Associations

Associations are all over the place in Rails and are incredibly useful because of all the new methods they give you access to.

Introduction

You've already had some familiarity with associations, especially the basic has_one, has_many, and belongs_to variety. Thus far, you've probably mostly used these associations to grab collections of objects like a user's posts (user.posts). There are a lot of other handy things that Rails lets you do with associations too. This brief section will highlight some of the more useful methods that come along with associations.

Learning outcomes

Look through these now and then use them to test yourself after doing the assignment:

  • How does Rails normally know which table and foreign key to use when you have an association (e.g. User.first.posts)?

  • When would you need to specify the :class_name option in an association?

  • What about the :foreign_key?

  • What about the :source?

  • What is a polymorphic association and when would you use one?

  • What are two ways to use the association to create a new object instead of just calling YourObject.new? Why is this useful? Which method is preferred?

  • How do you automatically destroy all a User's Post objects if that user is deleted?

  • How do you set up a self-association, like with Users following Users?

Basics

If you're still shaky on basic associations, go back and check out the Associations section of the Basic Active Record lesson first. This section is meant to just bring up some of the basic stuff you may not yet have been exposed to.

Foreign keys and class names

When you create an association, Rails makes two major assumptions -- first, that the class of the model your association points to is based directly off of the name of the association, and, second, that the foreign key in any belongs_to relationship will be called yourassociationname_id. Any time you go away from these defaults, you just need to let Rails know what kind of class to look for and which foreign key to use.

A very simple case would be a User who can create many Posts for a blog:

# app/models/user.rb
class User < ActiveRecord::Base
  has_many :posts
end
# app/models/post.rb
class Post < ActiveRecord::Base
  belongs_to :user
end

So you could ask the first user for all her posts with User.first.posts or the first post for its author user with Post.first.user. Rails knows to look for a foreign key called user_id in the Posts table. If you ask for Post.first.user, Rails will look in the Users table for the User with the ID corresponding to the user_id column in the Posts table. All is well in the world when your association names correspond directly to the names of your models and tables.

But what if you want to have two types of users that the post belongs to -- the "author" and the "editor"? In this case, you'll need two separate foreign keys in your Posts table, presumably one called author_id and another called editor_id. How do you tell Rails that each of these foreign keys actually point to a User (so it knows to look in the Users table for them)? Just specify the class your association should point to using the aptly named :class_name option:

# app/models/post.rb
class Post < ActiveRecord::Base
  belongs_to :author, class_name: 'User'
  belongs_to :editor, class_name: 'User'
end

In this case, Rails will automatically look for the foreign key named after the association, e.g. author_id or editor_id, in the Posts table.

If you called the association something which didn't correspond exactly to what you'd named the foreign key in your table, you need to tell Rails that as well. This should be obvious if you think of this relationship from the side of the User. The User will have some posts for which she is an author and others for which she is an editor. You'll need to rename the association on the User's side as well to keep things crystal clear, for instance splitting up authored_posts from edited_posts.

But now Rails doesn't have the foggiest idea where to look and what to look for. By default, if you ask for User.first.authored_posts it will go looking in the authored_posts table for a foreign key called user_id (neither of which exist). To get it pointing at the right table, we again need to specify the :class_name and to get it using the correct foreign key, we need to specify the right :foreign_key. For instance:

# app/models/user.rb
class User < ActiveRecord::Base
  has_many :authored_posts, foreign_key: 'author_id', class_name: 'Post'
  has_many :edited_posts, foreign_key: 'editor_id', class_name: 'Post'
end

The basic gist of this is simple -- assume that Rails is looking for the foreign key named after the association in the table also named after the association. If any of these are incorrect because of a creatively named association or foreign key, you'll need to specify. This is quite common to make your associations more legible.

Source

Now that it's clear you need to let Rails know when you've creatively named your associations or foreign keys, we should point out that there's one additional step required if you're using a creatively named has_many :through association. Recall that has-many-through associations are where you create a "through table" to act as a go-between for two models that have a many-to-many relationship.

