rules
is a tiny but powerful app providing object-level permissions to
Django, without requiring a database. At its core, it is a generic framework
for building rule-based systems, similar to decision trees. It can also be
used as a standalone library in other contexts and frameworks.
rules
has got you covered. rules
is:
- Documented, tested, reliable and easy to use.
- Versatile. Decorate callables to build complex graphs of predicates. Predicates can be any type of callable -- simple functions, lambdas, methods, callable class objects, partial functions, decorated functions, anything really.
- A good Django citizen. Seamless integration with Django views, templates and the Admin for testing for object-level permissions.
- Efficient and smart. No need to mess around with a database to figure out whether John really wrote that book.
- Simple. Dive in the code. You'll need 10 minutes to figure out how it works.
- Powerful.
rules
comes complete with advanced features, such as invocation context and storage for arbitrary data, skipping evaluation of predicates under specific conditions, logging of evaluated predicates and more!
- Requirements
- Upgrading from 1.x
- How to install
- Using Rules
- Using Rules with Django
- Advanced features
- Best practices
- API Reference
- Licence
rules
requires Python 2.7/3.4 or newer. It can optionally integrate with
Django, in which case requires Django 1.11 or newer.
Note: At any given moment in time, rules
will maintain support for all
currently supported Django versions, while dropping support for those versions
that reached end-of-life in minor releases. See the Supported Versions
section on Django Project website for the current state and timeline.
- Support Python 2.6 and 3.3, and Django versions before 1.11 has been dropped.
- The
SkipPredicate
exception andskip()
method ofPredicate
, that were used to signify that a predicate should be skipped, have been removed. You may returnNone
from your predicate to achieve this. - The APIs to replace a rule's predicate have been renamed and their
behaviour changed.
replace_rule
andreplace_perm
functions andreplace_rule
method ofRuleSet
have been renamed toset_rule
,set_perm
andRuleSet.set_perm
respectively. The old behaviour was to raise aKeyError
if a rule by the given name did not exist. Since version 2.0 this has changed and you can safely useset_*
to set a rule's predicate without having to ensure the rule exists first.
Using pip:
$ pip install rules
Manually:
$ git clone https://github.com/dfunckt/django-rules.git
$ cd django-rules
$ python setup.py install
Run tests with:
$ ./runtests.sh
You may also want to read Best practices for general advice on how to
use rules
.
Add rules
to INSTALLED_APPS
:
INSTALLED_APPS = (
# ...
'rules',
)
Add the authentication backend:
AUTHENTICATION_BACKENDS = (
'rules.permissions.ObjectPermissionBackend',
'django.contrib.auth.backends.ModelBackend',
)
rules
is based on the idea that you maintain a dict-like object that maps
string keys used as identifiers of some kind, to callables, called
predicates. This dict-like object is actually an instance of RuleSet
and
the predicates are instances of Predicate
.
Let's ignore rule sets for a moment and go ahead and define a predicate. The
easiest way is with the @predicate
decorator:
>>> @rules.predicate
>>> def is_book_author(user, book):
... return book.author == user
...
>>> is_book_author
<Predicate:is_book_author object at 0x10eeaa490>
This predicate will return True
if the book's author is the given user,
False
otherwise.
Predicates can be created from any callable that accepts anything from zero to two positional arguments:
fn(obj, target)
fn(obj)
fn()
This is their generic form. If seen from the perspective of authorization in Django, the equivalent signatures are:
fn(user, obj)
fn(user)
fn()
Predicates can do pretty much anything with the given arguments, but must
always return True
if the condition they check is true, False
otherwise. rules
comes with several predefined predicates that you may
read about later on in API Reference, that are mostly useful when dealing
with authorization in Django.
Let's pretend that we want to let authors edit or delete their books, but not books written by other authors. So, essentially, what determines whether an author can edit or can delete a given book is whether they are its author.
In rules
, such requirements are modelled as rules. A rule is a map of
a unique identifier (eg. "can edit") to a predicate. Rules are grouped
together into a rule set. rules
has two predefined rule sets:
- A default rule set storing shared rules.
- Another rule set storing rules that serve as permissions in a Django context.
