/ingress

Simple role based authorization for Ruby applications

Primary LanguageRubyMIT LicenseMIT

Ingress

A simple role based authorization framework inspired by CanCan (similar syntax) with a nicer interface for defining the permissions for the roles in your system.

The biggest problem I had with CanCan was the fact that it mostly forced you define the permissions for all the roles in one class (really one method). And when the set of permissions in your system grew very large, you had to bend over backwards to allow you to break things down.

In the OO world we're used to being able to break down functionality into multiple smaller classes which we can them compose into a greater whole. This is the main idea behind this gem, keep the nice syntax that CanCan had, but allow composing the main permission object in your system from many smaller classes.

Installation

Add this line to your application's Gemfile:

gem 'ingress'

And then execute:

$ bundle

Or install it yourself as:

$ gem install ingress

Usage

Let's say you have a user object in your system and the user can have multiple roles. Our set of roles will be guest, member, admin.

First we create the main permission object in our system, let's call it UserPermissions:

class UserPermissions < Ingress::Permissions
  def user_role_identifiers
    user.roles.map do |role|
      role.name.to_sym
    end
  end
end

A couple of things of note is that it inherits from Ingress::Permissions, from which it inherits the initializer:

attr_reader :user

def initialize(user)
  @user = user
end

So this object is always instantiated with a user. The second thing to note is that we have to provide a method called user_role_identifiers which needs to return a list of role identifier that this particular user has. Above we make the assumptions that a user has many roles and that a role has a name. So we iterate over the roles collect the symbolized names and return them. This is essentially what ties everything together. We haven't defined any permissions just yet, but we can already do the following:

user_permissions = UserPermissions.new(user)
user_permissions.can?(:do, :stuff) # returns false

So now we have an object that we can instantiate anywhere, and ask it if our user has a particular permission. Let us now define some permissions for a role. We'll start with the guest role. First, let's update our UserPermissions object:

class UserPermissions < Ingress::Permissions
  define_role_permissions :guest, GuestPermissions

  def user_role_identifiers
    user.roles.map do |role|
      role.name.to_sym
    end
  end
end

We've now said that the permission for the role with the guest identifier live in the GuestPermissions class. Let's create it:

class GuestPermissions < Ingress::Permissions
  define_role_permissions do
    can :view, :non_sensitive_info
    can [:create], :session
  end
end

It's pretty self explanatory, the class again inherits from Ingress::Permissions as that's where the simple DSL for defining permissions lives. The thing to note is that we called the class GuestPermissions, but it could be called anything, the permissions we define here are not attached to any role. They only get attached to the role via the define_role_permissions :guest, GuestPermissions line in the UserPermissions class. The syntax for defining permissions is:

can 'action', 'subject'
or
cannot 'action', 'subject'

Similar to CanCan, the action can be any string, symbol or array of strings or symbols. The subject can also be a string or symbol, or it can be a class constant. Let's define permissions for the next role in our system, member which is more complex. Firstly, update our UserPermissions.

class UserPermissions < Ingress::Permissions
  define_role_permissions :guest, GuestPermissions
  define_role_permissions :member, MemberPermissions

  def user_role_identifiers
    user.roles.map do |role|
      role.name.to_sym
    end
  end
end

Simple, next MemberPermissions class:

class MemberPermissions < Ingress::Permissions
  define_role_permissions do
    can [:show, :update, :destroy], :session
    can :accept, :terms
    can [:view, :create], Post
    can [:update, :destroy], Post, if: ->(user, post) do
      user.id == post.user_id
    end
  end
end

It's a little bit more complex, but still fairly self explanatory. As you can see, we have a Post object in our system. So we allow user with a member role to view and create posts, and they can update and destroy posts that they own. So we could do:

user_permissions = UserPermissions.new(user)
user_permissions.can?(:create, Post) # returns true
post = user.posts.first # assume we can get the list of posts form the user object
user_permissions.can?(:update, post) # returns true

The condition lambda always takes two parameters, the user and an object, the object is whatever we supply to the can? method, when we check permissions.

