Class StateMachine::Machine
In: lib/state_machine/machine.rb
Parent: Object

Represents a state machine for a particular attribute. State machines consist of states, events and a set of transitions that define how the state changes after a particular event is fired.

A state machine will not know all of the possible states for an object unless they are referenced somewhere in the state machine definition. As a result, any unused states should be defined with the other_states or state helper.

Actions

When an action is configured for a state machine, it is invoked when an object transitions via an event. The success of the event becomes dependent on the success of the action. If the action is successful, then the transitioned state remains persisted. However, if the action fails (by returning false), the transitioned state will be rolled back.

For example,

  class Vehicle
    attr_accessor :fail, :saving_state

    state_machine :initial => :parked, :action => :save do
      event :ignite do
        transition :parked => :idling
      end

      event :park do
        transition :idling => :parked
      end
    end

    def save
      @saving_state = state
      fail != true
    end
  end

  vehicle = Vehicle.new     # => #<Vehicle:0xb7c27024 @state="parked">
  vehicle.save              # => true
  vehicle.saving_state      # => "parked" # The state was "parked" was save was called

  # Successful event
  vehicle.ignite            # => true
  vehicle.saving_state      # => "idling" # The state was "idling" when save was called
  vehicle.state             # => "idling"

  # Failed event
  vehicle.fail = true
  vehicle.park              # => false
  vehicle.saving_state      # => "parked"
  vehicle.state             # => "idling"

As shown, even though the state is set prior to calling the save action on the object, it will be rolled back to the original state if the action fails. Note that this will also be the case if an exception is raised while calling the action.

Indirect transitions

In addition to the action being run as the result of an event, the action can also be used to run events itself. For example, using the above as an example:

  vehicle = Vehicle.new           # => #<Vehicle:0xb7c27024 @state="parked">

  vehicle.state_event = 'ignite'
  vehicle.save                    # => true
  vehicle.state                   # => "idling"
  vehicle.state_event             # => nil

As can be seen, the save action automatically invokes the event stored in the state_event attribute (:ignite in this case).

One important note about using this technique for running transitions is that if the class in which the state machine is defined also defines the action being invoked (and not a superclass), then it must manually run the StateMachine hook that checks for event attributes.

For example, in ActiveRecord, DataMapper, Mongoid, MongoMapper, and Sequel, the default action (save) is already defined in a base class. As a result, when a state machine is defined in a model / resource, StateMachine can automatically hook into the save action.

On the other hand, the Vehicle class from above defined its own save method (and there is no save method in its superclass). As a result, it must be modified like so:

    def save
      self.class.state_machines.transitions(self, :save).perform do
        @saving_state = state
        fail != true
      end
    end

This will add in the functionality for firing the event stored in the state_event attribute.

Callbacks

Callbacks are supported for hooking before and after every possible transition in the machine. Each callback is invoked in the order in which it was defined. See StateMachine::Machine#before_transition and StateMachine::Machine#after_transition for documentation on how to define new callbacks.

Note that callbacks only get executed within the context of an event. As a result, if a class has an initial state when it‘s created, any callbacks that would normally get executed when the object enters that state will not get triggered.

For example,

  class Vehicle
    state_machine :initial => :parked do
      after_transition all => :parked do
        raise ArgumentError
      end
      ...
    end
  end

  vehicle = Vehicle.new   # => #<Vehicle id: 1, state: "parked">
  vehicle.save            # => true (no exception raised)

If you need callbacks to get triggered when an object is created, this should be done by either:

  • Use a before :save or equivalent hook, or
  • Set an initial state of nil and use the correct event to create the object with the proper state, resulting in callbacks being triggered and the object getting persisted

Canceling callbacks

Callbacks can be canceled by throwing :halt at any point during the callback. For example,

  ...
  throw :halt
  ...

If a before callback halts the chain, the associated transition and all later callbacks are canceled. If an after callback halts the chain, the later callbacks are canceled, but the transition is still successful.

These same rules apply to around callbacks with the exception that any around callback that doesn‘t yield will essentially result in :halt being thrown. Any code executed after the yield will behave in the same way as after callbacks.

Note that if a before callback fails and the bang version of an event was invoked, an exception will be raised instead of returning false. For example,

  class Vehicle
    state_machine :initial => :parked do
      before_transition any => :idling, :do => lambda {|vehicle| throw :halt}
      ...
    end
  end

  vehicle = Vehicle.new
  vehicle.park        # => false
  vehicle.park!       # => StateMachine::InvalidTransition: Cannot transition state via :park from "idling"

Observers

Observers, in the sense of external classes and not Ruby‘s Observable mechanism, can hook into state machines as well. Such observers use the same callback api that‘s used internally.

