Module | StateMachine::Integrations::ActiveRecord |
In: |
lib/state_machine/integrations/active_record/versions.rb
lib/state_machine/integrations/active_record.rb |
Adds support for integrating state machines with ActiveRecord models.
Below is an example of a simple state machine defined within an ActiveRecord model:
class Vehicle < ActiveRecord::Base state_machine :initial => :parked do event :ignite do transition :parked => :idling end end end
The examples in the sections below will use the above class as a reference.
By default, the action that will be invoked when a state is transitioned is the save action. This will cause the record to save the changes made to the state machine‘s attribute. Note that if any other changes were made to the record prior to transition, then those changes will be saved as well.
For example,
vehicle = Vehicle.create # => #<Vehicle id: 1, name: nil, state: "parked"> vehicle.name = 'Ford Explorer' vehicle.ignite # => true vehicle.reload # => #<Vehicle id: 1, name: "Ford Explorer", state: "idling">
As described in StateMachine::InstanceMethods#state_machine, event attributes are created for every machine that allow transitions to be performed automatically when the object‘s action (in this case, :save) is called.
In ActiveRecord, these automated events are run in the following order:
For example,
vehicle = Vehicle.create # => #<Vehicle id: 1, name: nil, state: "parked"> vehicle.state_event # => nil vehicle.state_event = 'invalid' vehicle.valid? # => false vehicle.errors.full_messages # => ["State event is invalid"] vehicle.state_event = 'ignite' vehicle.valid? # => true vehicle.save # => true vehicle.state # => "idling" vehicle.state_event # => nil
Note that this can also be done on a mass-assignment basis:
vehicle = Vehicle.create(:state_event => 'ignite') # => #<Vehicle id: 1, name: nil, state: "idling"> vehicle.state # => "idling"
This technique is always used for transitioning states when the save action (which is the default) is configured for the machine.
Beware that public event attributes mean that events can be fired whenever mass-assignment is being used. If you want to prevent malicious users from tampering with events through URLs / forms, the attribute should be protected like so:
class Vehicle < ActiveRecord::Base attr_protected :state_event # attr_accessible ... # Alternative technique state_machine do ... end end
If you want to only have some events be able to fire via mass-assignment, you can build two state machines (one public and one protected) like so:
class Vehicle < ActiveRecord::Base attr_protected :state_event # Prevent access to events in the first machine state_machine do # Define private events here end # Public machine targets the same state as the private machine state_machine :public_state, :attribute => :state do # Define public events here end end
In order to ensure that any changes made during transition callbacks are rolled back during a failed attempt, every transition is wrapped within a transaction.
For example,
class Message < ActiveRecord::Base end Vehicle.state_machine do before_transition do |vehicle, transition| Message.create(:content => transition.inspect) false end end vehicle = Vehicle.create # => #<Vehicle id: 1, name: nil, state: "parked"> vehicle.ignite # => false Message.count # => 0
Note that only before callbacks that halt the callback chain and failed attempts to save the record will result in the transaction being rolled back. If an after callback halts the chain, the previous result still applies and the transaction is not rolled back.
To turn off transactions:
class Vehicle < ActiveRecord::Base state_machine :initial => :parked, :use_transactions => false do ... end end
If using the save action for the machine, this option will be ignored as the transaction will be created by ActiveRecord within save. To avoid this, use a different action like so:
class Vehicle < ActiveRecord::Base state_machine :initial => :parked, :use_transactions => false, :action => :save_state do ... end alias_method :save_state, :save end
As mentioned in StateMachine::Machine#state, you can define behaviors, like validations, that only execute for certain states. One important caveat here is that, due to a constraint in ActiveRecord‘s validation framework, custom validators will not work as expected when defined to run in multiple states. For example:
class Vehicle < ActiveRecord::Base state_machine do ... state :first_gear, :second_gear do validate :speed_is_legal end end end
In this case, the :speed_is_legal validation will only get run for the :second_gear state. To avoid this, you can define your custom validation like so:
class Vehicle < ActiveRecord::Base state_machine do ... state :first_gear, :second_gear do validate {|vehicle| vehicle.speed_is_legal} end end end
If an event fails to successfully fire because there are no matching transitions for the current record, a validation error is added to the record‘s state attribute to help in determining why it failed and for reporting via the UI.
For example,
vehicle = Vehicle.create(:state => 'idling') # => #<Vehicle id: 1, name: nil, state: "idling"> vehicle.ignite # => false vehicle.errors.full_messages # => ["State cannot transition via \"ignite\""]
If an event fails to fire because of a validation error on the record and not because a matching transition was not available, no error messages will be added to the state attribute.
