Pundit

Pundit provides a set of helpers which guide you in leveraging regular Ruby classes and object oriented design patterns to build a simple, robust and scaleable authorization system.

Links:

Installation

gem "pundit"

Include Pundit in your application controller:

class ApplicationController < ActionController::Base
  include Pundit
  protect_from_forgery
end

Optionally, you can run the generator, which will set up an application policy with some useful defaults for you:

rails g pundit:install

After generating your application policy, restart the Rails server so that Rails can pick up any classes in the new app/policies/ directory.

Policies

Pundit is focused around the notion of policy classes. We suggest that you put these classes in app/policies. This is a simple example that allows updating a post if the user is an admin, or if the post is unpublished:

class PostPolicy
  attr_reader :user, :post

  def initialize(user, post)
    @user = user
    @post = post
  end

  def update?
    user.admin? or not post.published?
  end
end

As you can see, this is just a plain Ruby class. Pundit makes the following assumptions about this class:

The class has the same name as some kind of model class, only suffixed with the word "Policy".
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 some kind of model object, whose authorization you want to check. This does not need to be an ActiveRecord or even an ActiveModel object, it can be anything really.
The class implements some kind of query method, in this case update?. Usually, this will map to the name of a particular controller action.

That's it really.

Usually you'll want to inherit from the application policy created by the generator, or set up your own base class to inherit from:

class PostPolicy < ApplicationPolicy
  def update?
    user.admin? or not record.published?
  end
end

In the generated ApplicationPolicy, the model object is called record.

Supposing that you have an instance of class Post, Pundit now lets you do this in your controller:

def update
  @post = Post.find(params[:id])
  authorize @post
  if @post.update(post_params)
    redirect_to @post
  else
    render :edit
  end
end

The authorize method automatically infers that Post will have a matching PostPolicy class, and instantiates this class, handing in the current user and the given record. It then infers from the action name, that it should call update? on this instance of the policy. In this case, you can imagine that authorize would have done something like this:

raise "not authorized" unless PostPolicy.new(current_user, @post).update?

You can pass a second argument to authorize if the name of the permission you want to check doesn't match the action name. For example:

def publish
  @post = Post.find(params[:id])
  authorize @post, :update?
  @post.publish!
  redirect_to @post
end

If you don't have an instance for the first argument to authorize, then you can pass the class. For example:

Policy:

class PostPolicy < ApplicationPolicy
  def admin_list?
    user.admin?
  end
end

Controller:

def admin_list
  authorize Post # we don't have a particular post to authorize
  # Rest of controller action
end

authorize returns the object passed to it, so you can chain it like this:

Controller:

def show
  @user = authorize User.find(params[:id])
end

You can easily get a hold of an instance of the policy through the policy method in both the view and controller. This is especially useful for conditionally showing links or buttons in the view:

<% if policy(@post).update? %>
  <%= link_to "Edit post", edit_post_path(@post) %>
<% end %>

Headless policies

Given there is a policy without a corresponding model / ruby class, you can retrieve it by passing a symbol.

# app/policies/dashboard_policy.rb
class DashboardPolicy < Struct.new(:user, :dashboard)
  # ...
end

# In controllers
authorize :dashboard, :show?

# In views
<% if policy(:dashboard).show? %>
  <%= link_to 'Dashboard', dashboard_path %>
<% end %>

Scopes

Often, you will want to have some kind of view listing records which a particular user has access to. When using Pundit, you are expected to define a class called a policy scope. It can look something like this:

class PostPolicy < ApplicationPolicy
  class Scope
    attr_reader :user, :scope

    def initialize(user, scope)
      @user  = user
      @scope = scope
    end

    def resolve
      if user.admin?
        scope.all
      else
        scope.where(published: true)
      end
    end
  end

  def update?
    user.admin? or not post.published?
  end
end

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 of some kind on which to perform some kind of query. It will usually be an ActiveRecord class or a ActiveRecord::Relation, but it could be something else entirely.
Instances of this class respond to the method resolve, which should return some kind of result which can be iterated over. For ActiveRecord classes, this would usually be an ActiveRecord::Relation.

