state_machine
state_machine adds support for creating state machines for attributes on any
Ruby class.
Resources
API
Bugs
Development
Testing
Source
- git://github.com/pluginaweek/state_machine.git
Mailing List
Description
State machines make it dead-simple to manage the behavior of a class. Too often,
the state of an object is kept by creating multiple boolean attributes and
deciding how to behave based on the values. This can become cumbersome and
difficult to maintain when the complexity of your class starts to increase.
state_machine simplifies this design by introducing the various parts of a real
state machine, including states, events, transitions, and callbacks. However,
the api is designed to be so simple you don't even need to know what a
state machine is :)
Some brief, high-level features include:
- Defining state machines on any Ruby class
- Multiple state machines on a single class
- Namespaced state machines
- before/after/around/failure transition hooks with explicit transition requirements
- Integration with ActiveModel, ActiveRecord, DataMapper, Mongoid, MongoMapper, and Sequel
- State predicates
- State-driven instance / class behavior
- State values of any data type
- Dynamically-generated state values
- Event parallelization
- Attribute-based event transitions
- Path analysis
- Inheritance
- Internationalization
- GraphViz visualization creator
- YARD integration (Ruby 1.9+ only)
- Flexible machine syntax
Examples of the usage patterns for some of the above features are shown below.
You can find much more detailed documentation in the actual API.
Usage
Example
Below is an example of many of the features offered by this plugin, including:
- Initial states
- Namespaced states
- Transition callbacks
- Conditional transitions
- State-driven instance behavior
- Customized state values
- Parallel events
- Path analysis
Class definition:
class Vehicle
attr_accessor :seatbelt_on, :time_used, :auto_shop_busy
state_machine :state, :initial => :parked do
before_transition :parked => any - :parked, :do => :put_on_seatbelt
after_transition :on => :crash, :do => :tow
after_transition :on => :repair, :do => :fix
after_transition any => :parked do |vehicle, transition|
vehicle.seatbelt_on = false
end
after_failure :on => :ignite, :do => :log_start_failure
around_transition do |vehicle, transition, block|
start = Time.now
block.call
vehicle.time_used += Time.now - start
end
event :park do
transition [:idling, :first_gear] => :parked
end
event :ignite do
transition :stalled => same, :parked => :idling
end
event :idle do
transition :first_gear => :idling
end
event :shift_up do
transition :idling => :first_gear, :first_gear => :second_gear, :second_gear => :third_gear
end
event :shift_down do
transition :third_gear => :second_gear, :second_gear => :first_gear
end
event :crash do
transition all - [:parked, :stalled] => :stalled, :if => lambda {|vehicle| !vehicle.passed_inspection?}
end
event :repair do
transition :stalled => :parked, :unless => :auto_shop_busy
transition :stalled => same
end
state :parked do
def speed
0
end
end
state :idling, :first_gear do
def speed
10
end
end
state all - [:parked, :stalled, :idling] do
def moving?
true
end
end
state :parked, :stalled, :idling do
def moving?
false
end
end
end
state_machine :alarm_state, :initial => :active, :namespace => 'alarm' do
event :enable do
transition all => :active
end
event :disable do
transition all => :off
end
state :active, :value => 1
state :off, :value => 0
end
def initialize
@seatbelt_on = false
@time_used = 0
@auto_shop_busy = true
super()
end
def put_on_seatbelt
@seatbelt_on = true
end
def passed_inspection?
false
end
def tow
end
def fix
end
def log_start_failure
end
end
Note the comment made on the initialize
method in the class. In order for
state machine attributes to be properly initialized, super()
must be called.
See StateMachine::MacroMethods
for more information about this.