So, let's define our first couple of rules, adding them to the shared rule
set. We can use the is_book_author
predicate we defined earlier:
>>> rules.add_rule('can_edit_book', is_book_author)
>>> rules.add_rule('can_delete_book', is_book_author)
Assuming we've got some data, we can now test our rules:
>>> from django.contrib.auth.models import User
>>> from books.models import Book
>>> guidetodjango = Book.objects.get(isbn='978-1-4302-1936-1')
>>> guidetodjango.author
<User: adrian>
>>> adrian = User.objects.get(username='adrian')
>>> rules.test_rule('can_edit_book', adrian, guidetodjango)
True
>>> rules.test_rule('can_delete_book', adrian, guidetodjango)
True
Nice... but not awesome.
Predicates by themselves are not so useful -- not more useful than any other function would be. Predicates, however, can be combined using binary operators to create more complex ones. Predicates support the following operators:
P1 & P2
: Returns a new predicate that returnsTrue
if both predicates returnTrue
, otherwiseFalse
. If P1 returnsFalse
, P2 will not be evaluated.P1 | P2
: Returns a new predicate that returnsTrue
if any of the predicates returnsTrue
, otherwiseFalse
. If P1 returnsTrue
, P2 will not be evaluated.P1 ^ P2
: Returns a new predicate that returnsTrue
if one of the predicates returnsTrue
and the other returnsFalse
, otherwiseFalse
.~P
: Returns a new predicate that returns the negated result of the original predicate.
Suppose the requirement for allowing a user to edit a given book was for them to be either the book's author, or a member of the "editors" group. Allowing users to delete a book should still be determined by whether the user is the book's author.
With rules
that's easy to implement. We'd have to define another
predicate, that would return True
if the given user is a member of the
"editors" group, False
otherwise. The built-in is_group_member
factory
will come in handy:
>>> is_editor = rules.is_group_member('editors')
>>> is_editor
<Predicate:is_group_member:editors object at 0x10eee1350>
We could combine it with the is_book_author
predicate to create a new one
that checks for either condition:
>>> is_book_author_or_editor = is_book_author | is_editor
>>> is_book_author_or_editor
<Predicate:(is_book_author | is_group_member:editors) object at 0x10eee1390>
We can now update our can_edit_book
rule:
>>> rules.set_rule('can_edit_book', is_book_author_or_editor)
>>> rules.test_rule('can_edit_book', adrian, guidetodjango)
True
>>> rules.test_rule('can_delete_book', adrian, guidetodjango)
True
Let's see what happens with another user:
>>> martin = User.objects.get(username='martin')
>>> list(martin.groups.values_list('name', flat=True))
['editors']
>>> rules.test_rule('can_edit_book', martin, guidetodjango)
True
>>> rules.test_rule('can_delete_book', martin, guidetodjango)
False
Awesome.
So far, we've only used the underlying, generic framework for defining and
testing rules. This layer is not at all specific to Django; it may be used in
any context. There's actually no import of anything Django-related in the
whole app (except in the rules.templatetags
module). rules
however can
integrate tightly with Django to provide authorization.
rules
is able to provide object-level permissions in Django. It comes
with an authorization backend and a couple template tags for use in your
templates.
In rules
, permissions are a specialised type of rules. You still define
rules by creating and combining predicates. These rules however, must be added
to a permissions-specific rule set that comes with rules
so that they can
be picked up by the rules
authorization backend.
The convention for naming permissions in Django is app_label.action_object
,
and we like to adhere to that. Let's add rules for the books.change_book
and books.delete_book
permissions:
>>> rules.add_perm('books.change_book', is_book_author | is_editor)
>>> rules.add_perm('books.delete_book', is_book_author)
See the difference in the API? add_perm
adds to a permissions-specific
rule set, whereas add_rule
adds to a default shared rule set. It's
important to know however, that these two rule sets are separate, meaning that
adding a rule in one does not make it available to the other.
Let's go ahead and check whether adrian
has change permission to the
guidetodjango
book:
>>> adrian.has_perm('books.change_book', guidetodjango)
False
When you call the User.has_perm
method, Django asks each backend in
settings.AUTHENTICATION_BACKENDS
whether a user has the given permission
for the object. When queried for object permissions, Django's default
authentication backend always returns False
. rules
comes with an
authorization backend, that is able to provide object-level permissions by
looking into the permissions-specific rule set.
