Rails applications can end up with models that get way too big, and so far, the
Ruby community response has been Service Objects. But sometimes app/services
can turn into another junk drawer that doesn't help you build and make concepts for your Domain Model.
ActiveRecord::AssociatedObject
takes that head on. Associated Objects are a new domain concept, a context object, that's meant to
help you tease out collaborator objects for your Active Record models.
They're essentially POROs that you associate with an Active Record model to get benefits both in simpler code as well as automatic app/models
organization.
Let's look at an example. Say you have a Post
model that encapsulates a blog post in a Content-Management-System:
class Post < ApplicationRecord
end
You've identified that several things need to happen when a post gets published.
But where does that behavior live; in Post
? That might get messy.
If we put it in a classic Service Object, we've got access to a def call
method and that's it βΒ what if we need other methods that operate on the state? And then having PublishPost
or a similar ad-hoc name in app/services
can pollute that folder over time.
What if we instead identified a Publisher
collaborator object, a Ruby class that handles publishing? What if we required it to be placed within Post::
to automatically help connote the object as belonging to and collaborating with Post
? Then we'd get app/models/post/publisher.rb
which guides naming and gives more organization in your app automatically through that convention β and helps prevent a junk drawer from forming.
This is what Associated Objects are! We'd define it like this:
# app/models/post/publisher.rb
class Post::Publisher < ActiveRecord::AssociatedObject
end
And then you can declare it in Post
:
# app/models/post.rb
class Post < ApplicationRecord
has_object :publisher
end
There isn't anything super special happening yet. Here's essentially what's happening under the hood:
class Post::Publisher
attr_reader :post
def initialize(post) = @post = post
end
class Post < ApplicationRecord
def publisher = (@associated_objects ||= {})[:publisher] ||= Post::Publisher.new(self)
end
Note: due to Ruby's Object Shapes, we use a single @associated_objects
instance variable that's assigned to nil
on Post.new
. This prevents Active Record's from ballooning into many different shapes in Ruby's internals.
We've fixed this so you don't need to care, but this is what's happening.
Tip
has_object
only requires a namespace and an initializer that takes a single argument. The above Post::Publisher
is perfectly valid as an Associated Object β same goes for class Post::Publisher < Data.define(:post); end
.
Tip
You can pass multiple names too: has_object :publisher, :classified, :fortification
. I recommend -[i]er
, -[i]ed
and -ion
as the general naming conventions for your Associated Objects.
See how we're always expecting a link to the model, here post
?
Because of that, you can rely on post
from the associated object:
class Post::Publisher < ActiveRecord::AssociatedObject
def publish
# `transaction` is syntactic sugar for `post.transaction` here.
transaction do
post.update! published: true
post.subscribers.post_published post
# There's also a `record` alias available if you prefer the more general reading version:
# record.update! published: true
# record.subscribers.post_published record
end
end
end
To further help illustrate how your collaborator Associated Objects interact with your domain model, you can forward callbacks.
Say we wanted to have our publisher
automatically publish posts after they're created. Or we need to refresh a publishing after a post has been touched. Or what if we don't want posts to be destroyed if they're published due to HAHA BUSINESS rules?
So has_object
can state this and forward those callbacks onto the Associated Object:
class Post < ActiveRecord::Base
# Passing `true` forwards the same name, e.g. `after_touch`.
has_object :publisher, after_touch: true, after_create_commit: :publish,
before_destroy: :prevent_errant_post_destroy
# The above is the same as writing:
after_create_commit { publisher.publish }
after_touch { publisher.after_touch }
before_destroy { publisher.prevent_errant_post_destroy }
end
class Post::Publisher < ActiveRecord::AssociatedObject
def publish
end
def after_touch
# Respond to the after_touch on the Post.
end
def prevent_errant_post_destroy
# Passed callbacks can throw :abort too, and in this example prevent post.destroy.
throw :abort if haha_business?
end
end
Since has_object
eager-loads the Associated Object class, you can also move
any integrating code into a provided extension
block:
Note
Technically, extension
is just Post.class_eval
but with syntactic sugar.
