Django friendly finite state machine support

django-fsm adds simple declarative states management for django models.

If you need parallel task execution, view and background task code reuse over different flows - check my new project django-viewflow:

https://github.com/viewflow/viewflow

Instead of adding some state field to a django model, and managing its values by hand, you could use FSMState field and mark model methods with the transition decorator. Your method could contain the side-effects of the state change.

Nice introduction is available here: https://gist.github.com/Nagyman/9502133

You may also take a look at django-fsm-admin project containing a mixin and template tags to integrate django-fsm state transitions into the django admin.

https://github.com/gadventures/django-fsm-admin

Transition logging support could be achived with help of django-fsm-log package

https://github.com/gizmag/django-fsm-log

FSM really helps to structure the code, especially when a new developer comes to the project. FSM is most effective when you use it for some sequential steps.

Installation

$ pip install django-fsm

Or, for the latest git version

$ pip install -e git://github.com/kmmbvnr/django-fsm.git#egg=django-fsm

The library has full Python 3 support

Usage

Add FSMState field to your model

from django_fsm import FSMField, transition

class BlogPost(models.Model):
    state = FSMField(default='new')

Use the transition decorator to annotate model methods

@transition(field=state, source='new', target='published')
def publish(self):
    """
    This function may contain side-effects,
    like updating caches, notifying users, etc.
    The return value will be discarded.
    """

source parameter accepts a list of states, or an individual state. You can use * for source, to allow switching to target from any state. The field parameter accepts both a string attribute name or an actual field instance.

If calling publish() succeeds without raising an exception, the state field will be changed, but not written to the database.

from django_fsm import can_proceed

def publish_view(request, post_id):
    post = get_object__or_404(BlogPost, pk=post_id)
    if not can_proceed(post.publish):
        raise PermissionDenied

    post.publish()
    post.save()
    return redirect('/')

If some conditions are required to be met before changing the state, use the conditions argument to transition. conditions must be a list of functions taking one argument, the model instance. The function must return either True or False or a value that evaluates to True or False. If all functions return True, all conditions are considered to be met and the transition is allowed to happen. If one of the functions returns False, the transition will not happen. These functions should not have any side effects.

You can use ordinary functions

def can_publish(instance):
    # No publishing after 17 hours
    if datetime.datetime.now().hour > 17:
        return False
    return True

Or model methods

def can_destroy(self):
    return self.is_under_investigation()

Use the conditions like this:

@transition(field=state, source='new', target='published', conditions=[can_publish])
    def publish(self):
    """
    Side effects galore
    """

@transition(field=state, source='*', target='destroyed', conditions=[can_destroy])
    def destroy(self):
    """
    Side effects galore
    """

You could instantiate a field with protected=True option, that prevents direct state field modification.

class BlogPost(models.Model):
    state = FSMField(default='new', protected=True)

model = BlogPost()
model.state = 'invalid' # Raises AttributeError

Note that calling refresh_from_db on a model instance with a protected FSMField will cause an exception.

target

target state parameter could point to a specific state or django_fsm.State implementation

from django_fsm import FSMField, transition, RETURN_VALUE, GET_STATE
@transition(field=state,
            source='*',
            target=RETURN_VALUE('for_moderators', 'published'))
def publish(self, is_public=False):
    return 'need_moderation' if is_public else 'published'

@transition(
    field=state,
    source='for_moderators',
    target=GET_STATE(
        lambda self, allowed: 'published' if allowed else 'rejected',
        states=['published', 'rejected']))
def moderate(self, allowed):
    self.allowed=allowed

`custom` properties

Custom properties can be added by providing a dictionary to the custom keyword on the transition decorator.

@transition(field=state,
            source='*',
            target='onhold',
            custom=dict(verbose='Hold for legal reasons'))
def legal_hold(self):
    """
    Side effects galore
    """

`on_error` state

In case of transition method would raise exception, you can provide specific target state

@transition(field=state, source='new', target='published', on_error='failed')
def publish(self):
   """
   Some exception could happen here
   """

`state_choices`

Instead of passing two elements list choices you could use three elements state_choices, the last element states for string reference to model proxy class.