Let's add the rules
authorization backend in settings:
AUTHENTICATION_BACKENDS = (
'rules.permissions.ObjectPermissionBackend',
'django.contrib.auth.backends.ModelBackend',
)
Now, checking again gives adrian
the required permissions:
>>> adrian.has_perm('books.change_book', guidetodjango)
True
>>> adrian.has_perm('books.delete_book', guidetodjango)
True
>>> martin.has_perm('books.change_book', guidetodjango)
True
>>> martin.has_perm('books.delete_book', guidetodjango)
False
rules
comes with a set of view decorators to help you enforce
authorization in your views.
For function-based views you can use the permission_required
decorator:
from django.shortcuts import get_object_or_404
from rules.contrib.views import permission_required
from posts.models import Post
def get_post_by_pk(request, post_id):
return get_object_or_404(Post, pk=post_id)
@permission_required('posts.change_post', fn=get_post_by_pk)
def post_update(request, post_id):
# ...
Usage is straight-forward, but there's one thing in the example above that
stands out and this is the get_post_by_pk
function. This function, given
the current request and all arguments passed to the view, is responsible for
fetching and returning the object to check permissions against -- i.e. the
Post
instance with PK equal to the given post_id
in the example.
This specific use-case is quite common so, to save you some typing, rules
comes with a generic helper function that you can use to do this declaratively.
The example below is equivalent to the one above:
from rules.contrib.views import permission_required, objectgetter
from posts.models import Post
@permission_required('posts.change_post', fn=objectgetter(Post, 'post_id'))
def post_update(request, post_id):
# ...
For more information on the decorator and helper function, refer to the
rules.contrib.views
module.
Django includes a set of access mixins that you can use in your class-based
views to enforce authorization. rules
extends this framework to provide
object-level permissions via a mixin, PermissionRequiredMixin
.
The following example will automatically test for permission against the
instance returned by the view's get_object
method:
from django.views.generic.edit import UpdateView
from rules.contrib.views import PermissionRequiredMixin
from posts.models import Post
class PostUpdate(PermissionRequiredMixin, UpdateView):
model = Post
permission_required = 'posts.change_post'
You can customise the object either by overriding get_object
or
get_permission_object
.
For more information refer to the Django documentation and the
rules.contrib.views
module.
rules
comes with two template tags to allow you to test for rules and
permissions in templates.
Add rules
to your INSTALLED_APPS
:
INSTALLED_APPS = (
# ...
'rules',
)
Then, in your template:
{% load rules %} {% has_perm 'books.change_book' author book as can_edit_book %} {% if can_edit_book %} ... {% endif %} {% test_rule 'has_super_feature' user as has_super_feature %} {% if has_super_feature %} ... {% endif %}
If you've setup rules
to be used with permissions in Django, you're almost
set to also use rules
to authorize any add/change/delete actions in the
Admin. The Admin asks for four different permissions, depending on action:
<app_label>.add_<modelname>
<app_label>.view_<modelname>
<app_label>.change_<modelname>
<app_label>.delete_<modelname>
<app_label>
Note: view permission is new in Django v2.1.
The first four are obvious. The fifth is the required permission for an app
to be displayed in the Admin's "dashboard". Here's some rules for our
imaginary books
app as an example:
>>> rules.add_perm('books', rules.always_allow)
>>> rules.add_perm('books.add_book', is_staff)
>>> rules.add_perm('books.view_book', is_staff | has_secret_access_code)
>>> rules.add_perm('books.change_book', is_staff)
>>> rules.add_perm('books.delete_book', is_staff)
Django Admin does not support object-permissions, in the sense that it will never ask for permission to perform an action on an object, only whether a user is allowed to act on (any) instances of a model.
If you'd like to tell Django whether a user has permissions on a specific
object, you'd have to override the following methods of a model's
ModelAdmin
:
has_view_permission(user, obj=None)
has_change_permission(user, obj=None)
has_delete_permission(user, obj=None)
rules
comes with a custom ModelAdmin
subclass,
rules.contrib.admin.ObjectPermissionsModelAdmin
, that overrides these
methods to pass on the edited model instance to the authorization backends,
thus enabling permissions per object in the Admin:
# books/admin.py
from django.contrib import admin
from rules.contrib.admin import ObjectPermissionsModelAdmin
from .models import Book
class BookAdmin(ObjectPermissionsModelAdmin):
pass
admin.site.register(Book, BookAdmin)
Now this allows you to specify permissions like this:
>>> rules.add_perm('books', rules.always_allow)
>>> rules.add_perm('books.add_book', has_author_profile)
>>> rules.add_perm('books.change_book', is_book_author_or_editor)
>>> rules.add_perm('books.delete_book', is_book_author)
To preserve backwards compatibility, Django will ask for either view or
change permission. For maximum flexibility, rules
behaves subtly
different: rules
will ask for the change permission if and only if no rule
exists for the view permission.