For example, perhaps we change the example above so a Post actually can have multiple Authors (but still only one editor). We'll need to create a new table, which we'll call post_authorings. post_authorings joins these two models together and contains columns for authored_post_id and post_author_id. You can probably see where this is going -- we've named our foreign keys something more descriptive and helpful than just simply post_id and user_id but it will require us to inform Rails of the change. Our models look like:

# app/models/post.rb
class Post < ActiveRecord::Base
  has_many :post_authorings, foreign_key: :authored_post_id
  has_many :authors, through: :post_authorings, source: :post_author
  belongs_to :editor, class_name: 'User'
end
# app/models/user.rb
class User < ActiveRecord::Base
  has_many :post_authorings, foreign_key: :post_author_id
  has_many :authored_posts, through: :post_authorings
  has_many :edited_posts, foreign_key: :editor_id, class_name: 'Post'
end
# app/models/post_authoring.rb
class PostAuthoring < ActiveRecord::Base
  belongs_to :post_author, class_name: 'User'
  belongs_to :authored_post, class_name: 'Post'
end

And our data model looks like:

users

name

string

created_at

datetime

updated_at

datetime

posts

content

text

editor_id

integer

created_at

datetime

updated_at

datetime

post_authorings

authored_post_id

integer

post_author_id

integer

created_at

datetime

updated_at

datetime

The major thing to note here is that with has-many-through associations, Rails uses the name of the association in the through table to determine which foreign key and table name to reach out to. If it's named anything irregular, you'll use the :source option to specify which association actually points where we'd like to go. You can think of :source as being just like :class_name but for the associations in the "through table".

It may be helpful to illustrate what Rails is doing. In the example above, if you ask for Post.first.authors, Rails sort of "thinks" like this:

  1. You just called for the "authors" association on the Post model

  2. To get "authors", we first need to go through the post_authorings association to get there.

  3. Once we're at the PostAuthoring model, to get to the author, we'll need to use the :belongs_to association and it'll be called post_author. We know this because we used the :source option. If we hadn't used the :source option in the original has-many-through association, we would have been looking for belongs_to :author instead.

  4. Now we've got all the information we need to structure our SQL joins and grab the list of authors for the post.

It sounds a bit wonky but it's just the same logic as before -- if Rails can't tell based on its assumptions which associations or class names or foreign keys you're supposed to use, you need to specify them yourself using :source or :foreign_key or :class_name. It takes some practice but you'll get it. Usually you know something's up if you get error messages of the flavor ActiveRecord::StatementInvalid: SQLite3::SQLException: no such column.

Once you do get things figured out, it can still be helpful to look in your Rails server output to see which joins are being done to build the SQL query. That's a great window into what your associations are doing behind the scenes (because in the end, it's all about figuring out the correct SQL query to run).

Polymorphic associations

We'll cover polymorphism here but, if your head is really spinning from the other concepts, feel free to just skim it. Consider this an "additional topic" instead of a core topic like foreign keys, class names and source.

Polymorphic associations can be a bit of a head scratcher at first and aren't terribly common, but are well suited for their use case. They use a big word to describe a pretty straightforward concept -- what if you have a single model that can belong to a bunch of different types of models? For example, let's say you're building a service like Facebook where users can comment on any of the different types of things posted by other users (like text, pictures, images). How do you make it okay to comment on all these different types of objects using just a single Comment model?

In a plain vanilla situation, the comment would belongs_to a Post or a Picture or a Video (or whatever you're commenting on). You would have a foreign key called something like post_id in your Comments table. Now if we want to be able to comment on multiple types of things, we need to figure out a different way of dealing with the foreign key because a single foreign key could be referencing a post, an image, a video etc and we don't know which one... it's ambiguous. You could just make a different column for each one, e.g. post_id, image_id, video_id, but that is terribly inelegant and hardcoded (imagine if there were 100 different types of posts we want to be able to comment on!). We need to stick with a single foreign key column.

We solve this by storing not just the foreign key id, but also a reference to which type of model it corresponds to. That way, whenever you want to retrieve a comment, by specifying which type of thing it belongs to it is no longer ambiguous what you're asking for. Note that Rails does this for you in the background as long as it knows you're working with a polymorphic association.