Using the above class as an example, you can interact with the state machine
like so:
vehicle = Vehicle.new
vehicle.state
vehicle.state_name
vehicle.human_state_name
vehicle.parked?
vehicle.can_ignite?
vehicle.ignite_transition
vehicle.state_events
vehicle.state_transitions
vehicle.speed
vehicle.moving?
vehicle.ignite
vehicle.parked?
vehicle.idling?
vehicle.speed
vehicle
vehicle.shift_up
vehicle.speed
vehicle.moving?
vehicle
vehicle.fire_state_event(:shift_up)
vehicle.speed
vehicle
vehicle.park!
vehicle.state?(:parked)
vehicle.state?(:invalid)
vehicle.alarm_state
vehicle.alarm_state_name
vehicle.can_disable_alarm?
vehicle.disable_alarm
vehicle.alarm_state
vehicle.alarm_state_name
vehicle.can_enable_alarm?
vehicle.alarm_off?
vehicle.alarm_active?
vehicle.fire_events(:shift_down, :enable_alarm)
vehicle.state_name
vehicle.alarm_state_name
vehicle.fire_events!(:ignite, :enable_alarm)
Vehicle.human_state_name(:first_gear)
Vehicle.human_alarm_state_name(:active)
Vehicle.human_state_event_name(:shift_down)
Vehicle.human_alarm_state_event_name(:enable)
Vehicle.human_state_name('first_gear')
Vehicle.human_state_event_name('shift_down')
vehicle.state_paths
vehicle.state_paths.to_states
vehicle.state_paths.events
vehicle.state_paths(:from => :parked, :to => :first_gear)
vehicle.state = "parked"
vehicle.state
vehicle.state_name
Integrations
In addition to being able to define state machines on all Ruby classes, a set of
out-of-the-box integrations are available for some of the more popular Ruby
libraries. These integrations add library-specific behavior, allowing for state
machines to work more tightly with the conventions defined by those libraries.
The integrations currently available include:
- ActiveModel classes
- ActiveRecord models
- DataMapper resources
- Mongoid models
- MongoMapper models
- Sequel models
A brief overview of these integrations is described below.
ActiveModel
The ActiveModel integration is useful for both standalone usage and for providing
the base implementation for ORMs which implement the ActiveModel API. This
integration adds support for validation errors, dirty attribute tracking, and
observers. For example,
class Vehicle
include ActiveModel::Dirty
include ActiveModel::Validations
include ActiveModel::Observing
attr_accessor :state
define_attribute_methods [:state]
state_machine :initial => :parked do
before_transition :parked => any - :parked, :do => :put_on_seatbelt
after_transition any => :parked do |vehicle, transition|
vehicle.seatbelt = 'off'
end
around_transition :benchmark
event :ignite do
transition :parked => :idling
end
state :first_gear, :second_gear do
validates_presence_of :seatbelt_on
end
end
def put_on_seatbelt
...
end
def benchmark
...
yield
...
end
end
class VehicleObserver < ActiveModel::Observer
def before_ignite(vehicle, transition)
end
def after_transition(vehicle, transition)
Audit.log(vehicle, transition)
end
def after_failure_to_transition(vehicle, transition)
Audit.error(vehicle, transition)
end
end
For more information about the various behaviors added for ActiveModel state
machines and how to build new integrations that use ActiveModel, see
StateMachine::Integrations::ActiveModel
.
ActiveRecord
The ActiveRecord integration adds support for database transactions, automatically
saving the record, named scopes, validation errors, and observers. For example,
class Vehicle < ActiveRecord::Base
state_machine :initial => :parked do
before_transition :parked => any - :parked, :do => :put_on_seatbelt
after_transition any => :parked do |vehicle, transition|
vehicle.seatbelt = 'off'
end
around_transition :benchmark
event :ignite do
transition :parked => :idling
end
state :first_gear, :second_gear do
validates_presence_of :seatbelt_on
end
end
def put_on_seatbelt
...
end
def benchmark
...
yield
...
end
end
class VehicleObserver < ActiveRecord::Observer
def before_ignite(vehicle, transition)
end
def after_transition(vehicle, transition)
Audit.log(vehicle, transition)
end
end
For more information about the various behaviors added for ActiveRecord state
machines, see StateMachine::Integrations::ActiveRecord
.