Let's add our admin role:

class UserPermissions < Ingress::Permissions
  define_role_permissions :guest, GuestPermissions
  define_role_permissions :member, MemberPermissions
  define_role_permissions :admin, AdminPermissions

  def user_role_identifiers
    user.roles.map do |role|
      role.name.to_sym
    end
  end
end

And the class:

class AdminPermissions < Ingress::Permissions
  define_role_permissions do
    can "*", "*" # you can also use can_do_anything
  end
end

As you can see both action and subject can be wildcards, so in this case an admin would be able to do anything in the system, i.e. any call to can? will always return true.

So what else can we do? Well let's say we wanted another role called limited_admin which would be similar to admin, but can't destroy comments:

class UserPermissions < Ingress::Permissions
  define_role_permissions :guest, GuestPermissions
  define_role_permissions :member, MemberPermissions
  define_role_permissions :admin, AdminPermissions
  define_role_permissions :limited_admin, LimitedAdminPermissions

  def user_role_identifiers
    user.roles.map do |role|
      role.name.to_sym
    end
  end
end

And the class:

class LimitedAdminPermissions < Ingress::Permissions
  inherits AdminPermissions

  define_role_permissions do
    cannot :destroy, Comment
  end
end

So basically, we can inherit permissions that are defined in other classes, and either switch off some or add others. Let's create some sort of super_member role, which can do everything a member can do, but can also update anything in the system:

class UserPermissions < Ingress::Permissions
  define_role_permissions :guest, GuestPermissions
  define_role_permissions :member, MemberPermissions
  define_role_permissions :admin, AdminPermissions
  define_role_permissions :limited_admin, LimitedAdminPermissions
  define_role_permissions :super_member, SuperMemberPermissions

  def user_role_identifiers
    user.roles.map do |role|
      role.name.to_sym
    end
  end
end

And the class:

class SuperMemberPermissions < Ingress::Permissions
  inherits MemberPermissions

  define_role_permissions do
    can :update, "*"
  end
end

We can inherit permissions, and we use a wildcard subject, to allow a user with the super_member role to be able to update anything. We can even define a common set of permissions which we want multiple roles to share and have the permission class for each of those roles inherit from the common set. Let's say we want a financial_officer role and a reporting_officer role both of which should have the ability to do anything with a Transaction object in our system (for whatever reason):

class UserPermissions < Ingress::Permissions
  define_role_permissions :guest, GuestPermissions
  define_role_permissions :member, MemberPermissions
  define_role_permissions :admin, AdminPermissions
  define_role_permissions :limited_admin, LimitedAdminPermissions
  define_role_permissions :super_member, SuperMemberPermissions
  define_role_permissions :financial_officer, FinancialOfficerPermissions
  define_role_permissions :reporting_officer, ReportingOfficerPermissions

  def user_role_identifiers
    user.roles.map do |role|
      role.name.to_sym
    end
  end
end

And the classes:

class CommonPermissions < Ingress::Permissions
  define_role_permissions do
    can "*", Transaction
  end
end

class FinancialOfficerPermissions < Ingress::Permissions
  inherits CommonPermissions
end

class ReportingOfficerPermissions < Ingress::Permissions
  inherits CommonPermissions
end

Now we wildcard the action, so we can do anything to Transaction objects. And we have to other sets of permission inherit from the CommonPermissions class.

I hope it's relatively clear that it's pretty flexible, you can almost endlessly decompose the permission definitions into smaller classes then combine via inherits and assign the final permission set to a role identifier via define_role_permissions on the main UserPermissions class.

So now the authorization in your system can be defined in a much more OO way, without nasty and complex tricks. And you can still enjoy a nice syntax very similar to CanCan.

This framework has no hooks into Rails (these would be trivial to write if necessary, e.g. you can instantiate the user_permissions object on your ApplicationController and then do the can? checks anywhere you want) and can therefore be used with any web framework, or even outside of the context of a web framework (if such a use case makes sense).

Development

After checking out the repo, run script/setup to install dependencies. Then, run rake spec to run the tests. You can also run script/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 tags, and push the .gem file to rubygems.org.

Contributing

Bug reports and pull requests are welcome on GitHub at https://github.com/skorks/ingress.

License

The gem is available as open source under the terms of the MIT License.