transitions
is a ruby state machine implementation.
This goes into your Gemfile:
gem "transitions", :require => ["transitions", "active_model/transitions"]
… and this into your ORM model:
include ActiveModel::Transitions
gem install transitions
… and this into your class:
include Transitions
class Product
include ActiveModel::Transitions
state_machine do
state :available # first one is initial state
state :out_of_stock, :exit => :exit_out_of_stock
state :discontinued, :enter => lambda { |product| product.cancel_orders }
event :discontinued do
transitions :to => :discontinued, :from => [:available, :out_of_stock], :on_transition => :do_discontinue
end
event :out_of_stock, :success => :reorder do
transitions :to => :out_of_stock, :from => [:available, :discontinued]
end
event :available do
transitions :to => :available, :from => [:out_of_stock], :guard => lambda { |product| product.in_stock > 0 }
end
end
end
In this example we assume that you are in a rails project using Bundler, which
would automatically require transitions
. If this is not the case then you
have to add
require 'transitions'
wherever you load your dependencies in your application.
Known limitations:
-
You can only use one state machine per model. While in theory you can define two or more, this won't work as you would expect. Not supporting this was intentional, if you're interested in the rational look up version 0.1.0 in the CHANGELOG.
-
Use symbols, not strings for declaring the state machine. Using strings is not supported as is using whitespace in names (because
transitions
possibly generates methods out of this).
Use the (surprise ahead) current_state
method - in case you didn't set a
state explicitly you'll get back the state that you defined as initial state.
>> Product.new.current_state
=> :available
You can also set a new state explicitly via update_current_state(new_state, persist = true / false)
but you should never do this unless you really know
what you're doing and why - rather use events / state transitions (see below).
Predicate methods are also available using the name of the state.
>> Product.new.available?
=> true
When you declare an event, say discontinue
, three methods are declared for
you: discontinue
, discontinue!
and can_discontinue?
. The first two
events will modify the state
attribute on successful transition, but only
the bang(!)-version will call save!
. The can_discontinue?
method will not
modify state but instead returns a boolean letting you know if a given
transition is possible.
In addition, a can_transition?
method is added to the object that expects one or more event names as arguments. This semi-verbose method name is used to avoid collisions with [https://github.com/ryanb/cancan](the authorization gem CanCan).
>> Product.new.can_transition? :out_of_stock
=> true
If you need to get all available transitions for the current state you can simply call:
>> Product.new.available_transitions
=> [:discontinued, :out_of_stock]
transitions
offers you the possibility to define a couple of callbacks during the different stages / places of transitioning from one state to another. So let's say you have an event discontinue
which transitions the current state from in_stock
to sold_out
. The callback sequence would look like this:
| discontinue event |
|
|
|
| current_state `in_stock` | ----> executes `exit` callback
|
|
|
| current_state `in_stock` | ----> executes `on_transition` callback if and only the `guard` check was successfull. If not successfull, the chain aborts here and the `event_failed` callback is executed
|
|
|
| current_state `in_stock` | ----> executes `enter` callback for new state `sold_out`
|
|
|
| current_state `in_stock` | ----> executes `event_fired` callback
|
|
|
| current_state `in_stock` | ----> move state from `in_stock` to `sold_out`
|
|
|
| current_state `sold_out` | ----> executes `success` callback of the `discontinue` event
This all looks very complicated (I know), but don't worry, in 99% of all cases you don't have to care about the details and the usage itself is straightforward as you can see in the examples below where each callback is explained a little more throrough.
If you want to trigger a method call when the object enters or exits a state regardless
of the transition that made that happen, use enter
and exit
.
exit
will be called before the transition out of the state is executed. If you want the method
to only be called if the transition is successful, then use another approach.
enter
will be called after the transition has been made but before the object is persisted. If you want
the method to only be called after a successful transition to a new state including persistence,
use the success
argument to an event instead.
An example:
class Motor < ActiveRecord::Base
include ActiveModel::Transitions
state_machine do
state :off, enter: :turn_power_off
state :on, exit: :prepare_shutdown
end
end
Each event definition takes an optional on_transition
argument, which allows
you to execute code on transition. This callback is executed after the exit
callback of the former state (if it has been defined) but before the enter
callback of the new state and only if the guard
check succeeds. There is no check if the callback itself succeeds (meaning that transitions
does not evaluate its return value somewhere). However, you can easily add some properly abstracted error handling yourself by raising an exception in this callback and then handling this exception in the (also defined by you) event_failed
callback (see below and / or the wonderful ascii diagram above).
You can pass in a Symbol, a String, a Proc or an Array containing method names as Symbol or String like this:
event :discontinue do
transitions :to => :discontinued, :from => [:available, :out_of_stock], :on_transition => [:do_discontinue, :notify_clerk]
end
Any arguments passed to the event method will be passed on to the on_transition
callback.