DataMapper
Like the ActiveRecord integration, the DataMapper integration adds support for
database transactions, automatically saving the record, named scopes, Extlib-like
callbacks, validation errors, and observers. For example,
class Vehicle
include DataMapper::Resource
property :id, Serial
property :state, String
state_machine :initial => :parked do
before_transition :parked => any - :parked, :do => :put_on_seatbelt
after_transition any => :parked do |transition|
self.seatbelt = 'off'
end
around_transition :benchmark
event :ignite do
transition :parked => :idling
end
state :first_gear, :second_gear do
validates_presence_of :seatbelt_on
end
end
def put_on_seatbelt
...
end
def benchmark
...
yield
...
end
end
class VehicleObserver
include DataMapper::Observer
observe Vehicle
before_transition :on => :ignite do |transition|
end
after_transition do |transition|
Audit.log(self, transition)
end
around_transition do |transition, block|
block.call
end
after_transition_failure do |transition|
Audit.log(self, transition)
end
end
Note that the DataMapper::Observer integration is optional and only available
when the dm-observer library is installed.
For more information about the various behaviors added for DataMapper state
machines, see StateMachine::Integrations::DataMapper
.
Mongoid
The Mongoid integration adds support for automatically saving the record,
basic scopes, validation errors, and observers. For example,
class Vehicle
include Mongoid::Document
state_machine :initial => :parked do
before_transition :parked => any - :parked, :do => :put_on_seatbelt
after_transition any => :parked do |vehicle, transition|
vehicle.seatbelt = 'off'
end
around_transition :benchmark
event :ignite do
transition :parked => :idling
end
state :first_gear, :second_gear do
validates_presence_of :seatbelt_on
end
end
def put_on_seatbelt
...
end
def benchmark
...
yield
...
end
end
class VehicleObserver < Mongoid::Observer
def before_ignite(vehicle, transition)
end
def after_transition(vehicle, transition)
Audit.log(vehicle, transition)
end
end
For more information about the various behaviors added for Mongoid state
machines, see StateMachine::Integrations::Mongoid
.
MongoMapper
The MongoMapper integration adds support for automatically saving the record,
basic scopes, validation errors and callbacks. For example,
class Vehicle
include MongoMapper::Document
state_machine :initial => :parked do
before_transition :parked => any - :parked, :do => :put_on_seatbelt
after_transition any => :parked do |vehicle, transition|
vehicle.seatbelt = 'off'
end
around_transition :benchmark
event :ignite do
transition :parked => :idling
end
state :first_gear, :second_gear do
validates_presence_of :seatbelt_on
end
end
def put_on_seatbelt
...
end
def benchmark
...
yield
...
end
end
For more information about the various behaviors added for MongoMapper state
machines, see StateMachine::Integrations::MongoMapper
.
Sequel
Like the ActiveRecord integration, the Sequel integration adds support for
database transactions, automatically saving the record, named scopes, validation
errors and callbacks. For example,
class Vehicle < Sequel::Model
plugin :validation_class_methods
state_machine :initial => :parked do
before_transition :parked => any - :parked, :do => :put_on_seatbelt
after_transition any => :parked do |transition|
self.seatbelt = 'off'
end
around_transition :benchmark
event :ignite do
transition :parked => :idling
end
state :first_gear, :second_gear do
validates_presence_of :seatbelt_on
end
end
def put_on_seatbelt
...
end
def benchmark
...
yield
...
end
end
For more information about the various behaviors added for Sequel state
machines, see StateMachine::Integrations::Sequel
.
Additional Topics
Explicit vs. Implicit Event Transitions
Every event defined for a state machine generates an instance method on the
class that allows the event to be explicitly triggered. Most of the examples in
the state_machine documentation use this technique. However, with some types of
integrations, like ActiveRecord, you can also implicitly fire events by
setting a special attribute on the instance.