Below are examples of defining observers for the following state machine:

  class Vehicle
    state_machine do
      event :park do
        transition :idling => :parked
      end
      ...
    end
    ...
  end

Event/Transition behaviors:

  class VehicleObserver
    def self.before_park(vehicle, transition)
      logger.info "#{vehicle} instructed to park... state is: #{transition.from}, state will be: #{transition.to}"
    end

    def self.after_park(vehicle, transition, result)
      logger.info "#{vehicle} instructed to park... state was: #{transition.from}, state is: #{transition.to}"
    end

    def self.before_transition(vehicle, transition)
      logger.info "#{vehicle} instructed to #{transition.event}... #{transition.attribute} is: #{transition.from}, #{transition.attribute} will be: #{transition.to}"
    end

    def self.after_transition(vehicle, transition)
      logger.info "#{vehicle} instructed to #{transition.event}... #{transition.attribute} was: #{transition.from}, #{transition.attribute} is: #{transition.to}"
    end

    def self.around_transition(vehicle, transition)
      logger.info Benchmark.measure { yield }
    end
  end

  Vehicle.state_machine do
    before_transition :on => :park, :do => VehicleObserver.method(:before_park)
    before_transition VehicleObserver.method(:before_transition)

    after_transition :on => :park, :do => VehicleObserver.method(:after_park)
    after_transition VehicleObserver.method(:after_transition)

    around_transition VehicleObserver.method(:around_transition)
  end

One common callback is to record transitions for all models in the system for auditing/debugging purposes. Below is an example of an observer that can easily automate this process for all models:

  class StateMachineObserver
    def self.before_transition(object, transition)
      Audit.log_transition(object.attributes)
    end
  end

  [Vehicle, Switch, Project].each do |klass|
    klass.state_machines.each do |attribute, machine|
      machine.before_transition StateMachineObserver.method(:before_transition)
    end
  end

Additional observer-like behavior may be exposed by the various integrations available. See below for more information on integrations.

Overriding instance / class methods

Hooking in behavior to the generated instance / class methods from the state machine, events, and states is very simple because of the way these methods are generated on the class. Using the class‘s ancestors, the original generated method can be referred to via super. For example,

  class Vehicle
    state_machine do
      event :park do
        ...
      end
    end

    def park(*args)
      logger.info "..."
      super
    end
  end

In the above example, the park instance method that‘s generated on the Vehicle class (by the associated event) is overridden with custom behavior. Once this behavior is complete, the original method from the state machine is invoked by simply calling super.

The same technique can be used for state, state_name, and all other instance and class methods on the Vehicle class.

Integrations

By default, state machines are library-agnostic, meaning that they work on any Ruby class and have no external dependencies. However, there are certain libraries which expose additional behavior that can be taken advantage of by state machines.

This library is built to work out of the box with a few popular Ruby libraries that allow for additional behavior to provide a cleaner and smoother experience. This is especially the case for objects backed by a database that may allow for transactions, persistent storage, search/filters, callbacks, etc.

When a state machine is defined for classes using any of the above libraries, it will try to automatically determine the integration to use (Agnostic, ActiveModel, ActiveRecord, DataMapper, Mongoid, MongoMapper, or Sequel) based on the class definition. To see how each integration affects the machine‘s behavior, refer to all constants defined under the StateMachine::Integrations namespace.

Methods

Included Modules

Assertions EvalHelpers MatcherHelpers StateMachine::InstanceMethods

Attributes

action  [R]  The action to invoke when an object transitions
callbacks  [R]  The callbacks to invoke before/after a transition is performed

Maps :before => callbacks and :after => callbacks

default_messages  [RW] 
events  [R]  The events that trigger transitions. These are sorted, by default, in the order in which they were defined.
name  [R]  The name of the machine, used for scoping methods generated for the machine as a whole (not states or events)
namespace  [R]  An identifier that forces all methods (including state predicates and event methods) to be generated with the value prefixed or suffixed, depending on the context.
owner_class  [RW]  The class that the machine is defined in
states  [R]  A list of all of the states known to this state machine. This will pull states from the following sources:
  • Initial state
  • State behaviors
  • Event transitions (:to, :from, and :except_from options)
  • Transition callbacks (:to, :from, :except_to, and :except_from options)
  • Unreferenced states (using other_states helper)

These are sorted, by default, in the order in which they were referenced.

use_transactions  [R]  Whether the machine will use transactions when firing events

Public Class methods

Draws the state machines defined in the given classes using GraphViz. The given classes must be a comma-delimited string of class names.