In addition, if you‘re using the ignite! version of the event, then the failure reason (such as the current validation errors) will be included in the exception that gets raised when the event fails. For example, assuming there‘s a validation on a field called name on the class:
vehicle = Vehicle.new vehicle.ignite! # => StateMachine::InvalidTransition: Cannot transition state via :ignite from :parked (Reason(s): Name cannot be blank)
To assist in filtering models with specific states, a series of named scopes are defined on the model for finding records with or without a particular set of states.
These named scopes are essentially the functional equivalent of the following definitions:
class Vehicle < ActiveRecord::Base named_scope :with_states, lambda {|*states| {:conditions => {:state => states}}} # with_states also aliased to with_state named_scope :without_states, lambda {|*states| {:conditions => ['state NOT IN (?)', states]}} # without_states also aliased to without_state end
Note, however, that the states are converted to their stored values before being passed into the query.
Because of the way named scopes work in ActiveRecord, they can be chained like so:
Vehicle.with_state(:parked).all(:order => 'id DESC')
Note that states can also be referenced by the string version of their name:
Vehicle.with_state('parked')
All before/after transition callbacks defined for ActiveRecord models behave in the same way that other ActiveRecord callbacks behave. The object involved in the transition is passed in as an argument.
For example,
class Vehicle < ActiveRecord::Base state_machine :initial => :parked do before_transition any => :idling do |vehicle| vehicle.put_on_seatbelt end before_transition do |vehicle, transition| # log message end event :ignite do transition :parked => :idling end end def put_on_seatbelt ... end end
Note, also, that the transition can be accessed by simply defining additional arguments in the callback block.
after_failure callbacks allow you to execute behaviors when a transition is allowed, but fails to save. This could be useful for something like auditing transition attempts. Since callbacks run within transactions in ActiveRecord, a save failure will cause any records that get created in your callback to roll back. You can work around this issue like so:
class TransitionLog < ActiveRecord::Base establish_connection Rails.env.to_sym end class Vehicle < ActiveRecord::Base state_machine do after_failure do |vehicle, transition| TransitionLog.create(:vehicle => vehicle, :transition => transition) end ... end end
The TransitionLog model establishes a second connection to the database that allows new records to be saved without being affected by rollbacks in the Vehicle model‘s transaction.
In addition to support for ActiveRecord-like hooks, there is additional support for ActiveRecord observers. Because of the way ActiveRecord observers are designed, there is less flexibility around the specific transitions that can be hooked in. However, a large number of hooks are supported. For example, if a transition for a record‘s state attribute changes the state from parked to idling via the ignite event, the following observer methods are supported:
The following class shows an example of some of these hooks:
class VehicleObserver < ActiveRecord::Observer def before_save(vehicle) # log message end # Callback for :ignite event *before* the transition is performed def before_ignite(vehicle, transition) # log message end # Callback for :ignite event *after* the transition has been performed def after_ignite(vehicle, transition) # put on seatbelt end # Generic transition callback *before* the transition is performed def after_transition(vehicle, transition) Audit.log(vehicle, transition) end end
More flexible transition callbacks can be defined directly within the model as described in StateMachine::Machine#before_transition and StateMachine::Machine#after_transition.
To define a single observer for multiple state machines:
class StateMachineObserver < ActiveRecord::Observer observe Vehicle, Switch, Project def after_transition(record, transition) Audit.log(record, transition) end end
In Rails 2.2+, any error message that is generated from performing invalid transitions can be localized. The following default translations are used:
en: activerecord: errors: messages: invalid: "is invalid" # %{value} = attribute value, %{state} = Human state name invalid_event: "cannot transition when %{state}" # %{value} = attribute value, %{event} = Human event name, %{state} = Human current state name invalid_transition: "cannot transition via %{event}"
Notice that the interpolation syntax is %{key} in Rails 3+. In Rails 2.x, the appropriate syntax is {{key}}.