Suppose you're using the ActiveRecord integration and the following model is
defined:
class Vehicle < ActiveRecord::Base
state_machine :initial => :parked do
event :ignite do
transition :parked => :idling
end
end
end
To trigger the ignite
event, you would typically call the Vehicle#ignite
method like so:
vehicle = Vehicle.create
vehicle.ignite
vehicle.state
This is referred to as an explicit event transition. The same behavior can
also be achieved implicitly by setting the state event attribute and invoking
the action associated with the state machine. For example:
vehicle = Vehicle.create
vehicle.state_event = "ignite"
vehicle.save
vehicle.state
vehicle.state_event
As you can see, the ignite
event was automatically triggered when the save
action was called. This is particularly useful if you want to allow users to
drive the state transitions from a web API.
See each integration's API documentation for more information on the implicit
approach.
Symbols vs. Strings
In all of the examples used throughout the documentation, you'll notice that
states and events are almost always referenced as symbols. This isn't a
requirement, but rather a suggested best practice.
You can very well define your state machine with Strings like so:
class Vehicle
state_machine :initial => 'parked' do
event 'ignite' do
transition 'parked' => 'idling'
end
end
end
You could even use numbers as your state / event names. The important thing
to keep in mind is that the type being used for referencing states / events in
your machine definition must be consistent. If you're using Symbols, then
all states / events must use Symbols. Otherwise you'll encounter the following
error:
class Vehicle
state_machine do
event :ignite do
transition :parked => 'idling'
end
end
end
There is an exception to this rule. The consistency is only required within
the definition itself. However, when the machine's helper methods are called
with input from external sources, such as a web form, state_machine will map
that input to a String / Symbol. For example:
class Vehicle
state_machine :initial => :parked do
event :ignite do
transition :parked => :idling
end
end
end
v = Vehicle.new
v.state?('parked')
v.state?(:parked)
Note that none of this actually has to do with the type of the value that
gets stored. By default, all state values are assumed to be string -- regardless
of whether the state names are symbols or strings. If you want to store states
as symbols instead you'll have to be explicit about it:
class Vehicle
state_machine :initial => :parked do
event :ignite do
transition :parked => :idling
end
states.each do |state|
self.state(state.name, :value => state.name.to_sym)
end
end
end
v = Vehicle.new
v.state?('parked')
v.state?(:parked)
Syntax flexibility
Although state_machine introduces a simplified syntax, it still remains
backwards compatible with previous versions and other state-related libraries by
providing some flexibility around how transitions are defined. See below for an
overview of these syntaxes.
Verbose syntax
In general, it's recommended that state machines use the implicit syntax for
transitions. However, you can be a little more explicit and verbose about
transitions by using the :from
, :except_from
, :to
,
and :except_to
options.
For example, transitions and callbacks can be defined like so:
class Vehicle
state_machine :initial => :parked do
before_transition :from => :parked, :except_to => :parked, :do => :put_on_seatbelt
after_transition :to => :parked do |transition|
self.seatbelt = 'off'
end
event :ignite do
transition :from => :parked, :to => :idling
end
end
end
Transition context
Some flexibility is provided around the context in which transitions can be
defined. In almost all examples throughout the documentation, transitions are
defined within the context of an event. If you prefer to have state machines
defined in the context of a state either out of preference or in order to
easily migrate from a different library, you can do so as shown below:
class Vehicle
state_machine :initial => :parked do
...
state :parked do
transition :to => :idling, :on => [:ignite, :shift_up], :if => :seatbelt_on?
def speed
0
end
end
state :first_gear do
transition :to => :second_gear, :on => :shift_up
def speed
10
end
end
state :idling, :first_gear do
transition :to => :parked, :on => :park
end
end
end
In the above example, there's no need to specify the from
state for each
transition since it's inferred from the context.
You can also define transitions completely outside the context of a particular
state / event. This may be useful in cases where you're building a state
machine from a data store instead of part of the class definition. See the
example below:
class Vehicle
state_machine :initial => :parked do
...
transition :parked => :idling, :on => [:ignite, :shift_up]
transition :first_gear => :second_gear, :second_gear => :third_gear, :on => :shift_up
transition [:idling, :first_gear] => :parked, :on => :park
transition [:idling, :first_gear] => :parked, :on => :park
transition all - [:parked, :stalled] => :stalled, :unless => :auto_shop_busy?
end
end
Notice that in these alternative syntaxes:
- You can continue to configure
:if
and :unless
conditions - You can continue to define
from
states (when in the machine context) using
the all
, any
, and same
helper methods
Static / Dynamic definitions
In most cases, the definition of a state machine is static. That is to say,
the states, events and possible transitions are known ahead of time even though
they may depend on data that's only known at runtime. For example, certain
transitions may only be available depending on an attribute on that object it's
being run on. All of the documentation in this library define static machines
like so:
class Vehicle
state_machine :state, :initial => :parked do
event :park do
transition [:idling, :first_gear] => :parked
end
...
end
end
However, there may be cases where the definition of a state machine is dynamic.