Configuration options:

  • :file - A comma-delimited string of files to load that contain the state machine definitions to draw
  • :path - The path to write the graph file to
  • :format - The image format to generate the graph in
  • :font - The name of the font to draw state names in

Attempts to find or create a state machine for the given class. For example,

  StateMachine::Machine.find_or_create(Vehicle)
  StateMachine::Machine.find_or_create(Vehicle, :initial => :parked)
  StateMachine::Machine.find_or_create(Vehicle, :status)
  StateMachine::Machine.find_or_create(Vehicle, :status, :initial => :parked)

If a machine of the given name already exists in one of the class‘s superclasses, then a copy of that machine will be created and stored in the new owner class (the original will remain unchanged).

Creates a new state machine for the given attribute

Public Instance methods

Determines whether an action hook was defined for firing attribute-based event transitions when the configured action gets called.

Creates a callback that will be invoked after a transition failures to be performed so long as the given requirements match the transition.

See before_transition for a description of the possible configurations for defining callbacks. Note however that you cannot define the state requirements in these callbacks. You may only define event requirements.

The callback

Failure callbacks get invoked whenever an event fails to execute. This can happen when no transition is available, a before callback halts execution, or the action associated with this machine fails to succeed. In any of these cases, any failure callback that matches the attempted transition will be run.

For example,

  class Vehicle
    state_machine do
      after_failure do |vehicle, transition|
        logger.error "vehicle #{vehicle} failed to transition on #{transition.event}"
      end

      after_failure :on => :ignite, :do => :log_ignition_failure

      ...
    end
  end

Creates a callback that will be invoked after a transition is performed so long as the given requirements match the transition.

See before_transition for a description of the possible configurations for defining callbacks.

Creates a callback that will be invoked around a transition so long as the given requirements match the transition.

The callback

Around callbacks wrap transitions, executing code both before and after. These callbacks are defined in the exact same manner as before / after callbacks with the exception that the transition must be yielded to in order to finish running it.

If defining around callbacks using blocks, you must yield within the transition by directly calling the block (since yielding is not allowed within blocks).

For example,

  class Vehicle
    state_machine do
      around_transition do |block|
        Benchmark.measure { block.call }
      end

      around_transition do |vehicle, block|
        logger.info "vehicle was #{state}..."
        block.call
        logger.info "...and is now #{state}"
      end

      around_transition do |vehicle, transition, block|
        logger.info "before #{transition.event}: #{vehicle.state}"
        block.call
        logger.info "after #{transition.event}: #{vehicle.state}"
      end
    end
  end

Notice that referencing the block is similar to doing so within an actual method definition in that it is always the last argument.

On the other hand, if you‘re defining around callbacks using method references, you can yield like normal:

  class Vehicle
    state_machine do
      around_transition :benchmark
      ...
    end

    def benchmark
      Benchmark.measure { yield }
    end
  end

See before_transition for a description of the possible configurations for defining callbacks.

Gets the actual name of the attribute on the machine‘s owner class that stores data with the given name.

Creates a callback that will be invoked before a transition is performed so long as the given requirements match the transition.

The callback

Callbacks must be defined as either an argument, in the :do option, or as a block. For example,

  class Vehicle
    state_machine do
      before_transition :set_alarm
      before_transition :set_alarm, all => :parked
      before_transition all => :parked, :do => :set_alarm
      before_transition all => :parked do |vehicle, transition|
        vehicle.set_alarm
      end
      ...
    end
  end

Notice that the first three callbacks are the same in terms of how the methods to invoke are defined. However, using the :do can provide for a more fluid DSL.

In addition, multiple callbacks can be defined like so:

  class Vehicle
    state_machine do
      before_transition :set_alarm, :lock_doors, all => :parked
      before_transition all => :parked, :do => [:set_alarm, :lock_doors]
      before_transition :set_alarm do |vehicle, transition|
        vehicle.lock_doors
      end
    end
  end

Notice that the different ways of configuring methods can be mixed.

State requirements

Callbacks can require that the machine be transitioning from and to specific states. These requirements use a Hash syntax to map beginning states to ending states. For example,

  before_transition :parked => :idling, :idling => :first_gear, :do => :set_alarm

In this case, the set_alarm callback will only be called if the machine is transitioning from parked to idling or from idling to parked.

To help define state requirements, a set of helpers are available for slightly more complex matching:

  • all - Matches every state/event in the machine
  • all - [:parked, :idling, …] - Matches every state/event except those specified
  • any - An alias for all (matches every state/event in the machine)
  • same - Matches the same state being transitioned from

See StateMachine::MatcherHelpers for more information.