class Post::Publisher < ActiveRecord::AssociatedObject
extension do
# Here we're within Post and can extend it:
has_many :contracts, dependent: :destroy do
def signed? = all?(&:signed?)
end
def self.with_contracts = includes(:contracts)
after_create_commit :publish_later, if: -> { contracts.signed? }
# An integrating method that operates on `publisher`.
private def publish_later = publisher.publish_later
end
end
This is meant as an alternative to having a wrapping ActiveSupport::Concern
in yet-another file like this:
class Post < ApplicationRecord
include Published
end
# app/models/post/published.rb
module Post::Published
extend ActiveSupport::Concern
included do
has_many :contracts, dependent: :destroy do
def signed? = all?(&:signed?)
end
has_object :publisher
after_create_commit :publish_later, if: -> { contracts.signed? }
end
class_methods do
def with_contracts = includes(:contracts)
end
# An integrating method that operates on `publisher`.
private def publish_later = publisher.publish_later
end
Note
Notice how in the extension
version you don't need to:
- have a naming convention for Concerns and where to place them.
- look up two files to read the feature (the concern and the associated object).
- wrap integrating code in an
included
block. - wrap class methods in a
class_methods
block.
The primary benefit for right now is that by focusing the concept of namespaced Collaborator Objects through Associated Objects, you will start seeing them when you're modelling new features and it'll change how you structure and write your apps.
This is what @natematykiewicz found when they started using the gem (we'll get to ActiveJob::Performs
soon):
We're running
ActiveRecord::AssociatedObject
andActiveJob::Performs
(via the associated object) in 3 spots in production so far. It massively improved how I was architecting a new feature. I put a PR up for review and a coworker loved how organized and easy to follow the large PR was because of those 2 gems. I'm now working on another PR in our app where I'm using them again. I keep seeing use-cases for them now. I love it. Thank you for these gems!Anyone reading this, if you haven't checked them out yet, I highly recommend it.
And about a month later it was still holding up:
Just checking in to say we've added like another 4 associated objects to production since my last message.
ActiveRecord::AssociatedObject
+ActiveJob::Performs
is like a 1-2 punch super power. I'm a bit surprised that this isn't Rails core to be honest. I want to migrate so much of our code over to this. It feels much more organized and sane. Then my app/jobs folder won't have much in it because most jobs will actually be via some associated object's _later method. app/jobs will then basically be cron-type things (deactivate any expired subscriptions).
Here's what @nshki found when they tried it:
Spent some time playing with @kaspth's
ActiveRecord::AssociatedObject
andActiveJob::Performs
and wow! The conventions these gems put in place help simplify a codebase drastically. I particularly loveActiveJob::Performs
βit helped me refactor out allApplicationJob
classes I had and keep important context in the right domain model.
Let's look at testing, then we'll get to passing these POROs to jobs like the quotes mentioned!
Follow the app/models/post.rb
and app/models/post/publisher.rb
naming structure in your tests and add test/models/post/publisher_test.rb
.
Then test it like any other object:
# test/models/post/publisher_test.rb
class Post::PublisherTest < ActiveSupport::TestCase
# You can use Fixtures/FactoryBot to get a `post` and then extract its `publisher`:
setup { @publisher = posts(:one).publisher }
setup { @publisher = FactoryBot.build(:post).publisher }
test "publish updates the post" do
@publisher.publish
assert @publisher.post.reload.published?
end
end
Associated Objects include GlobalID::Identification
and have automatic Active Job serialization support that looks like this:
class Post::Publisher < ActiveRecord::AssociatedObject
class PublishJob < ApplicationJob
def perform(publisher) = publisher.publish
end
def publish_later
PublishJob.perform_later self # We're passing this PORO to the job!
end
def publish
# β¦
end
end
Note
Internally, Active Job serializes Active Records as GlobalIDs. Active Record also includes GlobalID::Identification
, which requires the find
and where(id:)
class methods.
We've added Post::Publisher.find
& Post::Publisher.where(id:)
that calls Post.find(id).publisher
and Post.where(id:).map(&:publisher)
respectively.
This pattern of a job perform
consisting of calling an instance method on a sole domain object is ripe for a convention, here's how to do that.
If you also bundle active_job-performs
in your Gemfile like this:
gem "active_job-performs"
gem "active_record-associated_object"
Every Associated Object (and Active Records too) now has access to the performs
macro, so you can do this:
class Post::Publisher < ActiveRecord::AssociatedObject
performs queue_as: :important
performs :publish
performs :retract
def publish
end
def retract(reason:)
end
end
which spares you writing all this:
class Post::Publisher < ActiveRecord::AssociatedObject
# `performs` without a method defines a general job to share between method jobs.
class Job < ApplicationJob
queue_as :important
end
# Individual method jobs inherit from the `Post::Publisher::Job` defined above.
class PublishJob < Job
# Here's the GlobalID integration again, i.e. we don't have to do `post.publisher`.
def perform(publisher, *, **) = publisher.publish(*, **)
end
class RetractJob < Job
def perform(publisher, *, **) = publisher.retract(*, **)
end
def publish_later(*, **) = PublishJob.perform_later(self, *, **)
def retract_later(*, **) = RetractJob.perform_later(self, *, **)
end
Note: you can also pass more complex configuration like this:
performs :publish, queue_as: :important, discard_on: SomeError do
retry_on TimeoutError, wait: :exponentially_longer
end
See the ActiveJob::Performs
README for more details.
We've got automatic Kredis integration for Associated Objects, so you can use any kredis_*
type just like in Active Record classes:
class Post::Publisher < ActiveRecord::AssociatedObject
kredis_datetime :publish_at # Uses a namespaced "post:publishers:<post_id>:publish_at" key.
end
Note
Under the hood, this reuses the same info we needed for automatic Active Job support. Namely, the Active Record class, here Post
, and its id
.
If you have a namespaced Active Record like this:
# app/models/post/comment.rb
class Post::Comment < ApplicationRecord
belongs_to :post
belongs_to :creator, class_name: "User"
has_object :rating
end
You can define the associated object in the same way it was done for Post::Publisher
above, within the Post::Comment
namespace:
# app/models/post/comment/rating.rb
class Post::Comment::Rating < ActiveRecord::AssociatedObject
def good?
# A `comment` method is generated to access the associated comment. There's also a `record` alias available.
comment.creator.subscriber_of? comment.post.creator
end
end
And then test it in test/models/post/comment/rating_test.rb
:
class Post::Comment::RatingTest < ActiveSupport::TestCase
setup { @rating = posts(:one).comments.first.rating }
setup { @rating = FactoryBot.build(:post_comment).rating }
test "pretty, pretty, pretty, pretty good" do
assert @rating.good?
end
end
We support Active Record models with composite primary keys out of the box.
Just setup the associated objects like the above examples and you've got GlobalID/Active Job and Kredis support automatically.
This gem is relatively tiny and I'm not expecting more significant changes on it, for right now. It's unofficial and not affiliated with Rails core.
Though it's written and maintained by an ex-Rails core person, so I know my way in and out of Rails and how to safely extend it.
Install the gem and add to the application's Gemfile by executing:
$ bundle add active_record-associated_object
If bundler is not being used to manage dependencies, install the gem by executing:
$ gem install active_record-associated_object
After checking out the repo, run bin/setup
to install dependencies. Then, run rake test
to run the tests. You can also run bin/console
for an interactive prompt that will allow you to experiment.
To install this gem onto your local machine, run bundle exec rake install
. To release a new version, update the version number in version.rb
, and then run bundle exec rake release
, which will create a git tag for the version, push git commits and the created tag, and push the .gem
file to rubygems.org.
Bug reports and pull requests are welcome on GitHub at https://github.com/kaspth/active_record-associated_object.
The gem is available as open source under the terms of the MIT License.