This means that you don't know the possible states or events for a machine until
runtime. For example, you may allow users in your application to manage the
state machine of a project or task in your system. This means that the list of
transitions (and their associated states / events) could be stored externally,
such as in a database. In a case like this, you can define dynamically-generated
state machines like so:
class Vehicle
attr_accessor :state
def initialize(*)
super
machine
end
def transitions
[
{:parked => :idling, :on => :ignite},
{:idling => :first_gear, :first_gear => :second_gear, :on => :shift_up}
]
end
def machine
vehicle = self
@machine ||= Machine.new(vehicle, :initial => :parked, :action => :save) do
vehicle.transitions.each {|attrs| transition(attrs)}
end
end
def save
true
end
end
class Machine
def self.new(object, *args, &block)
machine_class = Class.new
machine = machine_class.state_machine(*args, &block)
attribute = machine.attribute
action = machine.action
machine_class.class_eval do
define_method(:definition) { machine }
define_method(attribute) { object.send(attribute) }
define_method("#{attribute}=") {|value| object.send("#{attribute}=", value) }
define_method(action) { object.send(action) } if action
end
machine_class.new
end
end
vehicle = Vehicle.new
vehicle.state
vehicle.machine.ignite
vehicle.machine.state
vehicle.state
vehicle.machine.state_transitions
vehicle.machine.definition.states.keys
As you can see, state_machine provides enough flexibility for you to be able
to create new machine definitions on the fly based on an external source of
transitions.
Core Extensions
By default, state_machine extends the Ruby core with a state_machine
method on
Class
. All other parts of the library are confined within the StateMachine
namespace. While this isn't wholly necessary, it also doesn't have any performance
impact and makes it truly feel like an extension to the language. This is very
similar to the way that you'll find yaml
, json
, or other libraries adding a
simple method to all objects just by loading the library.
However, if you'd like to avoid having state_machine add this extension to the
Ruby core, you can do so like so:
require 'state_machine/core'
class Vehicle
extend StateMachine::MacroMethods
state_machine do
end
end
If you're using a gem loader like Bundler, you can explicitly indicate which
file to load:
...
gem 'state_machine', :require => 'state_machine/core'
Tools
Generating graphs
This library comes with built-in support for generating di-graphs based on the
events, states, and transitions defined for a state machine using GraphViz.
This requires that both the ruby-graphviz
gem and graphviz library be
installed on the system.
Examples
To generate a graph for a specific file / class:
rake state_machine:draw FILE=vehicle.rb CLASS=Vehicle
To save files to a specific path:
rake state_machine:draw FILE=vehicle.rb CLASS=Vehicle TARGET=files
To customize the image format / orientation:
rake state_machine:draw FILE=vehicle.rb CLASS=Vehicle FORMAT=jpg ORIENTATION=landscape
See http://rdoc.info/github/glejeune/Ruby-Graphviz/Constants for the list of
supported image formats. If resolution is an issue, the svg format may offer
better results.
To generate multiple state machine graphs:
rake state_machine:draw FILE=vehicle.rb,car.rb CLASS=Vehicle,Car
To use human state / event names:
rake state_machine:draw FILE=vehicle.rb CLASS=Vehicle HUMAN_NAMES=true
Note that this will generate a different file for every state machine defined
in the class. The generated files will use an output filename of the format
#{class_name}_#{machine_name}.#{format}
.
For examples of actual images generated using this task, see those under the
examples folder.