Examples:

  before_transition :parked => [:idling, :first_gear], :do => ...     # Matches from parked to idling or first_gear
  before_transition all - [:parked, :idling] => :idling, :do => ...   # Matches from every state except parked and idling to idling
  before_transition all => :parked, :do => ...                        # Matches all states to parked
  before_transition any => same, :do => ...                           # Matches every loopback

Event requirements

In addition to state requirements, an event requirement can be defined so that the callback is only invoked on specific events using the on option. This can also use the same matcher helpers as the state requirements.

Examples:

  before_transition :on => :ignite, :do => ...                        # Matches only on ignite
  before_transition :on => all - :ignite, :do => ...                  # Matches on every event except ignite
  before_transition :parked => :idling, :on => :ignite, :do => ...    # Matches from parked to idling on ignite

Verbose Requirements

Requirements can also be defined using verbose options rather than the implicit Hash syntax and helper methods described above.

Configuration options:

  • :from - One or more states being transitioned from. If none are specified, then all states will match.
  • :to - One or more states being transitioned to. If none are specified, then all states will match.
  • :on - One or more events that fired the transition. If none are specified, then all events will match.
  • :except_from - One or more states not being transitioned from
  • :except_to - One more states not being transitioned to
  • :except_on - One or more events that *did not* fire the transition

Examples:

  before_transition :from => :ignite, :to => :idling, :on => :park, :do => ...
  before_transition :except_from => :ignite, :except_to => :idling, :except_on => :park, :do => ...

Conditions

In addition to the state/event requirements, a condition can also be defined to help determine whether the callback should be invoked.

Configuration options:

  • :if - A method, proc or string to call to determine if the callback should occur (e.g. :if => :allow_callbacks, or :if => lambda {|user| user.signup_step > 2}). The method, proc or string should return or evaluate to a true or false value.
  • :unless - A method, proc or string to call to determine if the callback should not occur (e.g. :unless => :skip_callbacks, or :unless => lambda {|user| user.signup_step <= 2}). The method, proc or string should return or evaluate to a true or false value.

Examples:

  before_transition :parked => :idling, :if => :moving?, :do => ...
  before_transition :on => :ignite, :unless => :seatbelt_on?, :do => ...

Accessing the transition

In addition to passing the object being transitioned, the actual transition describing the context (e.g. event, from, to) can be accessed as well. This additional argument is only passed if the callback allows for it.

For example,

  class Vehicle
    # Only specifies one parameter (the object being transitioned)
    before_transition all => :parked do |vehicle|
      vehicle.set_alarm
    end

    # Specifies 2 parameters (object being transitioned and actual transition)
    before_transition all => :parked do |vehicle, transition|
      vehicle.set_alarm(transition)
    end
  end

Note that the object in the callback will only be passed in as an argument if callbacks are configured to not be bound to the object involved. This is the default and may change on a per-integration basis.

See StateMachine::Transition for more information about the attributes available on the transition.

Examples

Below is an example of a class with one state machine and various types of before transitions defined for it:

  class Vehicle
    state_machine do
      # Before all transitions
      before_transition :update_dashboard

      # Before specific transition:
      before_transition [:first_gear, :idling] => :parked, :on => :park, :do => :take_off_seatbelt

      # With conditional callback:
      before_transition all => :parked, :do => :take_off_seatbelt, :if => :seatbelt_on?

      # Using helpers:
      before_transition all - :stalled => same, :on => any - :crash, :do => :update_dashboard
      ...
    end
  end

As can be seen, any number of transitions can be created using various combinations of configuration options.

Defines a new helper method in an instance or class scope with the given name. If the method is already defined in the scope, then this will not override it.

Example:

  # Instance helper
  machine.define_helper(:instance, :state_name) do |machine, object|
    machine.states.match(object).name
  end

  # Class helper
  machine.define_helper(:class, :state_machine_name) do |machine, klass|
    "State"
  end

You can also define helpers using string evaluation like so:

  # Instance helper
  machine.define_helper :instance, <<-end_eval, __FILE__, __LINE__ + 1
    def state_name
      self.class.state_machine(:state).states.match(self).name
    end
  end_eval

  # Class helper
  machine.define_helper :class, <<-end_eval, __FILE__, __LINE__ + 1
    def state_machine_name
      "State"
    end
  end_eval

Draws a directed graph of the machine for visualizing the various events, states, and their transitions.

This requires both the Ruby graphviz gem and the graphviz library be installed on the system.

