articles", blorgh.articles_path %>
```

If you wish to reference the application inside the engine in a similar way, use
the `main_app` helper:

```erb
<%= link_to "Home", main_app.root_path %>
```

If you were to use this inside an engine, it would **always** go to the
application's root. If you were to leave off the `main_app` "routing proxy"
method call, it could potentially go to the engine's or application's root,
depending on where it was called from.

If a template rendered from within an engine attempts to use one of the
application's routing helper methods, it may result in an undefined method call.
If you encounter such an issue, ensure that you're not attempting to call the
application's routing methods without the `main_app` prefix from within the
engine.

### Assets

Assets within an engine work in an identical way to a full application. Because
the engine class inherits from `Rails::Engine`, the application will know to
look up assets in the engine's 'app/assets' and 'lib/assets' directories.

Like all of the other components of an engine, the assets should be namespaced.
This means that if you have an asset called `style.css`, it should be placed at
`app/assets/stylesheets/[engine name]/style.css`, rather than
`app/assets/stylesheets/style.css`. If this asset isn't namespaced, there is a
possibility that the host application could have an asset named identically, in
which case the application's asset would take precedence and the engine's one
would be ignored.

Imagine that you did have an asset located at
`app/assets/stylesheets/blorgh/style.css` To include this asset inside an
application, just use `stylesheet_link_tag` and reference the asset as if it
were inside the engine:

```erb
<%= stylesheet_link_tag "blorgh/style.css" %>
```

You can also specify these assets as dependencies of other assets using Asset
Pipeline require statements in processed files:

```
/*
 *= require blorgh/style
*/
```

INFO. Remember that in order to use languages like Sass or CoffeeScript, you
should add the relevant library to your engine's `.gemspec`.

### Separate Assets & Precompiling

There are some situations where your engine's assets are not required by the
host application. For example, say that you've created an admin functionality
that only exists for your engine. In this case, the host application doesn't
need to require `admin.css` or `admin.js`. Only the gem's admin layout needs
these assets. It doesn't make sense for the host app to include
`"blorgh/admin.css"` in its stylesheets. In this situation, you should
explicitly define these assets for precompilation.  This tells sprockets to add
your engine assets when `bin/rails assets:precompile` is triggered.

You can define assets for precompilation in `engine.rb`:

```ruby
initializer "blorgh.assets.precompile" do |app|
  app.config.assets.precompile += %w( admin.js admin.css )
end
```

For more information, read the [Asset Pipeline guide](asset_pipeline.html).

### Other Gem Dependencies

Gem dependencies inside an engine should be specified inside the `.gemspec` file
at the root of the engine. The reason is that the engine may be installed as a
gem. If dependencies were to be specified inside the `Gemfile`, these would not
be recognized by a traditional gem install and so they would not be installed,
causing the engine to malfunction.

To specify a dependency that should be installed with the engine during a
traditional `gem install`, specify it inside the `Gem::Specification` block
inside the `.gemspec` file in the engine:

```ruby
s.add_dependency "moo"
```

To specify a dependency that should only be installed as a development
dependency of the application, specify it like this:

```ruby
s.add_development_dependency "moo"
```

Both kinds of dependencies will be installed when `bundle install` is run inside
of the application. The development dependencies for the gem will only be used
when the tests for the engine are running.

Note that if you want to immediately require dependencies when the engine is
required, you should require them before the engine's initialization. For
example:

```ruby
require 'other_engine/engine'
require 'yet_another_engine/engine'

module MyEngine
  class Engine < ::Rails::Engine
  end
end
```

Active Support On Load Hooks
----------------------------

Active Support is the Ruby on Rails component responsible for providing Ruby language extensions, utilities, and other transversal utilities.

Rails code can often be referenced on load of an application. Rails is responsible for the load order of these frameworks, so when you load frameworks, such as `ActiveRecord::Base`, prematurely you are violating an implicit contract your application has with Rails. Moreover, by loading code such as `ActiveRecord::Base` on boot of your application you are loading entire frameworks which may slow down your boot time and could cause conflicts with load order and boot of your application.

On Load hooks are the API that allow you to hook into this initialization process without violating the load contract with Rails. This will also mitigate boot performance degradation and avoid conflicts.

## What are `on_load` hooks?

Since Ruby is a dynamic language, some code will cause different Rails frameworks to load. Take this snippet for instance:

