A three-tiered permissions model for KegElements built atop Flask-User
A one-stop shop for all things related to authentication and authorization in a Keg app.
Built on top of Keg and KegElements, KegBouncer offers several features for managing authorization and authentication. KegBouncer allows you to pick and choose which features you want it to handle in you application. It achieves this by providing each feature as a Mixin class which you can optionally mixin to your entities (probably a
The available mixins cover:
- Three-tierd permission system
- Password-based authentication and password history
- Login history
Refer to the sections below on how to use each of these.
Note that each mixin will automatically determine the primary key of your entitiy. However, your entity must have exactly one primary key, and it must be specified as SQLAlchemy declarative class attribute.
In order to use KegBouncer’s authorization features to protect Keg views, you will also need Flask-Login. However, KegBouncer’s models do not require that dependency.
keg_bouncer.mixins.PermissionMixin provides a three-tiered permissions model. It manages four kinds of entities:
- Permissions (for describing actions that can be guarded within the system)
- User groups (for grouping users in a way that best models business needs)
- Permission bundles (for grouping permissions in a way that best models the system)
We call this a “three-tiered” permissions model because a user can be granted permissions in three ways:
- Through permission bundles
- Through user groups
This terminology is designed to distinguish this permissions model from other ones, like RBAC, which permit higherarchies of any depth. Technically, this three-tier model is a special case of RBAC.
Note about the term “role”: While this model is technically a special case of the widely-used Role-based access control (RBAC), we took great pains to avoid the highly ambiguous term “role.”
To add permission facilities to your user entity, inherit the
PermissionMixin like this:
import flask_login # Only necessary if using KegBouncer to protect you views. from sqlalchemy import import Column, Integer, String Base = sqlalchemy.ext.declarative.declarative_base() class User(Base, flask_login.UserMixin, keg_bouncer.model.mixins.PermissionMixin): __tablename__ = 'users' id = Column(Integer, primary_key=True)
Protecting Views and Components
To protect various parts of your application, you can use the tools provided in
In order to take advantage of these tools, your
User entity needs to also mixin
ifblock and check for permissions:
# ... if keg_bouncer.auth.current_user_has_permissions('launch-missiles'): launch_missiles()
Decorate a function:
# ... @keg_bouncer.auth.requires_permissions('launch-missiles') def launch_missiles(target=Enemy()) # ...
# ... class LaunchMissilesView(keg_bouncer.auth.ProtectedBaseView): requires_permission = 'launch-missiles'
KegBouncer uses Alembic to manage migrations and it assumes your app does as well.
To use the migrations that KegBouncer provides, you need to tell Alembic where
to find the revisions. In your
alembic.ini file for your application, adjust
your version_locations setting to include your KegBouncer’s versions
[alembic] version_locations = alembic/versions keg_bouncer:alembic/versions
If you run alembic heads you should now see two heads, one for your application and one for KegBouncer.
$ alembic heads 51ba1b47505e (application) (head) 13d265b97e4d (keg_bouncer) (head)
It is totally fine for the application to have multiple heads, but you will need to upgrade them independently. A better option is to merge the two heads into one. Do that with the alembic merge comand.
$ alembic merge -m "pull KegBouncer into application" 51ba1b 13d265 Generating /path/to/app/alembic/versions/31b094b2844f_pull_keg_bouncer_into_application.py ... done
If you run alembic heads again you will find that there is one head.
$ alembic heads 31b094b2844f (application, keg_bouncer) (head)
Also within this merge revision, you will need to create some linking tables for your
entity (which mixes in keg_bouncer.model.mixins.PermissionMixin).
To add password-based authentication to your entity, you need to dynamically construct a password mixin object and mix it in to your entity.
from keg_bouncer.model import mixins from passlib.context import CryptContext import sqlalchemy as sa crypt_context = CryptContext(schemes=['bcrypt']) # This mixin is optional but allows you to add additional fields to the password history table. class OptionalAdditionalFields(object): another_field = sa.Column(sa.Integer) password_history_mixin = mixins.make_password_mixin( OptionalAdditionalFields, # [optional] Allows you to add more fields to the password # history table via a mixin crypt_context=crypt_context # [optional, but must be provided here or via another means] # Configures the CryptContext for hashing and verifying ) class User(password_history_mixin): default_crypt_context = crypt_context # An alternative way of specifying your CryptContext # Yet another way to specify your CryptContext def get_crypt_context(self): return crypt_context help(User.set_password) # Adds password to password history help(User.verify_password) # Verifies a password against most recent password help(User.is_password_used_previously) # Looks for password in history help(User.password_history_entity) # SQLAlchemy entity defining password history entries User.password_history # SQLAlchemy relationship for past passwords; # sorted in reverse chronological order
Note: If you use
is_password_used_previously or a similar concept, your choice of a hashing algorithm can drastically impact performance since password verification is intentionally slow.
For example, using
bcrypt instead of
sha256_crypt will allow you to verify passwords about twice as quickly. This makes a big difference when you’re sifting through past passwords.
To add login history to your entity, you need to dynamically construct a history mixin object and mix it in to your entity.
from keg_bouncer.model import mixins import sqlalchemy as sa # This mixin is optional but allows you to add additional fields to the login history table. class OptionalAdditionalFields(object): another_field = sa.Column(sa.Integer) login_history_mixin = mixins.make_login_history_mixin( OptionalAdditionalFields, # [optional] Allows you to add more fields to the login history # table via a mixin ) class User(login_history_mixin): pass help(User.login_history_entity) # SQLAlchemy entity defining login history entries User.login_history # SQLAlchemy relationship for past logins; # sorted in reverse chronological order # Example use: def register_login(user): user.login_history.insert(0, user.login_history_entity(is_login_successful=True))
Branches & State
master: our “production” branch
All other branches are feature branches.
requirements directory for the files needed and note:
- You should clone Keg and KegElements and
pip install -e .to get working copies. Since these libraries are new, they will likely change frequently.
- Read the notes in the requirements files if you have problems.
- There is a
build-wheelhouse.pyscript that can be run if new requirements have been added. It always rebuilds libraries in
wheel-only.txtso Git will always show them modified. But, if they haven’t really been changed, you should revert those files so as to not add “static” to the commits.
To quickly setup a virtual environment for development, you can use one of the supplied scripts.
virtualenv is your thing, use
vex is your thing, use
Protect yourself from committing lint by installing the pre-commit hook:
ln -s scripts/pre-commit .git/hooks
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size & hash SHA256 hash help||File type||Python version||Upload date|
|KegBouncer-2.2.3-py2.py3-none-any.whl (20.8 kB) Copy SHA256 hash SHA256||Wheel||py2.py3||Apr 4, 2017|
|KegBouncer-2.2.3.tar.gz (23.4 kB) Copy SHA256 hash SHA256||Source||None||Apr 4, 2017|