River is a open source workflow system for Django which support on the fly changes on every item in workflow instead of hardcoding states and transitions.

Main goal of developing this framework is to be able to edit any workflow item on the fly. This means, all elements in workflow like states, transitions, user authorizations(permission), group authorization are editable. To do this, all data about the workflow item is persisted into DB. Hence, they can be changed without touching the code and re-deploying your application.

There is ordering aprovments for a transition functionality in django-river. It also provides skipping specific transition of a specific objects.

Online documentation is available at http://django-river.rtfd.org/.

• Python (2.7, 3.3 , 3.4, 3.5)
• PyPy (2, 3)
• Django (1.7, 1.8, 1.9)
• Django 1.7 is not for Python3.5
• Django 1.9 is not for Python3.3 and PyPy3 (Because of Django deprecation)

• Unlimited states
• Multiple destination
• Multiple end-point
• Circular state machines
• Transition authorization
• Skiping or disabling spesific step
• Custom transition hooks

• Install it

  pip install django-river

• Configure settings

  INSTALLED_APPS=[
  ...
  river
  ...
  ]

• In your model class which will be processing in workflow;

from django.db import models
from river.models.fields.state import StateField

class MyModel(models.Model):
    my_state_field = StateField()

my_model=MyModel.objects.get(....)

my_model.proceed(transactioner_user)
my_model.proceed(transactioner_user,next_state=State.objects.get(label='re-opened'))

That's it. Whenever your new model object is saved, it's state field will be initialized according to given meta data about workflow. proceed methods is injected into your model objects. The object will be in next state if the given user is authorized to do that transaction. When there is two destination states available from current state, next_state must be given to the function. If there is only one state can be at, no needs to give it; django-river will detect it.

1. Define your states.
2. Define your state transitions.
3. Define your proceeding metas which contains permissions and groups authorization for transitions. Proceeding order is also given here.

Whenever an object of MyModel is inserted in your system, all its workflow initialization is done by django-river.

pre_transition: fired before any transition occurrence.

post_transition: fired after any transition occurrence.

pre_proceed: fired before any is proceeded. Transition does not have to occur.

post_proceed: fired before any is proceeded occured. Transition does not have to occur.

pre_final: fired before any workflow is completed.

post_final: fired after any workflow is completed.

Handlers are different from django-river signals. These are for spesific object, spesific source_state, spesific destination_state etc. It is fired when the condition is matched.

Before an object is on final state, if the condition is match; means object is suitable, it is fired;

from river.handlers.completed import PreCompletedHandler

def handler(my_object,field,*args,**kwargs):
    do_something_with(object,field)

PreCompletedHandler.register(handler,my_object,'my_state_field')

register method parameter

After an object is on final state, if the condition is match; means object is suitable, it is fired;

from river.handlers.completed import PostCompletedHandler

def handler(my_object,field,*args,**kwargs):
    do_something_with(object,field)

PostCompletedHandler.register(handler,my_object,'my_state_field')

register method parameter

Before any transition occurred, if the condition is match; means object, source_state,destination state are suitable, it is fired;

from river.handlers.transition import PreTransitionHandler

def handler(my_object,field,*args,**kwargs):
    do_something_with(object,field)

PreTransitionHandler.register(handler,my_object,'my_state_field')

register method parameter

After any transition occurred, if the condition is match; means object, source_state,destination state are suitable, it is fired;

from river.handlers.transition import PostTransitionHandler

def handler(my_object,field,*args,**kwargs):
    do_something_with(object,field)

PostTransitionHandler.register(handler,my_object,'my_state_field')

register method parameter

Handlers can be persisted into different sources. This functionality is added for multiprocessing. Now, backends supports multiprocessing can be implemented.

Default backend is MemoryHandlerBackend which does not supports multiprocessing. It can be updated in settings file;

RIVER_HANDLER_BACKEND = {
    'backend':'river.handlers.backends.database.DatabaseHandlerBackend'
}

Indicates states in your state machine.

These are transition between your states. There must be only one initial state which is in a transition as destination state but no source state to make django-river find it on object creation.

These are proceeding meta of transitions that describes which user permission or user group will be allowed to proceed the transition. These are kind of template for proceedings will be created for each object. An order can also be given here for the transition. This means, If you want to order proceeding for a transition, you can define it. Assume s1 and s2 are our states and there is a transition defined between them and we have two proceeding meta on this transition. They shall be forpermission1 and permission2. If you want object available for proceeding; first permission1 and after it is proceeded by permission1, then it is on approval the second permission which is permission2, you can do it with djang-river by defining order in this model.

There are state machines paths which is needed to be proceeded for every particular object. Proceedings are generated on your model object creation by using proceeding meta. This is whole path for the created object. Do not add or edit this model data unless you don't need specific objects editing like skiping, overriding permissions and groups.

• Deprecation - ProceedingTrack is removed. ProceedingTracks were being used to keep any transaction track to handle even circular one. This was a workaround. So, it can be handled with Proceeding now by cloning them if there is circle. ProceedingTracks was just causing confusion. To fix this, ProceedingTrack model and its functions are removed from django-river.
• Improvement - Circular scenario test is added.

• Improvement - Python version 3.5 support is added. (not for Django1.7)
• Improvement - Django version 1.9 support is added. (not for Python3.3 and PyPy3)

• Bug - Migration 0002 and 0003 were not working properly for postgresql (maybe oracle). For these databases, data can not be fixed. Because, django migrates each in a transactional block and schema migration and data migration can not be done in a transactional block. To fix this, data fixing and schema fixing are seperated.
• Improvement - Timeline section is added into documentation.
• Improvement - State slug field is set as slug version of its label if it is not given on saving.

• Bug - After content_type and field are moved into ProceedingMeta model from Transition model in version 0.6.0, finding initial and final states was failing. This is fixed.
• Bug - 0002 migrations was trying to set default slug field of State model. There was a unique problem. It is fixed. 0002 can be migrated now.
• Improvement - The way of finding initial and final states is changed. ProceedingMeta now has parent-child tree structure to present state machine. This tree structure is used to define the way. This requires to migrate 0003. This migration will build the tree of your existed ProceedingMeta data.

• Improvement - content_type and field are moved into ProceedingMeta model from Transition model. This requires to migrate 0002. This migrations will move value of the fields from Transition to ProceedingMeta.
• Improvement - Slug field is added in State. It is unique field to describe state. This requires to migrate 0002. This migration will set the field as slug version of label field value. (Re Opened -> re-opened)
• Improvement - State model now has natural_key as slug field.
• Improvement - Transition model now has natural_key as (source_state_slug , destination_state_slug) fields
• Improvement - ProceedingMeta model now has natural_key as (content_type, field, transition, order) fields
• Improvement - Changelog is added into documentation.

• Bug - Authorization was not working properly when the user has irrelevant permissions and groups. This is fixed.
• Improvement - User permissions are now retreived from registered authentication backends instead of user.user_permissions

• Improvement - Removed unnecessary models.
• Improvement - Migrations are added
• Bug - content_type__0002 migrations cause failing for django1.7. Dependency is removed
• Bug - DatabaseHandlerBacked was trying to access database on django setup. This cause no table in db error for some django commands. This was happening; because there is no db created before some commands are executed; like makemigrations, migrate.

• Improvement - Example scenario diagrams are added into documentation.
• Bug - Migrations was failing because of injected ProceedingTrack relation. Relation is not injected anymore. But property proceeing_track remains. It still returns current one.