What is authorization?
Authorization is a process in which we allow or restrict access to certain resources.
Let's assume that task-1 is assigned to Oliver and task-2 is assigned to Eve. The way the code is right now, Oliver can see Eve's tasks and Eve can see Oliver's task. That's not right.
So, let us introduce authorization to our application.
Features
By the end of the chapter, our application will have the following changes.
- Only the tasks assigned to, or created by the logged in user, will be displayed in their dashboard.
- Only the creator or assignee will be allowed to view the details of a Task. For others, the request will result in authorization error.
- Only the creator of a task will be allowed to delete it.
- Only the creator is allowed to change title or to reassign the task.
Technical design
We will be using Pundit gem to do the authorization work.
Pundit helps us in managing role based authorization using policies defined in simple Ruby Classes.
- We will create a new Ruby class named TaskPolicy and define the authorization conditions in it.
- Pundit associates methods in this class with our controller actions. It executes the method corresponding to the action when a request is received. If its result is false, access is denied and an error is thrown. Otherwise the action will be executed as normal.
- This approach doesn't work for our index action. We need it to run for every authenticated user. The requirement is to narrow down the results to only the ones that are relevant to the user.
- We can use the policy scopes of pundit to filter the tasks. It will let us add additional conditions on the method we use to fetch tasks from database. Therefore the result will contain only those tasks that the requesting user is authorized to see.
Let us see this in action.
Pundit gem installation
Add the this line to the Gemfile:
1gem "pundit"
Now, install the gem:
1bundle install
Include Pundit in the ApplicationController class like so:
1class ApplicationController < ActionController::Base 2 # previous code 3 include Pundit 4end
Introducing policies
Policies contain the authorization for an action. Let's create a policy:
1mkdir app/policies 2touch app/policies/task_policy.rb
Open task_policy.rb and add following lines of code:
1class TaskPolicy 2 attr_reader :user, :task 3 4 def initialize(user, task) 5 @user = user 6 @task = task 7 end 8 9 def show? 10 # some condition which returns a boolean value 11 end 12end
Pundit makes the following assumptions about this class:
- The class has the same name as that of model class, only suffixed with the word "Policy". Hence, the name TaskPolicy.
- The first argument is a user. In your controller, Pundit will call the current_user method, which we had defined in ApplicationController, to retrieve what to send into this argument.
- The second argument is that of a model object, whose authorization you want to check.
- The class implements some kind of query method, in this case show?. Usually, this will map to the name of a particular controller action.
Adding policy checks in TaskPolicy
Now let's look at the required code for class TaskPolicy:
1class TaskPolicy 2 attr_reader :user, :task 3 4 def initialize(user, task) 5 @user = user 6 @task = task 7 end 8 9 # The show policy check is invoked when we call `authorize @task` 10 # from the show action of tasks controller. 11 # Here the condition we want to check is that 12 # whether the record's creator is current user or record is assigned to the current user. 13 def show? 14 task.task_owner_id == user.id || task.assigned_user_id == user.id 15 end 16 17 # The condition for edit policy is the same as that of the show. 18 # Hence, we can simply call `show?` inside the edit? policy here. 19 def edit? 20 show? 21 end 22 23 # Only owner is allowed to update a task. 24 def update? 25 task.task_owner_id == user.id 26 end 27 28 # Every user can create a task, hence create? will always returns true. 29 def create? 30 true 31 end 32 33 # Only the user that has created the task, can delete it. 34 def destroy? 35 task.task_owner_id == user.id 36 end 37end
Now, let's add the authorize method to our controller actions. The required code is added as follows:
1class TasksController < ApplicationController 2 before_action :load_task!, only: %i[show update destroy] 3 4 def index 5 tasks = Task.all.as_json(include: { assigned_user: { only: %i[name id] } }) 6 respond_with_json(tasks) 7 end 8 9 def create 10 task = current_user.created_tasks.new(task_params) 11 authorize task 12 task.save! 13 respond_with_success(t("successfully_created", entity: "Task")) 14 end 15 16 def show 17 authorize @task 18 end 19 20 def update 21 authorize @task 22 @task.update!(task_params) 23 respond_with_success(t("successfully_updated", entity: "Task") 24 end 25 26 def destroy 27 authorize @task 28 @task.destroy! 29 respond_with_json 30 end 31 32 private 33 34 def task_params 35 params.require(:task).permit(:title, :assigned_user_id) 36 end 37 38 def load_task! 39 @task = Task.find_by!(slug: params[:slug]) 40 end 41end
We have added the line authorize @task to show, update and destroy actions after initializing our @task instance variable.
The authorize method automatically infers that Task will have a matching TaskPolicy class, and instantiates this class, handing in the current user and the given record (@task in this case).
It then infers from the action name, that it should call the respective action of class TaskPolicy.
For example, the instance created by the authorize @task inside the show action, will call the show? action of Policy class.