Configuration options:

  • :name - The name of the file to write to (without the file extension). Default is "#{owner_class.name}_#{name}"
  • :path - The path to write the graph file to. Default is the current directory (".").
  • :format - The image format to generate the graph in. Default is "png’.
  • :font - The name of the font to draw state names in. Default is "Arial".
  • :orientation - The direction of the graph ("portrait" or "landscape"). Default is "portrait".
  • :output - Whether to generate the output of the graph

Whether a dynamic initial state is being used in the machine

Defines one or more events for the machine and the transitions that can be performed when those events are run.

This method is also aliased as on for improved compatibility with using a domain-specific language.

Configuration options:

  • :human_name - The human-readable version of this event‘s name. By default, this is either defined by the integration or stringifies the name and converts underscores to spaces.

Instance methods

The following instance methods are generated when a new event is defined (the "park" event is used as an example):

  • park(…, run_action = true) - Fires the "park" event, transitioning from the current state to the next valid state. If the last argument is a boolean, it will control whether the machine‘s action gets run.
  • park!(…, run_action = true) - Fires the "park" event, transitioning from the current state to the next valid state. If the transition fails, then a StateMachine::InvalidTransition error will be raised. If the last argument is a boolean, it will control whether the machine‘s action gets run.
  • can_park?(requirements = {}) - Checks whether the "park" event can be fired given the current state of the object. This will not run validations in ORM integrations. To check whether an event can fire and passes validations, use event attributes (e.g. state_event) as described in the "Events" documentation of each ORM integration.
  • park_transition(requirements = {}) - Gets the next transition that would be performed if the "park" event were to be fired now on the object or nil if no transitions can be performed.

With a namespace of "car", the above names map to the following methods:

  • can_park_car?
  • park_car_transition
  • park_car
  • park_car!

The can_park? and park_transition helpers both take an optional set of requirements for determining what transitions are available for the current object. These requirements include:

  • :from - One or more states to transition from. If none are specified, then this will be the object‘s current state.
  • :to - One or more states to transition to. If none are specified, then this will match any to state.
  • :guard - Whether to guard transitions with the if/unless conditionals defined for each one. Default is true.

Defining transitions

event requires a block which allows you to define the possible transitions that can happen as a result of that event. For example,

  event :park, :stop do
    transition :idling => :parked
  end

  event :first_gear do
    transition :parked => :first_gear, :if => :seatbelt_on?
    transition :parked => same # Allow to loopback if seatbelt is off
  end

See StateMachine::Event#transition for more information on the possible options that can be passed in.

Note that this block is executed within the context of the actual event object. As a result, you will not be able to reference any class methods on the model without referencing the class itself. For example,

  class Vehicle
    def self.safe_states
      [:parked, :idling, :stalled]
    end

    state_machine do
      event :park do
        transition Vehicle.safe_states => :parked
      end
    end
  end

Defining additional arguments

Additional arguments on event actions can be defined like so:

  class Vehicle
    state_machine do
      event :park do
        ...
      end
    end

    def park(kind = :parallel, *args)
      take_deep_breath if kind == :parallel
      super
    end

    def take_deep_breath
      sleep 3
    end
  end

Note that super is called instead of super(*args). This allows the entire arguments list to be accessed by transition callbacks through StateMachine::Transition#args like so:

  after_transition :on => :park do |vehicle, transition|
    kind = *transition.args
    ...
  end

Remember that if the last argument is a boolean, it will be used as the run_action parameter to the event action. Using the park action example from above, you can might call it like so:

  vehicle.park                    # => Uses default args and runs machine action
  vehicle.park(:parallel)         # => Specifies the +kind+ argument and runs the machine action
  vehicle.park(:parallel, false)  # => Specifies the +kind+ argument and *skips* the machine action

Example

  class Vehicle
    state_machine do
      # The park, stop, and halt events will all share the given transitions
      event :park, :stop, :halt do
        transition [:idling, :backing_up] => :parked
      end

      event :stop do
        transition :first_gear => :idling
      end

      event :ignite do
        transition :parked => :idling
        transition :idling => same # Allow ignite while still idling
      end
    end
  end

Generates the message to use when invalidating the given object after failing to transition on a specific event

Gets the initial state of the machine for the given object. If a dynamic initial state was configured for this machine, then the object will be passed into the lambda block to help determine the actual state.

Examples

With a static initial state:

  class Vehicle
    state_machine :initial => :parked do
      ...
    end
  end

  vehicle = Vehicle.new
  Vehicle.state_machine.initial_state(vehicle)  # => #<StateMachine::State name=:parked value="parked" initial=true>

With a dynamic initial state:

  class Vehicle
    attr_accessor :force_idle

    state_machine :initial => lambda {|vehicle| vehicle.force_idle ? :idling : :parked} do
      ...
    end
  end

  vehicle = Vehicle.new

  vehicle.force_idle = true
  Vehicle.state_machine.initial_state(vehicle)  # => #<StateMachine::State name=:idling value="idling" initial=false>

  vehicle.force_idle = false
  Vehicle.state_machine.initial_state(vehicle)  # => #<StateMachine::State name=:parked value="parked" initial=false>

Sets the initial state of the machine. This can be either the static name of a state or a lambda block which determines the initial state at creation time.