You may create as many rule sets as you need:
>>> features = rules.RuleSet()
And manipulate them by adding, removing, querying and testing rules:
>>> features.rule_exists('has_super_feature')
False
>>> is_special_user = rules.is_group_member('special')
>>> features.add_rule('has_super_feature', is_special_user)
>>> 'has_super_feature' in features
True
>>> features['has_super_feature']
<Predicate:is_group_member:special object at 0x10eeaa500>
>>> features.test_rule('has_super_feature', adrian)
True
>>> features.remove_rule('has_super_feature')
Note however that custom rule sets are not available in Django templates -- you need to provide integration yourself.
A new context is created as a result of invoking Predicate.test()
and is
only valid for the duration of the invocation. A context is a simple dict
that you can use to store arbitrary data, (eg. caching computed values,
setting flags, etc.), that can be used by predicates later on in the chain.
Inside a predicate function it can be used like so:
>>> @predicate
... def mypred(a, b):
... value = compute_expensive_value(a)
... mypred.context['value'] = value
... return True
Other predicates can later use stored values:
>>> @predicate
... def myotherpred(a, b):
... value = myotherpred.context.get('value')
... if value is not None:
... return do_something_with_value(value)
... else:
... return do_something_without_value()
Predicate.context
provides a single args
attribute that contains the
arguments as given to test()
at the beginning of the invocation.
In a predicate's function body, you can refer to the predicate instance itself
by its name, eg. is_book_author
. Passing bind=True
as a keyword
argument to the predicate
decorator will let you refer to the predicate
with self
, which is more convenient. Binding self
is just syntactic
sugar. As a matter of fact, the following two are equivalent:
>>> @predicate
... def is_book_author(user, book):
... if is_book_author.context.args:
... return user == book.author
... return False
>>> @predicate(bind=True)
... def is_book_author(self, user, book):
... if self.context.args:
... return user == book.author
... return False
You may skip evaluation by returning None
from your predicate:
>>> @predicate(bind=True)
... def is_book_author(self, user, book):
... if len(self.context.args) > 1:
... return user == book.author
... else:
... return None
Returning None
signifies that the predicate need not be evaluated, thus
leaving the predicate result up to that point unchanged.
rules
can optionally be configured to log debug information as rules are
evaluated to help with debugging your predicates. Messages are sent at the
DEBUG level to the 'rules'
logger. The following dictConfig configures
a console logger (place this in your project's settings.py if you're using
rules with Django):
LOGGING = {
'version': 1,
'disable_existing_loggers': False,
'handlers': {
'console': {
'level': 'DEBUG',
'class': 'logging.StreamHandler',
},
},
'loggers': {
'rules': {
'handlers': ['console'],
'level': 'DEBUG',
'propagate': True,
},
},
}
When this logger is active each individual predicate will have a log message printed when it is evaluated.
Before you can test for rules, these rules must be registered with a rule set, and for this to happen the modules containing your rule definitions must be imported.
For complex projects with several predicates and rules, it may not be practical to define all your predicates and rules inside one module. It might be best to split them among any sub-components of your project. In a Django context, these sub-components could be the apps for your project.
On the other hand, because importing predicates from all over the place in order to define rules can lead to circular imports and broken hearts, it's best to further split predicates and rules in different modules.
rules
may optionally be configured to autodiscover rules.py
modules in
your apps and import them at startup. To have rules
do so, just edit your
INSTALLED_APPS
setting:
INSTALLED_APPS = (
# replace 'rules' with:
'rules.apps.AutodiscoverRulesConfig',
)
Note: On Python 2, you must also add the following to the top of your
rules.py
file, or you'll get import errors trying to import rules
itself:
from __future__ import absolute_import
The core APIs are accessible from the root rules
module. Django-specific
functionality for the Admin and views is available from rules.contrib
.