Base class instance would be dynamically changed to corresponding Proxy class instance, depending on the state. Even for queryset results, you will get Proxy class instances, even if QuerySet executed on base class.

Check the test case for example usage. Or read about implementation internals

Permissions

It is common to have permissions attached to each model transition. django-fsm handles this with permission keyword on the transition decorator. permission accepts a permission string, or callable that expects instance and user arguments and returns True if user can perform the transition.

@transition(field=state, source='*', target='publish',
            permission=lambda instance, user: not user.has_perm('myapp.can_make_mistakes'))
def publish(self):
    pass

@transition(field=state, source='*', target='publish',
            permission='myapp.can_remove_post')
def remove(self):
    pass

You can check permission with has_transition_permission method

from django_fsm import has_transition_perm
def publish_view(request, post_id):
    post = get_object_or_404(BlogPost, pk=post_id)
    if not has_transition_perm(post.publish, request.user):
        raise PermissionDenied

    post.publish()
    post.save()
    return redirect('/')

Model methods

get_all_FIELD_transitions Enumerates all declared transitions

get_available_FIELD_transitions Returns all transitions data available in current state

get_available_user_FIELD_transitions Enumerates all transitions data available in current state for provided user

Foreign Key constraints support

If you store the states in the db table you could use FSMKeyField to ensure Foreign Key database integrity.

In your model :

class DbState(models.Model):
    id = models.CharField(primary_key=True, max_length=50)
    label = models.CharField(max_length=255)

    def __unicode__(self):
        return self.label


class BlogPost(models.Model):
    state = FSMKeyField(DbState, default='new')

    @transition(field=state, source='new', target='published')
    def publish(self):
        pass

In your fixtures/initial_data.json :

[
    {
        "pk": "new",
        "model": "myapp.dbstate",
        "fields": {
            "label": "_NEW_"
        }
    },
    {
        "pk": "published",
        "model": "myapp.dbstate",
        "fields": {
            "label": "_PUBLISHED_"
        }
    }
]

Note : source and target parameters in @transition decorator use pk values of DBState model as names, even if field "real" name is used, without _id postfix, as field parameter.

Integer Field support

You can also use FSMIntegerField. This is handy when you want to use enum style constants.

class BlogPostStateEnum(object):
    NEW = 10
    PUBLISHED = 20
    HIDDEN = 30

class BlogPostWithIntegerField(models.Model):
    state = FSMIntegerField(default=BlogPostStateEnum.NEW)

    @transition(field=state, source=BlogPostStateEnum.NEW, target=BlogPostStateEnum.PUBLISHED)
    def publish(self):
        pass

Signals

django_fsm.signals.pre_transition and django_fsm.signals.post_transition are called before and after allowed transition. No signals on invalid transition are called.

Arguments sent with these signals:

sender The model class.

instance The actual instance being proceed

name Transition name

source Source model state

target Target model state

Optimistic locking

django-fsm provides optimistic locking mixin, to avoid concurrent model state changes. If model state was changed in database django_fsm.ConcurrentTransition exception would be raised on model.save()

from django_fsm import FSMField, ConcurrentTransitionMixin

class BlogPost(ConcurrentTransitionMixin, models.Model):
    state = FSMField(default='new')

For guaranteed protection against race conditions caused by concurrently executed transitions, make sure:

Your transitions do not have any side effects except for changes in the database,
You always run the save() method on the object within django.db.transaction.atomic() block.

Following these recommendations, you can rely on ConcurrentTransitionMixin to cause a rollback of all the changes that have been executed in an inconsistent (out of sync) state, thus practically negating their effect.

Drawing transitions

Renders a graphical overview of your models states transitions

You need pip install graphviz>=0.4 library and add django_fsm to your INSTALLED_APPS:

INSTALLED_APPS = (
    ...
    'django_fsm',
    ...
)

# Create a dot file
$ ./manage.py graph_transitions > transitions.dot

# Create a PNG image file only for specific model
$ ./manage.py graph_transitions -o blog_transitions.png myapp.Blog

Changelog

django-fsm 2.4.0 2016-05-14

graph_transition command now works with multiple FSM's per model
Add ability to set target state from transition return value or callable

marcgibbons/django-fsm