Initializes the state on the given object. Initial values are only set if the machine‘s attribute hasn‘t been previously initialized.

Configuration options:

  • :force - Whether to initialize the state regardless of its current value
  • :to - A hash to set the initial value in instead of writing directly to the object

Marks the given object as invalid with the given message.

By default, this is a no-op.

on(*names, &block)

Alias for event

other_states(*names, &block)

Alias for state

Sets the class which is the owner of this state machine. Any methods generated by states, events, or other parts of the machine will be defined on the given owner class.

Generates a list of the possible transition sequences that can be run on the given object. These paths can reveal all of the possible states and events that can be encountered in the object‘s state machine based on the object‘s current state.

Configuration options:

  • from - The initial state to start all paths from. By default, this is the object‘s current state.
  • to - The target state to end all paths on. By default, paths will end when they loop back to the first transition on the path.
  • deep - Whether to allow the target state to be crossed more than once in a path. By default, paths will immediately stop when the target state (if specified) is reached. If this is enabled, then paths can continue even after reaching the target state; they will stop when reaching the target state a second time.

Note that the object is never modified when the list of paths is generated.

Examples

  class Vehicle
    state_machine :initial => :parked do
      event :ignite do
        transition :parked => :idling
      end

      event :shift_up do
        transition :idling => :first_gear, :first_gear => :second_gear
      end

      event :shift_down do
        transition :second_gear => :first_gear, :first_gear => :idling
      end
    end
  end

  vehicle = Vehicle.new   # => #<Vehicle:0xb7c27024 @state="parked">
  vehicle.state           # => "parked"

  vehicle.state_paths
  # => [
  #     [#<StateMachine::Transition attribute=:state event=:ignite from="parked" from_name=:parked to="idling" to_name=:idling>,
  #      #<StateMachine::Transition attribute=:state event=:shift_up from="idling" from_name=:idling to="first_gear" to_name=:first_gear>,
  #      #<StateMachine::Transition attribute=:state event=:shift_up from="first_gear" from_name=:first_gear to="second_gear" to_name=:second_gear>,
  #      #<StateMachine::Transition attribute=:state event=:shift_down from="second_gear" from_name=:second_gear to="first_gear" to_name=:first_gear>,
  #      #<StateMachine::Transition attribute=:state event=:shift_down from="first_gear" from_name=:first_gear to="idling" to_name=:idling>],
  #
  #     [#<StateMachine::Transition attribute=:state event=:ignite from="parked" from_name=:parked to="idling" to_name=:idling>,
  #      #<StateMachine::Transition attribute=:state event=:shift_up from="idling" from_name=:idling to="first_gear" to_name=:first_gear>,
  #      #<StateMachine::Transition attribute=:state event=:shift_down from="first_gear" from_name=:first_gear to="idling" to_name=:idling>]
  #    ]

  vehicle.state_paths(:from => :parked, :to => :second_gear)
  # => [
  #     [#<StateMachine::Transition attribute=:state event=:ignite from="parked" from_name=:parked to="idling" to_name=:idling>,
  #      #<StateMachine::Transition attribute=:state event=:shift_up from="idling" from_name=:idling to="first_gear" to_name=:first_gear>,
  #      #<StateMachine::Transition attribute=:state event=:shift_up from="first_gear" from_name=:first_gear to="second_gear" to_name=:second_gear>]
  #    ]

In addition to getting the possible paths that can be accessed, you can also get summary information about the states / events that can be accessed at some point along one of the paths. For example:

  # Get the list of states that can be accessed from the current state
  vehicle.state_paths.to_states # => [:idling, :first_gear, :second_gear]

  # Get the list of events that can be accessed from the current state
  vehicle.state_paths.events    # => [:ignite, :shift_up, :shift_down]

Gets the current value stored in the given object‘s attribute.

For example,

  class Vehicle
    state_machine :initial => :parked do
      ...
    end
  end

  vehicle = Vehicle.new                           # => #<Vehicle:0xb7d94ab0 @state="parked">
  Vehicle.state_machine.read(vehicle, :state)     # => "parked" # Equivalent to vehicle.state
  Vehicle.state_machine.read(vehicle, :event)     # => nil      # Equivalent to vehicle.state_event

Resets any errors previously added when invalidating the given object.