We have to call our foreign key something a bit different from the normal case since it's ambiguous which model it's referencing and you can't just use post_id or picture_id. A convention is to come up with an abstract term for what type of action you're doing and use that to name the association. So in this case we're commenting on things and can thus call the foreign key "commentable". You'll see the *able convention used a fair bit. So the migration for that model might look like:

class CreateComments < ActiveRecord::Migration
  def change
    create_table :comments do |t|
      t.string :title
      t.text :content
      t.integer :commentable_id
      t.string :commentable_type
      t.timestamps
    end
  end
end

"Commentable" will be used to refer to the associations as well. You'll need to tell your Comment model that it is actually polymorphic so Rails knows to also check for a commentable_type column when using it. This is done simply:

# app/models/comments.rb
class Comment < ActiveRecord::Base
  belongs_to :commentable, polymorphic: true
end

On the other side of the association, you just treat your comment like any other association (which happens to have a nonstandard name). You just need to specify the association on every model that has_many of it. The only wrinkle is that, because it's using the "commentable" name, you need to specify it in an alias just like you would if any other association had a nonstandard name:

# app/models/post.rb
class Post < ActiveRecord::Base
  has_many :comments, as: :commentable
end

# app/models/picture.rb
class Picture < ActiveRecord::Base
  has_many :comments, as: :commentable
end

Rails does the rest of the work for you. Any time you ask a Picture for all its comments (Picture.first.comments), Rails will return just the comments that belong to that picture without you having to worry about anything else.

Self joins

Often times you have relationships between the same type of model, for instance users who can follow other users. In this case, you need to specify both associations in your User model but name them differently. You will need to specify in your has_many association what the name of the foreign_key will be:

class Employee < ActiveRecord::Base
  has_many :subordinates, class_name: 'Employee', foreign_key: 'manager_id'

  belongs_to :manager, class_name: 'Employee'
end

Handy methods

As mentioned in the intro, associations give you access to some nifty tricks that you might not think of.

Creating association objects

There's a couple of shortcuts for creating new association objects. The first is to call #new or #create on the association to automatically populate the foreign key. For instance, if a User has_many Posts and Post belongs_to a User:

  # Long version:
  > user = User.first
  > post = Post.create(title: "sample", user_id: user.id)

  # Shorter version:
  > user = User.first
  > user.posts.create(title: "sample")

Another nifty trick is to create a new object and then use the shovel operator to add it to the association. This is just one of the ways that associations sometimes act like arrays:

  > user = User.create(name: "foobar")
  > post = Post.new(title: "sample")
  > user.posts << post

This will save the post to the database with that User's foreign key. It's roughly equivalent to calling:

  > post.user_id = user.id
  > post.save

If you really want to, you can actually replace the entire association with a new collection by setting it equal to the new collection:

  > user = User.first
  > post1 = Post.find(1)
  > post2 = Post.find(2)
  > user.posts = [post1, post2]  # posts added to that user's collection

Destroying dependents

If your user has created a bunch of posts and then decides to delete her account, how do you delete all the associated posts? Specify the dependent: :destroy option when first declaring the association:

# app/models/user.rb
class User < ActiveRecord::Base
  has_many :posts, dependent: :destroy
end

This is just the most common among several options to specify for :dependent. It will run the destroy method on all objects that belong to that user when the user is destroyed.

Assignment

  1. Read the Rails Guide on Associations. Start by skimming sections 1 to 2.7 (which you should have already done).

  2. Read from 2.8 to the end of chapter 3.

  3. Skim chapter 4. It contains all the methods that you gain by using various associations. You certainly don't need to memorize the whole list, but poke through it. You'll end up using most of them.

Conclusion

In this lesson we covered some of the more advanced associations material. Associations are all over the place in Rails and are incredibly useful because of all the new methods they give you access to. As long as you pause and think about what Rails is assuming when you set them up, you should be able to modify them to your liking without too much trouble. Practice makes perfect, though, so keep building projects with associations in them and it'll eventually stick.

Additional resources

This section contains helpful links to other content. It isn't required, so consider it supplemental for if you need to dive deeper into something.

Last updated