You create Predicate
instances by passing in a callable:
>>> def is_book_author(user, book):
... return book.author == user
...
>>> pred = Predicate(is_book_author)
>>> pred
<Predicate:is_book_author object at 0x10eeaa490>
You may optionally provide a different name for the predicate that is used when inspecting it:
>>> pred = Predicate(is_book_author, name='another_name')
>>> pred
<Predicate:another_name object at 0x10eeaa490>
Also, you may optionally provide bind=True
in order to be able to access
the predicate instance with self
:
>>> def is_book_author(self, user, book):
... if self.context.args:
... return user == book.author
... return False
...
>>> pred = Predicate(is_book_author, bind=True)
>>> pred
<Predicate:is_book_author object at 0x10eeaa490>
test(obj=None, target=None)
- Returns the result of calling the passed in callable with zero, one or two positional arguments, depending on how many it accepts.
RuleSet
extends Python's built-in dict type. Therefore, you may create
and use a rule set any way you'd use a dict.
add_rule(name, predicate)
- Adds a predicate to the rule set, assigning it to the given rule name.
Raises
KeyError
if another rule with that name already exists. set_rule(name, predicate)
- Set the rule with the given name, regardless if one already exists.
remove_rule(name)
- Remove the rule with the given name. Raises
KeyError
if a rule with that name does not exist. rule_exists(name)
- Returns
True
if a rule with the given name exists,False
otherwise. test_rule(name, obj=None, target=None)
- Returns the result of calling
predicate.test(obj, target)
wherepredicate
is the predicate for the rule with the given name. ReturnsFalse
if a rule with the given name does not exist.
@predicate
Decorator that creates a predicate out of any callable:
>>> @predicate ... def is_book_author(user, book): ... return book.author == user ... >>> is_book_author <Predicate:is_book_author object at 0x10eeaa490>
Customising the predicate name:
>>> @predicate(name='another_name') ... def is_book_author(user, book): ... return book.author == user ... >>> is_book_author <Predicate:another_name object at 0x10eeaa490>
Binding
self
:>>> @predicate(bind=True) ... def is_book_author(self, user, book): ... if 'user_has_special_flag' in self.context: ... return self.context['user_has_special_flag'] ... return book.author == user
always_allow()
,always_true()
- Always returns
True
. always_deny()
,always_false()
- Always returns
False
. is_authenticated(user)
- Returns the result of calling
user.is_authenticated()
. ReturnsFalse
if the given user does not have anis_authenticated
method. is_superuser(user)
- Returns the result of calling
user.is_superuser
. ReturnsFalse
if the given user does not have anis_superuser
property. is_staff(user)
- Returns the result of calling
user.is_staff
. ReturnsFalse
if the given user does not have anis_staff
property. is_active(user)
- Returns the result of calling
user.is_active
. ReturnsFalse
if the given user does not have anis_active
property. is_group_member(*groups)
- Factory that creates a new predicate that returns
True
if the given user is a member of all the given groups,False
otherwise.
add_rule(name, predicate)
- Adds a rule to the shared rule set. See
RuleSet.add_rule
. set_rule(name, predicate)
- Set the rule with the given name from the shared rule set. See
RuleSet.set_rule
. remove_rule(name)
- Remove a rule from the shared rule set. See
RuleSet.remove_rule
. rule_exists(name)
- Returns whether a rule exists in the shared rule set. See
RuleSet.rule_exists
. test_rule(name, obj=None, target=None)
- Tests the rule with the given name. See
RuleSet.test_rule
.
add_perm(name, predicate)
- Adds a rule to the permissions rule set. See
RuleSet.add_rule
. set_perm(name, predicate)
- Replace a rule from the permissions rule set. See
RuleSet.set_rule
. remove_perm(name)
- Remove a rule from the permissions rule set. See
RuleSet.remove_rule
. perm_exists(name)
- Returns whether a rule exists in the permissions rule set. See
RuleSet.rule_exists
. has_perm(name, user=None, obj=None)
- Tests the rule with the given name. See
RuleSet.test_rule
.
django-rules
is distributed under the MIT licence.
Copyright (c) 2014 Akis Kesoglou
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.