Huge News!Announcing our $40M Series B led by Abstract Ventures.Learn More
Socket
Sign inDemoInstall
Socket

access_allow

Package Overview
Dependencies
Maintainers
1
Alerts
File Explorer

Advanced tools

Socket logo

Install Socket

Detect and block malicious and high-risk dependencies

Install

access_allow

  • 0.2.0
  • Rubygems
  • Socket score

Version published
Maintainers
1
Created
Source

AccessAllow

Permissions and access control gem for Rails.

Roles, Abilities and the permissions model

Users should be assigned a role where a role is a named grouping of specific permissions (or abilities as we call them). Roles are configured in the application configuration.

Abilities are named permissions that live inside a namespace. These are context dependant. For example we might think of the ability for being able to check out a shopping cart as shopping_cart: :check_out where shopping_cart is the ability namespace for anything to do with the shopping cart and check_out is the specific ability name.

Thus abilities are acquired by user either through their assigned role, or an ability can be directly assign in the database, via the User association permissions.

Role and abilities utility methods

AccessAllow::Roles provides a bunch of utility methods that

  • check if a given role name is for a specific user type or not
  • returns humanized versions of the role names

AccessAllow::Abilities currently provides utility methods to convert between string and hash representations of abilities.

Configuring Roles & their abilities

Schema configured in Configuration and should be configured to create roles with their abilities.

The structure consists of:

<user_type_key>:
  <user_role_key>:
    <ability_namespace_key>:
      <ability_key>: [true/false]

where

  • user_type_key is determined from the model name of the user class (eg User => user)
  • user_role_key is the name of the user role (eg account_owner)
  • ability_namespace_key is the name of the group of abilites (eg product_management)
  • ability_key is the name of the actual ability (eg edit_product) and is set to a boolean to indicate if the ability is available to the specific configuration or not

Note: ability names must be defined in the correct user type, role, namespace key space otherwise the app will raise an exception. This is to prevent accidentally forgetting to define the default permissions of a role around a specific feature.

Setting a user specific ability

User specific abilities are persisted in Permissions where the attribute ability_name stores the ability namespace and name in one combined string. The format is / separated. Eg tag_management/edit_tag. Use AccessAllow::Abilities to convert between string and hash representations of abilities.

The existence of a Permission sets the specific ability in the above described structure of abilities.

The AbilitiesManager handles mixing these assigned abilities into the users specific total ability list.

Note that an ability defined in a Permission must also exist in the role assigned abilities. If it does not then it is ignored. In other words a Permission can only override abilities defined for the role that are set to false. This allows a user to be given a specific ability that normally their role has not got, but does not allow you to assign arbitrary abilities to a user, thus preventing dangerous situations where an ability that say is only for Admins is assigned to a User role.

Manually checking user abilities

The class AccessAllow::Check implements ability check logic. Using this class one can check if a user has a specific ability and optionally raise if not.

You can either build a new instance of the check class and then use #possible? and #possible! of use the class helper methods

  • .call(user, <ability_namespace>: <ability_name>): checks if user has ability_name in ability_namespace. Returns a boolean result
  • .call!(user, <ability_namespace>: <ability_name>): checks if user has ability_name in ability_namespace. Returns true or raises AccessAllow::ViolationError

The methods exposed by Check are useful for checking for abilities in other objects. To define abilities checks around controller actions see the next section.

Controller DSL for specifying requirements and abilities needed to perform actions

Much of the time permissions checks will occur in Controllers. Also many controller actions have specific checks and requirements around the user or other entities related to the controller action. For example, when editing a user's profile, one must check that the user who is trying to execute the update action has the ability (permission) to do it, but also that the user is even from the same company as the user being edited.

As such a DSL exists that can be used in controllers to define sets of required checks and rules around actions that then define what abilities or checks are needed to allow a specific action to execute. The rules can also define what should happen if the checks do not pass, or if no rules match the current situation.

The DSL allows us to define 3 categories of our so called access rules:

Required check rules

Many times we want to specify that certain requirements are required to allow a user to perform a certain action. These requirements maybe certain checks on the user, or they maybe related to their role or abilities.

These checks must all pass to allow the user to continue. They are checked before any other access checks are executed. If the checks do not pass then a 'violation' is returned, which is then handled by the controller accordingly.

