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.
$ 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
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 state parameter could points to the 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 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
"""
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
"""
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
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 user
argument 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('/')
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
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.
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
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
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.
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
- graph_transition command now works with multiple FSM's per model
- Add ability to set target state from transition return value or callable