Interactive graphs
Jean Bovet's Visual Automata Simulator
is a great tool for "simulating, visualizing and transforming finite state
automata and Turing Machines". It can help in the creation of states and events
for your models. It is cross-platform, written in Java.
Generating documentation
If you use YARD to generate documentation for your projects, state_machine can
be enabled to generate API docs for auto-generated methods from each state machine
definition as well as providing embedded visualizations.
See the generated API documentation under the examples folder to see what the
output looks like.
To enable the YARD integration, you'll need to add state_machine to the list of
YARD's plugins by editing the global YARD config:
~/.yard/config:
load_plugins: true
autoload_plugins:
- state_machine
Once enabled, simply generate your documentation like you normally do.
Note that this only works for Ruby 1.9+.
Web Frameworks
Ruby on Rails
Integrating state_machine into your Ruby on Rails application is straightforward
and provides a few additional features specific to the framework. To get
started, following the steps below.
1. Install the gem
If using Rails 2.x:
...
Rails::Initializer.run do |config|
...
config.gem 'state_machine', :version => '~> 1.0'
...
end
If using Rails 3.x or up:
...
gem 'state_machine'
gem 'ruby-graphviz', :require => 'graphviz'
As usual, run bundle install
to load the gems.
2. Create a model
Create a model with a field to store the state, along with other any other
fields your application requires:
$ rails generate model Vehicle state:string
$ rake db:migrate
3. Configure the state machine
Add the state machine to your model. Following the examples above,
app/models/vehicle.rb might become:
class Vehicle < ActiveRecord::Base
state_machine :initial => :parked do
before_transition :parked => any - :parked, :do => :put_on_seatbelt
...
end
end
Rake tasks
There is a special integration Rake task for generating state machines for
classes used in a Ruby on Rails application. This task will load the application
environment, meaning that it's unnecessary to specify the actual file to load.
For example,
rake state_machine:draw CLASS=Vehicle
If you are using this library as a gem in Rails 2.x, the following must be added
to the end of your application's Rakefile in order for the above task to work:
require 'tasks/state_machine'
Merb
Rake tasks
Like Ruby on Rails, there is a special integration Rake task for generating
state machines for classes used in a Merb application. This task will load the
application environment, meaning that it's unnecessary to specify the actual
files to load.
For example,
rake state_machine:draw CLASS=Vehicle
Testing
To run the core test suite (does not test any of the integrations):
bundle install
bundle exec rake test
To run integration tests:
bundle install
rake appraisal:install
rake appraisal:test
You can also test a specific version:
rake appraisal:active_model-3.0.0 test
rake appraisal:active_record-2.0.0 test
rake appraisal:data_mapper-0.9.4 test
rake appraisal:mongoid-2.0.0 test
rake appraisal:mongo_mapper-0.5.5 test
rake appraisal:sequel-2.8.0 test
Caveats
The following caveats should be noted when using state_machine:
- Overridden event methods won't get invoked when using attribute-based event transitions
- DataMapper: Attribute-based event transitions are disabled when using dm-validations 0.9.4 - 0.9.6
- DataMapper: Transitions cannot persist states when run from after :create / :save callbacks
- JRuby / Rubinius: around_transition callbacks in ORM integrations won't work on JRuby since it doesn't support continuations
- Factory Girl: Dynamic initial states don't work because of the way factory_girl
builds objects. You can work around this in a few ways:
- Use a default state that is common across all objects and rely on events to
determine the actual initial state for your object.
- Assuming you're not using state-driven behavior on initialization, you can
re-initialize states after the fact:
FactoryGirl.define do
factory :vehicle do
after_build {|user| user.send(:initialize_state_machines, :dynamic => :force)}
end
end
class Vehicle < ActiveRecord::Base
...
before_validation :on => :create {|user| user.send(:initialize_state_machines, :dynamic => :force)}
end
Dependencies
Ruby versions officially supported and tested:
- Ruby (MRI) 1.8.6+
- JRuby (1.8, 1.9)
- Rubinius (1.8, 1.9)
ORM versions officially supported and tested:
If graphing state machine: