Roleback is a simple DSL for writing static RBAC rules for your application. Roleback is not concerned with how you store or enforce your roles, it only cares about how you define them. Storing roles against a user class is easy enough, and there are plenty of gems out there to help you enforce them, like Pundit and CanCanCan.
Add this line to your application's Gemfile:
gem 'roleback'
And then execute:
$ bundle
Or install it yourself as:
$ gem install roleback
Using Roleback is simple. Define your roles in a ruby file, and then load them into your application. For example in Rails, you can create a file loaded during your application load, like config/initializers/roles.rb
:
# config/initializers/roles.rb
Roleback.define do
role :admin do
can :manage
end
end
Roleback defines permissions on by roles. In the example above, we've defined a role called admin
that can manage
anything. Usually permissions are defined with three pieces of information: scope
, resource
and action
.
resource
is the object you want to check permissions against. action
is the action you want to check permissions for. For example, you might want to check read
action, on a blog post
:
# config/initializers/roles.rb
Roleback.define do
role :admin do
resource :post do
can :read
end
end
end
resource
however, always includes 7 more actions: create
, read
, update
, delete
, list
, edit
and new
to make it easier to define permissions for common actions. You can change this behavior using the only
and except
options:
# config/initializers/roles.rb
Roleback.define do
role :admin do
resource :post, only: [:read, :create, :update, :destroy] do
can :read
end
end
end
scope
adds context to your permissions. For example, you might want to grant read
on a post
in the web, but not in other contexts (like an API):
# config/initializers/roles.rb
Roleback.define do
role :admin do
scope :web do
resource :post do
can :read
end
end
end
end
Permissions are granted using can
and denied using cannot
:
# config/initializers/roles.rb
Roleback.define do
role :admin do
scope :web do
resource :post do
can :read
cannot :write
end
end
end
end
By default, all resource
default permissions (create, read, update, delete, list, edit and new) are granted (ie can
).
Roles can inherit from other roles:
# config/initializers/roles.rb
Roleback.define do
role :admin do
can :manage
end
role :editor, inherits_from: :admin do
cannot :destroy
end
end
Roles can also inherit from multiple roles:
# config/initializers/roles.rb
Roleback.define do
role :admin do
can :manage
end
role :author do
can :write
end
role :editor, inherits_from: [:admin, :author] do
cannot :destroy
end
end
While you don't need to define a parent role before the child role, circular dependencies are not allowed, within the same parental line. For example, if :moderator
inherits from :admin
, :admin
cannot inherit from :moderator
, directly or indirectly. However, when inheriting from multiple parents, circular dependencies are allowed, as long as they are not in the same parental line. For example, :editor
can inherit from :admin
and :author
, and :author
can inherit from :editor
, as long as :editor
does not inherit from :author
directly or indirectly.
When it comes to consolidating the rules of inherited roles, Roleback allows repeated rules as long as they don't belong to the same role. For example, it is not allowed to define a rule twice, even with the same outcome:
# config/initializers/roles.rb
Roleback.define do
role :admin do
can :manage
can :manage # <- not allowed
end
end
You can however, define the same rule in different roles, as long as they don't contradict each other:
# config/initializers/roles.rb
Roleback.define do
role :admin do
can :manage
end
role :editor, inherits_from: :admin do
can :manage
end
end
# config/initializers/roles.rb
Roleback.define do
role :admin do
can :manage
end
role :editor, inherits_from: :admin do
cannot :manage # <- not allowed
end
end
Roleback doesn't care how you check permissions, but it does provide a simple API for doing so:
Roleback.can?(:admin, resource: :post, action: :read) # => true
Roleback.can?(:editor, resource: :post, action: :destroy) # => false
After the definition of roles is finished (Roleback.define
), all each role, ends up with the collection of all rules it has plus all the rules it has inherited from other roles. These rules are used to check for permissions. When the can?
method is called with scope
, resource
and action
, can?
will return the outcome of the most specific rule that matches the given scope
, resource
and action
. If no rule matches, can?
will return false
. If you have both can
and cannot
rules for a check, cannot
will take precedence (deny over grant).
If you have a User
class, Roleback will automatically, add a can?
method to it:
user = User.find(1) # user.roles => [:admin]
user.can?(resource: :post, action: :read) # => true
user.can?(resource: :post, action: :destroy) # => false
Your User
class has to have a method called roles
that returns an array of role names as symbols.
The User
class returns an array of roles, then Roleback will check each role for a match and will return true
(grant) when the first role matches. If no role matches, can?
will return false
. This is an important point to remember when using class extension, which basically means if you grant the user multiple rules, it will return true
if any of the rules match, even you have rules that deny access to the same resource and action.
You can change the class to be extended from User
, using user_class
option in define
:
Roleback.define user_class: Admin do
# ...
end
If you don't want to extend your User
class, pass in nil
as the user_class
option:
Roleback.define user_class: nil do
# ...
end
Even though Roleback doesn't impose any opinions on how define your rules (sacrilegious in Rails world, I know), here are some recommendations that might help using it with more ease:
- Although Roleback, support deny permissions (
cannot
), I recommend against using those and always define your rules with grant permissions (can
). This will make it easier to reason about your rules and will make it easier to debug them. - Either map your roles to actual organizational roles (marketing, support, etc), or define them based on their access context (commenter, editor, etc). Don't mix the two. Use multiple inheritance when defining the roles based on access context and use single inheritance when defining them based on organizational roles.
- Define your roles in a single file, and load them during application load. (
config/initializers/roles.rb
in Rails is a good place).
Bug reports and pull requests are welcome on this GitHub repository. PRs are welcome and more likely to be accepted if they include tests.