By default, this is a no-op.

Customizes the definition of one or more states in the machine.

Configuration options:

  • :value - The actual value to store when an object transitions to the state. Default is the name (stringified).
  • :cache - If a dynamic value (via a lambda block) is being used, then setting this to true will cache the evaluated result
  • :if - Determines whether an object‘s value matches the state (e.g. :value => lambda {Time.now}, :if => lambda {|state| !state.nil?}). By default, the configured value is matched.
  • :human_name - The human-readable version of this state‘s name. By default, this is either defined by the integration or stringifies the name and converts underscores to spaces.

Customizing the stored value

Whenever a state is automatically discovered in the state machine, its default value is assumed to be the stringified version of the name. For example,

  class Vehicle
    state_machine :initial => :parked do
      event :ignite do
        transition :parked => :idling
      end
    end
  end

In the above state machine, there are two states automatically discovered: :parked and :idling. These states, by default, will store their stringified equivalents when an object moves into that state (e.g. "parked" / "idling").

For legacy systems or when tying state machines into existing frameworks, it‘s oftentimes necessary to need to store a different value for a state than the default. In order to continue taking advantage of an expressive state machine and helper methods, every defined state can be re-configured with a custom stored value. For example,

  class Vehicle
    state_machine :initial => :parked do
      event :ignite do
        transition :parked => :idling
      end

      state :idling, :value => 'IDLING'
      state :parked, :value => 'PARKED
    end
  end

This is also useful if being used in association with a database and, instead of storing the state name in a column, you want to store the state‘s foreign key:

  class VehicleState < ActiveRecord::Base
  end

  class Vehicle < ActiveRecord::Base
    state_machine :attribute => :state_id, :initial => :parked do
      event :ignite do
        transition :parked => :idling
      end

      states.each do |state|
        self.state(state.name, :value => lambda { VehicleState.find_by_name(state.name.to_s).id }, :cache => true)
      end
    end
  end

In the above example, each known state is configured to store it‘s associated database id in the state_id attribute. Also, notice that a lambda block is used to define the state‘s value. This is required in situations (like testing) where the model is loaded without any existing data (i.e. no VehicleState records available).

One caveat to the above example is to keep performance in mind. To avoid constant db hits for looking up the VehicleState ids, the value is cached by specifying the :cache option. Alternatively, a custom caching strategy can be used like so:

  class VehicleState < ActiveRecord::Base
    cattr_accessor :cache_store
    self.cache_store = ActiveSupport::Cache::MemoryStore.new

    def self.find_by_name(name)
      cache_store.fetch(name) { find(:first, :conditions => {:name => name}) }
    end
  end

Dynamic values

In addition to customizing states with other value types, lambda blocks can also be specified to allow for a state‘s value to be determined dynamically at runtime. For example,

  class Vehicle
    state_machine :purchased_at, :initial => :available do
      event :purchase do
        transition all => :purchased
      end

      event :restock do
        transition all => :available
      end

      state :available, :value => nil
      state :purchased, :if => lambda {|value| !value.nil?}, :value => lambda {Time.now}
    end
  end

In the above definition, the :purchased state is customized with both a dynamic value and a value matcher.

When an object transitions to the purchased state, the value‘s lambda block will be called. This will get the current time and store it in the object‘s purchased_at attribute.

Note that the custom matcher is very important here. Since there‘s no way for the state machine to figure out an object‘s state when it‘s set to a runtime value, it must be explicitly defined. If the :if option were not configured for the state, then an ArgumentError exception would be raised at runtime, indicating that the state machine could not figure out what the current state of the object was.

Behaviors

Behaviors define a series of methods to mixin with objects when the current state matches the given one(s). This allows instance methods to behave a specific way depending on what the value of the object‘s state is.

For example,

  class Vehicle
    attr_accessor :driver
    attr_accessor :passenger

    state_machine :initial => :parked do
      event :ignite do
        transition :parked => :idling
      end

      state :parked do
        def speed
          0
        end

        def rotate_driver
          driver = self.driver
          self.driver = passenger
          self.passenger = driver
          true
        end
      end

      state :idling, :first_gear do
        def speed
          20
        end

        def rotate_driver
          self.state = 'parked'
          rotate_driver
        end
      end

      other_states :backing_up
    end
  end

In the above example, there are two dynamic behaviors defined for the class:

  • speed
  • rotate_driver

Each of these behaviors are instance methods on the Vehicle class. However, which method actually gets invoked is based on the current state of the object. Using the above class as the example:

  vehicle = Vehicle.new
  vehicle.driver = 'John'
  vehicle.passenger = 'Jane'

  # Behaviors in the "parked" state
  vehicle.state             # => "parked"
  vehicle.speed             # => 0
  vehicle.rotate_driver     # => true
  vehicle.driver            # => "Jane"
  vehicle.passenger         # => "John"

  vehicle.ignite            # => true

  # Behaviors in the "idling" state
  vehicle.state             # => "idling"
  vehicle.speed             # => 20
  vehicle.rotate_driver     # => true
  vehicle.driver            # => "John"
  vehicle.passenger         # => "Jane"

As can be seen, both the speed and rotate_driver instance method implementations changed how they behave based on what the current state of the vehicle was.

