Controller Validations

Learning Goals

Check the validity of a model in a controller
Render a response with the error messages
Use HTTP status codes to provide additional context

Introduction

Now that we've seen how Active Record can be used to validate our data, let's see how we can use that in our controllers to give our user access to the validation errors, so they can fix their typos or other problems with their request.

To get set up, run:

$ bundle install
$ rails db:migrate db:seed
$ rails s

Manually Checking Validation

Up until this point, our create action has looked something like this:

# app/controllers/birds_controller.rb
def create
  bird = Bird.create(bird_params)
  render json: bird, status: :created
end

Let's add some validation to our Bird model, so that we don't end up with bad data:

# app/models/bird.rb
class Bird < ApplicationRecord
  validates :name, presence: true, uniqueness: true
end

Now, if we try to create a bird using Postman with bad data, we've got a problem!

{
  "species": "Archilochus colubris"
}

Our server still returns a Bird object, but we can clearly see that it wasn't saved successfully:

{
  "id": null,
  "name": null,
  "species": "Archilochus colubris",
  "created_at": null,
  "updated_at": null,
  "likes": 0
}

From this process, we can tell:

Our model validation prevented this bad data from being saved in the database (yay!)
The response doesn't tell us anything about why the data wasn't saved (boo.)

To provide this additional context, we need to update our controller action to change the response based on whether or not the bird was saved successfully.

def create
  bird = Bird.create(bird_params)
  if bird.valid?
    render json: bird, status: :created
  else
    render json: { errors: bird.errors }, status: :unprocessable_entity
  end
end

Now, we get a different response after sending that same request in Postman:

{
  "errors": {
    "name": ["can't be blank"]
  }
}

From the controller, bird.errors will give a serializable object with all the error messages from our Active Record validations.

We also included the status code of 422 Unprocessable Entity, indicating this was a bad request.

We can clean up this controller action by handling the ActiveRecord::RecordInvalid exception class along with create! or update!:

def create
  bird = Bird.create!(bird_params)
  render json: bird, status: :created
rescue ActiveRecord::RecordInvalid => invalid
  render json: { errors: invalid.record.errors }, status: :unprocessable_entity
end

In the rescue block, the invalid variable is an instance of the exception itself. From that invalid variable, we can access the actual Active Record instance with the record method, where we can retrieve its errors.

We can take a similar approach to validation in our update method, since validations will also run when a model is updated:

def update
  bird = find_bird
  bird.update!(bird_params)
  render json: bird
rescue ActiveRecord::RecordInvalid => invalid
  render json: { errors: invalid.record.errors }, status: :unprocessable_entity
end

We could also handle all ActiveRecord::RecordInvalid exceptions in the controller with the rescue_from method. We'll add it along with the rescue_from that we already implemented for ActiveRecord::RecordNotFound:

class BirdsController < ApplicationController
  rescue_from ActiveRecord::RecordNotFound, with: :render_not_found_response
  # added rescue_from
  rescue_from ActiveRecord::RecordInvalid, with: :render_unprocessable_entity_response

  # rest of controller actions...

  private

  def render_unprocessable_entity_response(invalid)
    render json: { errors: invalid.record.errors }, status: :unprocessable_entity
  end

  # rest of private methods...
end

Now, our create and update actions can focus on the happy path:

def create
  # create! exceptions will be handled by the rescue_from ActiveRecord::RecordInvalid code
  bird = Bird.create!(bird_params)
  render json: bird, status: :created
end

def update
  bird = find_bird
  # update! exceptions will be handled by the rescue_from ActiveRecord::RecordInvalid code
  bird.update!(bird_params)
  render json: bird
end

Formatting the Error Response

When we're sending back error messages, we should take care to format the error messages in a way that can be easily displayed by our frontend. Take another look at the current implementation:

render json: { errors: invalid.record.errors }, status: :unprocessable_entity

This will return a JSON object in the body of the response with a key of errors pointing to a nested object where the keys are the invalid attributes, and values are the validation error messages, like this:

{
  "errors": {
    "name": ["can't be blank"],
    "species": ["must be unique"]
  }
}

We could also return a different format by using the #full_messages method to output an array of pre-formatted error messages:

render json: { errors: invalid.record.errors.full_messages }, status: :unprocessable_entity

That would produce a slightly different output:

{
  "errors": ["Name can't be blank", "Species must be unique"]
}

Notice in either case, the key on our JSON object is errors since we are returning a collection of error messages (either an object or an array).

Which format you choose will depend largely on how you plan on using this data on the frontend. It's good to know you have options!

Conclusion

With model validations in place, we can help protect our database against bad data. Active Record validations also help provide error messages to indicate why a certain value wasn't considered valid data. We can access the model's validity and error messages in the controller. By sending this data in the response, we'll be able to provide additional context to our clients about what went wrong with their request so they can fix it.

Check For Understanding

Before you move on, make sure you can answer the following questions:

If we have Active Record validations in place, what is the difference in how validation errors will be handled if we use create vs. create!?
In the following line of code, what is the invalid variable and how can we use it: rescue ActiveRecord::RecordInvalid => invalid?