These rules consist of a 'check', an optional set of required abilities, and optionally what violation type to raised if the check does not pass.

Action allow rules

Action allow rules are defined to provide specific rules which allow a user to perform a specific controller action.

Note that AccessAllow prevents an action from being executed unless it is explicitly allowed for the given user trying to execute it.

For a user to be allowed to perform a given controller action, there must be a matching "action allow" rule for that action for which the check passes and permissions requirements are met. Any matching rule will allow the user to execute the action. Note if no rules match successfully then the no-match behaviour is executed.

These rules consist of a check, an optional set of required abilities, a set of action names to which the rule applies and optionally a name to alias the check as a "named check" rule (see below).

Named check rule

These are named checks that can then be referenced by name in action logic or in views to say perform some conditional logic. The method provided for checking if any named check rule is valid for the current context is access_allowed?. There is more details below on this.

Say for example you want a user to be approved and have the ability company_profile/edit to edit the company profile, and want to conditionally display a "Edit" button in the view. You could define a named check, say :approved_can_edit (that checks approved and that the user has the ability) and use it in the view to conditionally display the button:

<% if access_allowed? :approved_can_edit %>
  The button...
<% end %>

Note that when you define an "action allow" rule it is automatically also added to these 'named checks' by the action name, for example, if there is action allow rule for :create then we can use access_allowed? :create.

Also note that it is possible to specify a custom 'named check' name for the "action allow" rule (see more below).

No-match behaviour

The DSL also allows us to define what should happen when executing an action and no rule matches the current situation.

What should happen is defined using one of a set of predefined 'violations' which are handled in specific ways. See the discussion below on violations.

Violations

The behaviour when a "required check" or an action has no matching "allow rule" is defined with so called "violation" configuration. These violations are handled in a standardised way by the controller callback that performs the rule checks.

The violation types are:

  • severe: this violation is considered something unusual and is logged. The end user will simply see a 404 page to avoid exposing to them that there is in fact an actionable endpoint at the route they tried to access.
  • hidden: this violation type is considered less severe, but still aims to avoid leaking information to the end user about the actual available routes on the app. If this violation is raised the user will see a 404 page and the violation is logged to the app logs.
  • not_permitted: this violation is used when a user can know that an action and route exists but that they do not have the assigned 'abilities' to perform the action. The end user will see a 403 (forbidden) page and the violation is simply logged to the app logs.
  • redirect: this violation type is used when we want to perform a redirect if the user does not have the necessary permissions. By default it will redirect to root_path but you can use a block to specify the destination path. The block must return a string or other structure that is accepted by redirect_to.

The DSL & defining checks

The methods are as follows:

access_require(check, with:, violation: :severe, &block)

Used to define a "required check rule".

Takes a check name (a symbol or array of symbols) (see details below), an optional violation type (defaults to severe) for when the check does not pass, and a block for when the violation type is redirect and you want to specify custom logic to determine the redirection destination. Also can take an an optional set of abilities (a hash) passed to with: to check against the user.

access_allow(check, with: nil, to:, as: nil)

Used to define an "action allow rule" with optional named check alias.

Takes a check name (a symbol or array of symbols) (see details below), an optional set of abilities (a hash) passed to with: to check against the user, and an optional name (symbol) passed with as: to allow the rule to be used as a "named check". The controller actions the rule applies to is passed to to: (symbol or array of symbols).

access_allow(check, with: nil, as:)

Used to define a "named check rule".

Similar to the "action allow rule" but without the actions. This rule is thus only available to be used as a "named check".

access_no_match(violation, &block)

Used to define the "no match" behaviour, ie what happens when an action is trying to be executed by no access rule matches or passes for the given user and action.

Takes a violation type and optionally a block for when the violation type is redirect and you want to specify custom logic to determine the redirection destination.

Defining abilities needed

Permissions requirements are specified for the rule with with:.

The permissions are defined as a hash containing keys representing the ability namespaces and associated values representing the required abilities.

For example, {tag_management: [:add_new, :edit_existing], product_management: :edit_variants} would mean that the user must have all 3 of the abilities, tag_management: :add_new, tag_management: :edit_existing and product_management: :edit_variants.

Defining Checks & predefined checks

Access rules must specify one or more 'checks' as part of their rule definition.

'Checks' are basically controller methods which return a boolean to determine if the check 'passed' or 'failed'. Checks are normally custom code written for the given context of the feature. Note that checks do not need to perform the abilities checks specified by with:, these are performed by the gem logic for you.

Checks are specified by providing an instance method on the controller named allow_(name)?, where name is the check name, and which returns a boolean.

For example, if defining a check for an action allow rule where the user must be approved on the platform, and have a specific ability assigned to them, then the 'check' part (named say approved_user) is "user must be approved on the platform" part of the rule, and would be defined on the controller as an instance method allow_approved_user?.

There are some predefined 'common' checks, where you do not need to define the allow_(name)? method. These are:

  • :public: anyone, logged in or not
  • :authenticated_user: any logged in user (uses current_user or whatever is set as the current_user_method in the config)

View helper to check permissions of user for conditional view sections

It is also possible to check access rules from inside views using the access_allowed? view helper, which takes a list of "named check" names. If any of those check names passes the method returns true.

Note that check names also include the actions for which rules exists, as described earlier.

    # in controller
    allow_access :admin, to: :new

    # in view
    <% if access_allowed? :new %>
      Only 'admin' users who are allowed to execute action `:new` can see this
    <% end %>

and

    # in controller
    allow_access :my_check, as: named_rule

    # in view
    <% if access_allowed? :named_rule %>
      Only users for whom the `named_rule` check passes can see this
    <% end %>

Example

Consider the following view fragment, and then the controller heirarchy defined below:

tags_controller.rb

class TagsController < AdminController
  # Allow any admin to access the :index and :show actions
  access_allow :admin, to: [:index, :show]
  # Only let admins with the ability `tag_management: :manage` to execute other actions.
  # Also in our view we can use the check name `tag_management` to conditionally add say an "Add new Tag" button
  access_allow :admin, with: {tag_management: :manage}, to: :all, as: :tag_management
  # On the index page we also conditionally show some statistics about Tag usage, but only to admins with the right
  # ability. This is done with the named check `:view_usage_stats`
  access_allow :admin, with: {tag_management: :usage_stats}, as: :view_usage_stats
  # Admins with a special flag called "im_magic" can also access the :magic action
  access_allow :magic_admin, to: :magic
  
  def allow_admin?
    current_user.admin?
  end
  
  def allow_magic_admin?
    current_user.im_magic? && allow_admin?
  end
  
  # ...
end

class AdminController < AuthenticatedController
  # Only admins can access actions on this controller or its sub controllers. Any authenticated user who is not an
  # admin user will generate a severe access violation. They will see a 404 but the violation will be logged.
  access_require :admin, violation: :severe
  # Once we have verified the user is an admin we can 403 them instead of 404 when they try to access a page they
  # dont have permission for. We don't need to hide the existence of the action from them.
  access_no_match :not_permitted
  # ...
end

class AuthenticatedController < ApplicationController
  # Any action requires an authenticated user. The defined behaviour is that if the user trying to access the action
  # is not authenticated they are redirected to the sign-in page.
  access_require :authenticated_user, violation: :redirect do
    sign_in_path
  end

  # ...
end

class ApplicationController < ActionController::Base
  # By default, if no access rules match when executing an action then show the user a 404 to prevent leaking the
  # existence of the end point
  access_no_match :hidden
  # ...
end

tags/index.html.erb

    <p>Tags Index</p>
    <% if access_allowed? :tag_management %>
      <button>Add new tag</button>
    <% end %>
    <% if access_allowed? :view_usage_stats %>
      <div> ... </div>
    <% end %>
    <ul> ... </ul>

Usage

Add this to your ApplicationController

class ApplicationController < ActionController::Base
  include AccessAllow::ControllerAccessDsl
end

Installation

Add this line to your application's Gemfile:

gem "access_allow"

And then execute:

$ bundle

Or install it yourself as:

$ gem install access_allow

Then run the generator to add the initializer

rails g access_allow:install

Contributing

Contribution directions go here.

License

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

FAQs

Package last updated on 08 Mar 2024

Did you know?

Socket

Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.

Install

Related posts

SocketSocket SOC 2 Logo

Product

  • Package Alerts
  • Integrations
  • Docs
  • Pricing
  • FAQ
  • Roadmap
  • Changelog

Packages

npm

Stay in touch

Get open source security insights delivered straight into your inbox.


  • Terms
  • Privacy
  • Security

Made with ⚡️ by Socket Inc