You'll probably want to inherit from the application policy scope generated by the generator, or create your own base class to inherit from:

class PostPolicy < ApplicationPolicy
  class Scope < Scope
    def resolve
      if user.admin?
        scope.all
      else
        scope.where(published: true)
      end
    end
  end

  def update?
    user.admin? or not post.published?
  end
end

You can now use this class from your controller via the policy_scope method:

def index
  @posts = policy_scope(Post)
end

Just as with your policy, this will automatically infer that you want to use the PostPolicy::Scope class, it will instantiate this class and call resolve on the instance. In this case it is a shortcut for doing:

def index
  @posts = PostPolicy::Scope.new(current_user, Post).resolve
end

You can, and are encouraged to, use this method in views:

<% policy_scope(@user.posts).each do |post| %>
  <p><%= link_to post.title, post_path(post) %></p>
<% end %>

Ensuring policies and scopes are used

Pundit adds a method called verify_authorized to your controllers. This method will raise an exception if authorize has not yet been called. You should run this method in an after_action to ensure that you haven't forgotten to authorize the action. For example:

class ApplicationController < ActionController::Base
  include Pundit
  after_action :verify_authorized
end

Likewise, Pundit also adds verify_policy_scoped to your controller. This will raise an exception in the vein of verify_authorized. However, it tracks if policy_scope is used instead of authorize. This is mostly useful for controller actions like index which find collections with a scope and don't authorize individual instances.

class ApplicationController < ActionController::Base
  include Pundit
  after_action :verify_authorized, except: :index
  after_action :verify_policy_scoped, only: :index
end

If you're using verify_authorized in your controllers but need to conditionally bypass verification, you can use skip_authorization. For bypassing verify_policy_scoped, use skip_policy_scope. These are useful in circumstances where you don't want to disable verification for the entire action, but have some cases where you intend to not authorize.

def show
  record = Record.find_by(attribute: "value")
  if record.present?
    authorize record
  else
    skip_authorization
  end
end

If you need to perform some more sophisticated logic or you want to raise a custom exception you can use the two lower level methods pundit_policy_authorized? and pundit_policy_scoped? which return true or false depending on whether authorize or policy_scope have been called, respectively.

Why is this important?

Having a mechanism that ensures authorization happens allows developers to thoroughly test authorization scenarios as units on the policy objects themselves. The benefits include:

request tests are lean and focused on happy path behavior
unit tests execute much faster keeping your test suite speedy

What is this not?

It's not a failsafe for authorization. Imagine a scenario where there exists a controller action that mutates your service, e.g. PATCH /users/:id. Assuming the update action for the UsersController is not authorized, a request would result in a 500 error due to verify_authorized. However the update would already be persisted leaving a vector for updates being applied without proper authorization.

Manually specifying policy classes

Sometimes you might want to explicitly declare which policy to use for a given class, instead of letting Pundit infer it. This can be done like so:

class Post
  def self.policy_class
    PostablePolicy
  end
end

Just plain old Ruby

As you can see, Pundit doesn't do anything you couldn't have easily done yourself. It's a very small library, it just provides a few neat helpers. Together these give you the power of building a well structured, fully working authorization system without using any special DSLs or funky syntax or anything.

Remember that all of the policy and scope classes are just plain Ruby classes, which means you can use the same mechanisms you always use to DRY things up. Encapsulate a set of permissions into a module and include them in multiple policies. Use alias_method to make some permissions behave the same as others. Inherit from a base set of permissions. Use metaprogramming if you really have to.

Generator

Use the supplied generator to generate policies:

rails g pundit:policy post

Closed systems

In many applications, only logged in users are really able to do anything. If you're building such a system, it can be kind of cumbersome to check that the user in a policy isn't nil for every single permission.

We suggest that you define a filter that redirects unauthenticated users to the login page. As a secondary defence, if you've defined an ApplicationPolicy, it might be a good idea to raise an exception if somehow an unauthenticated user got through. This way you can fail more gracefully.

class ApplicationPolicy
  def initialize(user, record)
    raise Pundit::NotAuthorizedError, "must be logged in" unless user
    @user   = user
    @record = record
  end
end

Rescuing a denied Authorization in Rails

Pundit raises a Pundit::NotAuthorizedError you can rescue_from in your ApplicationController. You can customize the user_not_authorized method in every controller.

class ApplicationController < ActionController::Base
  protect_from_forgery
  include Pundit

  rescue_from Pundit::NotAuthorizedError, with: :user_not_authorized

  private

  def user_not_authorized
    flash[:alert] = "You are not authorized to perform this action."
    redirect_to(request.referrer || root_path)
  end
end

Alternatively, you can globally handle Pundit::NotAuthorizedError's by having rails handle them as a 403 error and serving a 403 error page. Add the following to application.rb:

config.action_dispatch.rescue_responses["Pundit::NotAuthorizedError"] = :forbidden

Creating custom error messages

NotAuthorizedErrors provide information on what query (e.g. :create?), what record (e.g. an instance of Post), and what policy (e.g. an instance of PostPolicy) caused the error to be raised.