```ruby
ActiveRecord::Base.include(MyActiveRecordHelper)
```

This snippet means that when this file is loaded, it will encounter `ActiveRecord::Base`. This encounter causes Ruby to look for the definition of that constant and will require it. This causes the entire Active Record framework to be loaded on boot.

`ActiveSupport.on_load` is a mechanism that can be used to defer the loading of code until it is actually needed. The snippet above can be changed to:

```ruby
ActiveSupport.on_load(:active_record) { include MyActiveRecordHelper }
```

This new snippet will only include `MyActiveRecordHelper` when `ActiveRecord::Base` is loaded.

## How does it work?

In the Rails framework these hooks are called when a specific library is loaded. For example, when `ActionController::Base` is loaded, the `:action_controller_base` hook is called. This means that all `ActiveSupport.on_load` calls with `:action_controller_base` hooks will be called in the context of `ActionController::Base` (that means `self` will be an `ActionController::Base`).

## Modifying code to use `on_load` hooks

Modifying code is generally straightforward. If you have a line of code that refers to a Rails framework such as `ActiveRecord::Base` you can wrap that code in an `on_load` hook.

### Example 1

```ruby
ActiveRecord::Base.include(MyActiveRecordHelper)
```

becomes

```ruby
ActiveSupport.on_load(:active_record) { include MyActiveRecordHelper } # self refers to ActiveRecord::Base here, so we can simply #include
```

### Example 2

```ruby
ActionController::Base.prepend(MyActionControllerHelper)
```

becomes

```ruby
ActiveSupport.on_load(:action_controller_base) { prepend MyActionControllerHelper } # self refers to ActionController::Base here, so we can simply #prepend
```

### Example 3

```ruby
ActiveRecord::Base.include_root_in_json = true
```

becomes

```ruby
ActiveSupport.on_load(:active_record) { self.include_root_in_json = true } # self refers to ActiveRecord::Base here
```

## Available Hooks

These are the hooks you can use in your own code.

To hook into the initialization process of one of the following classes use the available hook.

| Class                             | Available Hooks                      |
| --------------------------------- | ------------------------------------ |
| `ActionCable`                     | `action_cable`                       |
| `ActionController::API`           | `action_controller_api`              |
| `ActionController::API`           | `action_controller`                  |
| `ActionController::Base`          | `action_controller_base`             |
| `ActionController::Base`          | `action_controller`                  |
| `ActionController::TestCase`      | `action_controller_test_case`        |
| `ActionDispatch::IntegrationTest` | `action_dispatch_integration_test`   |
| `ActionMailer::Base`              | `action_mailer`                      |
| `ActionMailer::TestCase`          | `action_mailer_test_case`            |
| `ActionView::Base`                | `action_view`                        |
| `ActionView::TestCase`            | `action_view_test_case`              |
| `ActiveJob::Base`                 | `active_job`                         |
| `ActiveJob::TestCase`             | `active_job_test_case`               |
| `ActiveRecord::Base`              | `active_record`                      |
| `ActiveSupport::TestCase`         | `active_support_test_case`           |
| `i18n`                            | `i18n`                               |

## Configuration hooks

These are the available configuration hooks. They do not hook into any particular framework, instead they run in context of the entire application.

| Hook                   | Use Case                                                                              |
| ---------------------- | ------------------------------------------------------------------------------------- |
| `before_configuration` | First configurable block to run. Called before any initializers are run.              |
| `before_initialize`    | Second configurable block to run. Called before frameworks initialize.                |
| `before_eager_load`    | Third configurable block to run. Does not run if `config.cache_classes` set to false. |
| `after_initialize`     | Last configurable block to run. Called after frameworks initialize.                   |

### Example

`config.before_configuration { puts 'I am called before any initializers' }`