Invalid behaviors

If a specific behavior has not been defined for a state, then a NoMethodError exception will be raised, indicating that that method would not normally exist for an object with that state.

Using the example from before:

  vehicle = Vehicle.new
  vehicle.state = 'backing_up'
  vehicle.speed               # => NoMethodError: undefined method 'speed' for #<Vehicle:0xb7d296ac> in state "backing_up"

State-aware class methods

In addition to defining scopes for instance methods that are state-aware, the same can be done for certain types of class methods.

Some libraries have support for class-level methods that only run certain behaviors based on a conditions hash passed in. For example:

  class Vehicle < ActiveRecord::Base
    state_machine do
      ...
      state :first_gear, :second_gear, :third_gear do
        validates_presence_of   :speed
        validates_inclusion_of  :speed, :in => 0..25, :if => :in_school_zone?
      end
    end
  end

In the above ActiveRecord model, two validations have been defined which will only run when the Vehicle object is in one of the three states: first_gear, second_gear, or +third_gear. Notice, also, that if/unless conditions can continue to be used.

This functionality is not library-specific and can work for any class-level method that is defined like so:

  def validates_presence_of(attribute, options = {})
    ...
  end

The minimum requirement is that the last argument in the method be an options hash which contains at least :if condition support.

Runs a transaction, rolling back any changes if the yielded block fails.

This is only applicable to integrations that involve databases. By default, this will not run any transactions since the changes aren‘t taking place within the context of a database.

Sets a new value in the given object‘s attribute.

For example,

  class Vehicle
    state_machine :initial => :parked do
      ...
    end
  end

  vehicle = Vehicle.new                                   # => #<Vehicle:0xb7d94ab0 @state="parked">
  Vehicle.state_machine.write(vehicle, :state, 'idling')  # => Equivalent to vehicle.state = 'idling'
  Vehicle.state_machine.write(vehicle, :event, 'park')    # => Equivalent to vehicle.state_event = 'park'
  vehicle.state                                           # => "idling"
  vehicle.event                                           # => "park"

Protected Instance methods

The method to hook into for triggering transitions when invoked. By default, this is the action configured for the machine.

Since the default hook technique relies on module inheritance, the action must be defined in an ancestor of the owner classs in order for it to be the action hook.

Adds a new transition callback of the given type.

Tracks the given set of events in the list of all known events for this machine

Tracks the given set of states in the list of all known states for this machine

Runs additional initialization hooks. By default, this is a no-op.

Creates a scope for finding objects with a particular value or values for the attribute.

By default, this is a no-op.

Creates a scope for finding objects without a particular value or values for the attribute.

By default, this is a no-op.

Adds helper methods for automatically firing events when an action is invoked

Determines whether action helpers should be defined for this machine. This is only true if there is an action configured and no other machines have process this same configuration already.

Hooks directly into actions by defining the same method in an included module. As a result, when the action gets invoked, any state events defined for the object will get run. Method visibility is preserved.

Adds helper methods for getting information about this state machine‘s events

Adds helper methods for interacting with the state machine, including for states, events, and transitions

Adds helper methods for accessing naming information about states and events on the owner class

Adds helper methods for getting information about this state machine‘s available transition paths

Defines the with/without scope helpers for this attribute. Both the singular and plural versions of the attribute are defined for each scope helper. A custom plural can be specified if it cannot be automatically determined by either calling pluralize on the attribute name or adding an "s" to the end of the name.

Adds reader/writer methods for accessing the state attribute

Defines the initial values for state machine attributes. Static values are set prior to the original initialize method and dynamic values are set after the initialize method in case it is dependent on it.

Adds predicate method to the owner class for determining the name of the current state

Determines if the machine‘s attribute needs to be initialized. This will only be true if the machine‘s attribute is blank.

Determines whether any of the ancestors for this machine‘s owner class has the given method defined, even if it‘s private.

Pluralizes the given word using pluralize (if available) or simply adding an "s" to the end of the word

[Validate]