You can override these for a specific model like so:
en: activerecord: errors: models: user: invalid: "is not valid"
In addition to the above, you can also provide translations for the various states / events in each state machine. Using the Vehicle example, state translations will be looked for using the following keys, where model_name = "vehicle", machine_name = "state" and state_name = "parked":
Event translations will be looked for using the following keys, where model_name = "vehicle", machine_name = "state" and event_name = "ignite":
An example translation configuration might look like so:
es: activerecord: state_machines: states: parked: 'estacionado' events: park: 'estacionarse'
# File lib/state_machine/integrations/active_record/versions.rb, line 5 5: def self.active? 6: ::ActiveRecord::VERSION::MAJOR == 2 || ::ActiveRecord::VERSION::MAJOR == 3 && ::ActiveRecord::VERSION::MINOR == 0 7: end
# File lib/state_machine/integrations/active_record/versions.rb, line 22 22: def self.active? 23: ::ActiveRecord::VERSION::MAJOR == 2 24: end
# File lib/state_machine/integrations/active_record/versions.rb, line 128 128: def self.active? 129: ::ActiveRecord::VERSION::MAJOR == 3 && ::ActiveRecord::VERSION::MINOR == 0 130: end
# File lib/state_machine/integrations/active_record/versions.rb, line 108 108: def self.active? 109: ::ActiveRecord::VERSION::MAJOR == 2 && (::ActiveRecord::VERSION::MINOR < 3 || ::ActiveRecord::VERSION::TINY < 2) 110: end
# File lib/state_machine/integrations/active_record/versions.rb, line 98 98: def self.active? 99: ::ActiveRecord::VERSION::MAJOR == 2 && ::ActiveRecord::VERSION::MINOR < 3 100: end
# File lib/state_machine/integrations/active_record/versions.rb, line 118 118: def self.active? 119: ::ActiveRecord::VERSION::MAJOR == 2 && ::ActiveRecord::VERSION::MINOR == 3 && ::ActiveRecord::VERSION::TINY >= 2 120: end
Whether this integration is available. Only true if ActiveRecord::Base is defined.
# File lib/state_machine/integrations/active_record.rb, line 412 412: def self.available? 413: defined?(::ActiveRecord::Base) 414: end
Should this integration be used for state machines in the given class? Classes that inherit from ActiveRecord::Base will automatically use the ActiveRecord integration.
# File lib/state_machine/integrations/active_record.rb, line 419 419: def self.matches?(klass) 420: klass <= ::ActiveRecord::Base 421: end
# File lib/state_machine/integrations/active_record/versions.rb, line 85 85: def action_hook 86: action == :save ? :create_or_update : super 87: end
# File lib/state_machine/integrations/active_record/versions.rb, line 122 122: def ancestors_for(klass) 123: klass.self_and_descendants_from_active_record 124: end
# File lib/state_machine/integrations/active_record/versions.rb, line 112 112: def ancestors_for(klass) 113: klass.self_and_descendents_from_active_record 114: end
# File lib/state_machine/integrations/active_record/versions.rb, line 30 30: def create_scope(name, scope) 31: if owner_class.respond_to?(:named_scope) 32: name = name.to_sym 33: machine_name = self.name 34: 35: # Since ActiveRecord does not allow direct access to the model 36: # being used within the evaluation of a dynamic named scope, the 37: # scope must be generated manually. It's necessary to have access 38: # to the model so that the state names can be translated to their 39: # associated values and so that inheritance is respected properly. 40: owner_class.named_scope(name) 41: owner_class.scopes[name] = lambda do |model, *states| 42: machine_states = model.state_machine(machine_name).states 43: values = states.flatten.map {|state| machine_states.fetch(state).value} 44: 45: ::ActiveRecord::NamedScope::Scope.new(model, :conditions => scope.call(values)) 46: end 47: end 48: 49: # Prevent the Machine class from wrapping the scope 50: false 51: end
# File lib/state_machine/integrations/active_record/versions.rb, line 102 102: def default_error_message_options(object, attribute, message) 103: {:default => @messages[message]} 104: end
# File lib/state_machine/integrations/active_record/versions.rb, line 132 132: def define_action_hook 133: # +around+ callbacks don't have direct access to results until AS 3.1 134: owner_class.set_callback(:save, :after, 'value', :prepend => true) if action_hook == :save 135: super 136: end
# File lib/state_machine/integrations/active_record/versions.rb, line 9 9: def define_static_state_initializer 10: define_helper :instance, "def attributes_from_column_definition(*)\nresult = super\nself.class.state_machines.initialize_states(self, :dynamic => false, :to => result)\nresult\nend\n", __FILE__, __LINE__ + 1 11: end
# File lib/state_machine/integrations/active_record/versions.rb, line 81 81: def i18n_scope(klass) 82: :activerecord 83: end
# File lib/state_machine/integrations/active_record/versions.rb, line 53 53: def invalidate(object, attribute, message, values = []) 54: if defined?