One way to use these query, record, and policy properties is to connect them with I18n to generate error messages. Here's how you might go about doing that.

class ApplicationController < ActionController::Base
 rescue_from Pundit::NotAuthorizedError, with: :user_not_authorized

 private

 def user_not_authorized(exception)
   policy_name = exception.policy.class.to_s.underscore

   flash[:error] = t "#{policy_name}.#{exception.query}", scope: "pundit", default: :default
   redirect_to(request.referrer || root_path)
 end
end

en:
 pundit:
   default: 'You cannot perform this action.'
   post_policy:
     update?: 'You cannot edit this post!'
     create?: 'You cannot create posts!'

Of course, this is just an example. Pundit is agnostic as to how you implement your error messaging.

Manually retrieving policies and scopes

Sometimes you want to retrieve a policy for a record outside the controller or view. For example when you delegate permissions from one policy to another.

You can easily retrieve policies and scopes like this:

Pundit.policy!(user, post)
Pundit.policy(user, post)

Pundit.policy_scope!(user, Post)
Pundit.policy_scope(user, Post)

The bang methods will raise an exception if the policy does not exist, whereas those without the bang will return nil.

Customize Pundit user

In some cases your controller might not have access to current_user, or your current_user is not the method that should be invoked by Pundit. Simply define a method in your controller called pundit_user.

def pundit_user
  User.find_by_other_means
end

Additional context

Pundit strongly encourages you to model your application in such a way that the only context you need for authorization is a user object and a domain model that you want to check authorization for. If you find yourself needing more context than that, consider whether you are authorizing the right domain model, maybe another domain model (or a wrapper around multiple domain models) can provide the context you need.

Pundit does not allow you to pass additional arguments to policies for precisely this reason.

However, in very rare cases, you might need to authorize based on more context than just the currently authenticated user. Suppose for example that authorization is dependent on IP address in addition to the authenticated user. In that case, one option is to create a special class which wraps up both user and IP and passes it to the policy.

class UserContext
  attr_reader :user, :ip

  def initialize(user, ip)
    @user = user
    @ip   = ip
  end
end

class ApplicationController
  include Pundit

  def pundit_user
    UserContext.new(current_user, request.ip)
  end
end

Strong parameters

In Rails 4 (or Rails 3.2 with the strong_parameters gem), mass-assignment protection is handled in the controller. With Pundit you can control which attributes a user has access to update via your policies. You can set up a permitted_attributes method in your policy like this:

# app/policies/post_policy.rb
class PostPolicy < ApplicationPolicy
  def permitted_attributes
    if user.admin? || user.owner_of?(post)
      [:title, :body, :tag_list]
    else
      [:tag_list]
    end
  end
end

You can now retrieve these attributes from the policy:

# app/controllers/posts_controller.rb
class PostsController < ApplicationController
  def update
    @post = Post.find(params[:id])
    if @post.update_attributes(post_params)
      redirect_to @post
    else
      render :edit
    end
  end

  private

  def post_params
    params.require(:post).permit(policy(@post).permitted_attributes)
  end
end

However, this is a bit cumbersome, so Pundit provides a convenient helper method:

# app/controllers/posts_controller.rb
class PostsController < ApplicationController
  def update
    @post = Post.find(params[:id])
    if @post.update_attributes(permitted_attributes(@post))
      redirect_to @post
    else
      render :edit
    end
  end
end

If you want to permit different attributes based on the current action, you can define a permitted_attributes_for_#{action} method on your policy:

# app/policies/post_policy.rb
class PostPolicy < ApplicationPolicy
  def permitted_attributes_for_create
    [:title, :body]
  end

  def permitted_attributes_for_edit
    [:body]
  end
end

If you have defined an action-specific method on your policy for the current action, the permitted_attributes helper will call it instead of calling permitted_attributes on your controller.

RSpec

Policy Specs

Pundit includes a mini-DSL for writing expressive tests for your policies in RSpec. Require pundit/rspec in your spec_helper.rb:

require "pundit/rspec"

Then put your policy specs in spec/policies, and make them look somewhat like this:

describe PostPolicy do
  subject { described_class }

  permissions :update?, :edit? do
    it "denies access if post is published" do
      expect(subject).not_to permit(User.new(admin: false), Post.new(published: true))
    end

    it "grants access if post is published and user is an admin" do
      expect(subject).to permit(User.new(admin: true), Post.new(published: true))
    end

    it "grants access if post is unpublished" do
      expect(subject).to permit(User.new(admin: false), Post.new(published: false))
    end
  end
end

An alternative approach to Pundit policy specs is scoping them to a user context as outlined in this excellent post and implemented in the third party pundit-matchers gem.

External Resources

License

Licensed under the MIT license, see the separate LICENSE.txt file.

xmjw/pundit