In case you need to trigger a method call after a successful transition you
can use success
. This will be called after the save!
is complete (if you
use the state_name!
method) and should be used for any methods that require
that the object be persisted.
event :discontinue, :success => :notify_admin do
transitions :to => :discontinued, :from => [:available, :out_of_stock]
end
In addition to just specify the method name on the record as a symbol you can pass a lambda to perfom some more complex success callbacks:
event :discontinue, :success => lambda { |order| AdminNotifier.notify_about_discontinued_order(order) } do
transitions :to => :discontinued, :from => [:available, :out_of_stock]
end
If you need it, you can even call multiple methods or lambdas just passing an array:
event :discontinue, :success => [:notify_admin, lambda { |order| AdminNotifier.notify_about_discontinued_order(order) }] do
transitions :to => :discontinued, :from => [:available, :out_of_stock]
end
Since callbacks will not be called by you but by transitions
the scope is different when they are called and you'll run into problems if you use classes / modules in those callbacks that have the same names like transitions
ones, e.g. "Event":
def event_fired(current_state, new_state, event)
Event.create!
end
This will crash because transitions
uses an Event class as well, and, since the scope has changed when transitions
calls this method, transitions
will use it's own Event class here, not yours.
In this case you can try to prefix your models with the "::" operator and see if that solves your problems. See troessner#123 for details.
transitions
will automatically generate scopes for you if you are using
ActiveRecord and tell it to do so via the auto_scopes
option:
Given a model like this:
class Order < ActiveRecord::Base
include ActiveModel::Transitions
state_machine :auto_scopes => true do
state :pick_line_items
state :picking_line_items
event :move_cart do
transitions to: :pick_line_items, from: :picking_line_items
end
end
end
you can use this feature a la:
>> Order.pick_line_items
=> []
>> Order.create!
=> #<Order id: 3, state: "pick_line_items", description: nil, created_at: "2011-08-23 15:48:46", updated_at: "2011-08-23 15:48:46">
>> Order.pick_line_items
=> [#<Order id: 3, state: "pick_line_items", description: nil, created_at: "2011-08-23 15:48:46", updated_at: "2011-08-23 15:48:46">]
Each event definition takes an optional guard
argument, which acts as a
predicate for the transition.
You can pass in Symbols, Strings, or Procs like this:
event :discontinue do
transitions :to => :discontinued, :from => [:available, :out_of_stock], :guard => :can_discontinue
end
or
event :discontinue do
transitions :to => :discontinued, :from => [:available, :out_of_stock], :guard => [:can_discontinue, :super_sure?]
end
Any arguments passed to the event method will be passed on to the guard
predicate.
Note that guards will not raise on failure on their own. This means that if you want to treat the failure of a guard exceptional you'll need to raise an exception yourself explicitly in that guard (see here for the corresponding discussion).
If you'd like to note the time of a state change, Transitions comes with
timestamps free! To activate them, simply pass the timestamp
option to the
event definition with a value of either true or the name of the timestamp
column. NOTE - This should be either true, a String or a Symbol
# This will look for an attribute called exploded_at or exploded_on (in that order)
# If present, it will be updated
event :explode, :timestamp => true do
transitions :from => :complete, :to => :exploded
end
# This will look for an attribute named repaired_on to update upon save
event :rebuild, :timestamp => :repaired_on do
transitions :from => :exploded, :to => :rebuilt
end
In case you define event_fired
and / or event_failed
, transitions
will
use those callbacks correspondingly.
You can use those callbacks like this:
def event_fired(current_state, new_state, event)
MyLogger.info "Event fired #{event.inspect}"
end
def event_failed(event)
MyLogger.warn "Event failed #{event.inspect}"
end
You can easily get a listing of all available states:
Order.available_states # Uses the <tt>default</tt> state machine
# => [:pick_line_items, :picking_line_items]
Same goes for the available events:
Order.available_events
# => [:move_cart]
state_machine :initial => :closed do
state :open
state :closed
end
The explicitly specified state must be one of the states listed in the state definition below, otherwise transitions
will raise a rather unhelpful exception like "NoMethodError: undefined method `call_action' for nil:NilClass" (there's a ticket to fix this already: troessner#112)
To use a different column than state
to track it's value simply do this:
class Product < ActiveRecord::Base
include Transitions
state_machine :attribute_name => :different_column do
...
end
end
Supported versions:
- 2.1
- 2.2
Supported implementations:
- MRI
- Rubinius
- Jruby
- 3
- 4
- Right now it seems like
transitions
does not play well withmongoid
. A possible fix had to be rolled back due to other side effects: troessner#76. Since I know virtually zero about mongoid, a pull request would be highly appreciated. - Multiple state machines are not and will not be supported. For the rationale behind this see the Changelog.
- GraphvizTransitions - Adds support for generating graphs based on states, events and transitions
- Online API Documentation
- Railscasts #392: A Tour of State Machines (requires Pro subscription)
Copyright (c) 2010 Jakub Kuźma, Timo Rößner. See LICENSE for details.