(I18n) 55: super 56: else 57: object.errors.add(self.attribute(attribute), generate_message(message, values)) 58: end 59: end
# File lib/state_machine/integrations/active_record/versions.rb, line 26 26: def load_locale 27: super if defined?(I18n) 28: end
# File lib/state_machine/integrations/active_record/versions.rb, line 89 89: def load_observer_extensions 90: super 91: ::ActiveRecord::Observer.class_eval do 92: include StateMachine::Integrations::ActiveModel::Observer 93: end unless ::ActiveRecord::Observer < StateMachine::Integrations::ActiveModel::Observer 94: end
# File lib/state_machine/integrations/active_record/versions.rb, line 77 77: def supports_mass_assignment_security? 78: true 79: end
# File lib/state_machine/integrations/active_record/versions.rb, line 69 69: def supports_observers? 70: true 71: end
# File lib/state_machine/integrations/active_record/versions.rb, line 73 73: def supports_validations? 74: true 75: end
# File lib/state_machine/integrations/active_record/versions.rb, line 61 61: def translate(klass, key, value) 62: if defined?(I18n) 63: super 64: else 65: value ? value.to_s.humanize.downcase : 'nil' 66: end 67: end
Runs state events around the machine‘s :save action
# File lib/state_machine/integrations/active_record.rb, line 481 481: def around_save(object) 482: object.class.state_machines.transitions(object, action).perform { yield } 483: end
Generates the fully-qualifed column name for this machine‘s attribute
# File lib/state_machine/integrations/active_record.rb, line 498 498: def attribute_column 499: connection = owner_class.connection 500: "#{connection.quote_table_name(owner_class.table_name)}.#{connection.quote_column_name(attribute)}" 501: end
Defines a new named scope with the given name
# File lib/state_machine/integrations/active_record.rb, line 511 511: def create_scope(name, scope) 512: lambda {|model, values| model.where(scope.call(values))} 513: end
Creates a scope for finding records with a particular state or states for the attribute
# File lib/state_machine/integrations/active_record.rb, line 487 487: def create_with_scope(name) 488: create_scope(name, lambda {|values| ["#{attribute_column} IN (?)", values]}) 489: end
Creates a scope for finding records without a particular state or states for the attribute
# File lib/state_machine/integrations/active_record.rb, line 493 493: def create_without_scope(name) 494: create_scope(name, lambda {|values| ["#{attribute_column} NOT IN (?)", values]}) 495: end
Uses around callbacks to run state events if using the :save hook
# File lib/state_machine/integrations/active_record.rb, line 472 472: def define_action_hook 473: if action_hook == :save 474: owner_class.set_callback(:save, :around, self, :prepend => true) 475: else 476: super 477: end 478: end
Initializes dynamic states
# File lib/state_machine/integrations/active_record.rb, line 459 459: def define_dynamic_state_initializer 460: define_helper :instance, "def initialize(*)\nsuper do |*args|\nself.class.state_machines.initialize_states(self, :static => false)\nyield(*args) if block_given?\nend\nend\n", __FILE__, __LINE__ + 1 461: end
Defines an initialization hook into the owner class for setting the initial state of the machine before any attributes are set on the object
# File lib/state_machine/integrations/active_record.rb, line 437 437: def define_state_initializer 438: define_static_state_initializer 439: define_dynamic_state_initializer 440: end
Initializes static states
# File lib/state_machine/integrations/active_record.rb, line 443 443: def define_static_state_initializer 444: # This is the only available hook where the default set of attributes 445: # can be overridden for a new object *prior* to the processing of the 446: # attributes passed into #initialize 447: define_helper :class, "def column_defaults(*) #:nodoc:\nresult = super\n# No need to pass in an object, since the overrides will be forced\nself.state_machines.initialize_states(nil, :dynamic => false, :to => result)\nresult\nend\n", __FILE__, __LINE__ + 1 448: end
ActiveModel‘s use of method_missing / respond_to for attribute methods breaks both ancestor lookups and defined?(super). Need to special-case the existence of query attribute methods.
# File lib/state_machine/integrations/active_record.rb, line 518 518: def owner_class_ancestor_has_method?(scope, method) 519: scope == :instance && method == "#{attribute}?" ? owner_class : super 520: end
Only runs validations on the action if using :save
# File lib/state_machine/integrations/active_record.rb, line 430 430: def runs_validations_on_action? 431: action == :save 432: end
Runs a new database transaction, rolling back any changes by raising an ActiveRecord::Rollback exception if the yielded block fails (i.e. returns false).
# File lib/state_machine/integrations/active_record.rb, line 506 506: def transaction(object) 507: object.class.transaction {raise ::ActiveRecord::Rollback unless yield} 508: end