From a high level overview, pundit authorize method in TasksController#show action, works something similar to below mentioned code:
1unless TaskPolicy.new(current_user, @task).show? 2 raise Pundit::NotAuthorizedError, "not allowed to show? this #{@task.inspect}" 3end
This raises an exception. But Pundit allows us to rescue the exception with a method of our choice. So, let's add the exception handling for Pundit inside the application_controller.rb file like so:
1class ApplicationController < ActionController::Base 2 # previous code 3 rescue_from Pundit::NotAuthorizedError, with: :handle_authorization_error 4 5 private 6 7 # previous code 8 9 def handle_authorization_error 10 respond_with_error("Access denied. You are not authorized to perform this action.", :forbidden) 11 end 12end
So whenever an action in TaskPolicy returns false it means the authorization has failed. When an authorization fails then we are raising an exception. The exception is rescued by method handle_authorization_error.
Let's say there exists a task with slug of task-slug and a user with name "Oliver". Task with slug of task-slug is neither created by Oliver nor assigned to Oliver. Oliver logs in by entering his email and password. Now Oliver, enters the URL localhost:3000/tasks/task-slug/show.
Since Oliver is not authorized to view the task because of the above assumptions, rather than throwing him an error (the red page), Pundit raises an exception and rescues it by calling the method handle_authorization_error.
Let's define authorization.denied message in our en.yml file:
1en: 2 authorization: 3 denied: "Access denied. You are not authorized to perform this action."
Let's use the authorization.denied translation string. Update the application_controller.rb file like so:
1class ApplicationController < ActionController::Base 2 # previous code 3 4 private 5 6 # previous code 7 8 def handle_authorization_error 9 respond_with_error(t("authorization.denied"), :forbidden) 10 end 11end
Introducing policy scope
If you closely observe, we did not make a change to our index action. As index action returns a collection of records, we need to apply a condition on a collection, and we do that by using Policy Scope.
As we want to have our index view to display only the tasks which are either created by the current_user or assigned to the current_user, we will define a class called a policy scope. This class is nested inside the class TaskPolicy:
1class TaskPolicy 2 3 #------previous code ------- 4 5 #------add new lines here---- 6 class Scope 7 attr_reader :user, :scope 8 9 def initialize(user, scope) 10 @user = user 11 @scope = scope 12 end 13 14 def resolve 15 scope.where(task_owner_id: user.id).or(scope.where(assigned_user_id: user.id)) 16 end 17 end 18end
Pundit makes the following assumptions about this class:
-
The class has the name Scope and is nested under the policy class.
-
The first argument is a user. In your controller, Pundit will call the current_user method to retrieve what to send into this argument.
-
The second argument is a scope which is a collection of records.
-
Instances of this class respond to the method resolve.
-
This method contains the query run on the scope defined and then returns a result which is a collection and can be iterated over.
The corresponding change we make in the index action of Tasks controller is as follows:
1def index 2 #------new line added here------ 3 tasks = TaskPolicy::Scope.new(current_user, Task).resolve 4 #-----end of added line------- 5end
Let's observe what's going on here-
-
Pundit creates an instance of class Scope (nested inside the class TaskPolicy) passing along the current_user and the Task model (our scope in this case) as parameters which get set in the @user and @scope instance variables respectively inside the initialize method.
-
Now this instance calls the method resolve where we run a query on our scope and it returns a collection of tasks which only have the tasks that are either created by or assigned to the current_user.
Now to make it easier Pundit provides syntactic sugar where we can replace the line TaskPolicy::Scope.new(current_user, Task).resolve simply with the syntax policy_scope(Task). It works exactly the same way as described before.
Now, replace your index method with the following lines:
1def index 2 tasks = policy_scope(Task) 3 tasks_with_assigned_user = tasks.as_json(include: { assigned_user: { only: %i[name id] } }) 4 respond_with_json(tasks_with_assigned_user) 5end
Now, we have added authorization to all our actions in task controller. But what if we forgot to call authorize function in every action? Or if someone changed the code and removed it by accident?
It would be a serious security loophole.
To catch such errors in a fail-fast way, Pundit provides us some helper methods.
Add these lines to our TasksController:
1class TasksController < ApplicationController 2 after_action :verify_authorized, except: :index 3 after_action :verify_policy_scoped, only: :index 4 before_action :load_task!, only: %i[show update destroy] 5 # previous code...
In addition to before filters, you can also run filters after an action has been executed. The after filters are registered via after_action.
These Rails filters help us keep the code DRY.
The above code will ensure we get an error if policy_scope or authorize methods were not called during execution of the action. Refer the official documentation of Pundit for a better insight on verify_authorized and verify_policy_scoped methods.
The reason we call verify_authorized and verify_policy_scoped with after_action is because we can only check whether policies have been applied or not, only after the policies have had an opportunity to act i.e. that the controller action has run completely. Thus we can only use the after filter.
You can learn more about how Pundit helps in ensuring policies and scopes are used from the official documentation.
We have successfully implemented authorization in our application.
Now let's commit these changes:
1git add -A 2git commit -m "Added pundit task policy"