spree-state_machine 2.0.0.beta1

data/lib/state_machine/state.rb

@@ -0,0 +1,297 @@
+require 'state_machine/assertions'
+require 'state_machine/state_context'
+
+module StateMachine
+  # A state defines a value that an attribute can be in after being transitioned
+  # 0 or more times.  States can represent a value of any type in Ruby, though
+  # the most common (and default) type is String.
+  # 
+  # In addition to defining the machine's value, a state can also define a
+  # behavioral context for an object when that object is in the state.  See
+  # StateMachine::Machine#state for more information about how state-driven
+  # behavior can be utilized.
+  class State
+    include Assertions
+    
+    # The state machine for which this state is defined
+    attr_accessor :machine
+    
+    # The unique identifier for the state used in event and callback definitions
+    attr_reader :name
+    
+    # The fully-qualified identifier for the state, scoped by the machine's
+    # namespace
+    attr_reader :qualified_name
+    
+    # The human-readable name for the state
+    attr_writer :human_name
+    
+    # The value that is written to a machine's attribute when an object
+    # transitions into this state
+    attr_writer :value
+    
+    # Whether this state's value should be cached after being evaluated
+    attr_accessor :cache
+    
+    # Whether or not this state is the initial state to use for new objects
+    attr_accessor :initial
+    alias_method :initial?, :initial
+    
+    # A custom lambda block for determining whether a given value matches this
+    # state
+    attr_accessor :matcher
+    
+    # Creates a new state within the context of the given machine.
+    # 
+    # Configuration options:
+    # * <tt>:initial</tt> - Whether this state is the beginning state for the
+    #   machine. Default is false.
+    # * <tt>:value</tt> - The value to store when an object transitions to this
+    #   state.  Default is the name (stringified).
+    # * <tt>:cache</tt> - If a dynamic value (via a lambda block) is being used,
+    #   then setting this to true will cache the evaluated result
+    # * <tt>:if</tt> - Determines whether a value matches this state
+    #   (e.g. :value => lambda {Time.now}, :if => lambda {|state| !state.nil?}).
+    #   By default, the configured value is matched.
+    # * <tt>:human_name</tt> - The human-readable version of this state's name
+    def initialize(machine, name, options = {}) #:nodoc:
+      assert_valid_keys(options, :initial, :value, :cache, :if, :human_name)
+      
+      @machine = machine
+      @name = name
+      @qualified_name = name && machine.namespace ? :"#{machine.namespace}_#{name}" : name
+      @human_name = options[:human_name] || (@name ? @name.to_s.tr('_', ' ') : 'nil')
+      @value = options.include?(:value) ? options[:value] : name && name.to_s
+      @cache = options[:cache]
+      @matcher = options[:if]
+      @initial = options[:initial] == true
+      @context = StateContext.new(self)
+      
+      if name
+        conflicting_machines = machine.owner_class.state_machines.select {|other_name, other_machine| other_machine != machine && other_machine.states[qualified_name, :qualified_name]}
+        
+        # Output a warning if another machine has a conflicting qualified name
+        # for a different attribute
+        if conflict = conflicting_machines.detect {|other_name, other_machine| other_machine.attribute != machine.attribute}
+          name, other_machine = conflict
+          warn "State #{qualified_name.inspect} for #{machine.name.inspect} is already defined in #{other_machine.name.inspect}"
+        elsif conflicting_machines.empty?
+          # Only bother adding predicates when another machine for the same
+          # attribute hasn't already done so
+          add_predicate
+        end
+      end
+    end
+    
+    # Creates a copy of this state, excluding the context to prevent conflicts
+    # across different machines.
+    def initialize_copy(orig) #:nodoc:
+      super
+      @context = StateContext.new(self)
+    end
+    
+    # Determines whether there are any states that can be transitioned to from
+    # this state.  If there are none, then this state is considered *final*.
+    # Any objects in a final state will remain so forever given the current
+    # machine's definition.
+    def final?
+      !machine.events.any? do |event|
+        event.branches.any? do |branch|
+          branch.state_requirements.any? do |requirement|
+            requirement[:from].matches?(name) && !requirement[:to].matches?(name, :from => name)
+          end
+        end
+      end
+    end
+    
+    # Transforms the state name into a more human-readable format, such as
+    # "first gear" instead of "first_gear"
+    def human_name(klass = @machine.owner_class)
+      @human_name.is_a?(Proc) ? @human_name.call(self, klass) : @human_name
+    end
+    
+    # Generates a human-readable description of this state's name / value:
+    # 
+    # For example,
+    # 
+    #   State.new(machine, :parked).description                               # => "parked"
+    #   State.new(machine, :parked, :value => :parked).description            # => "parked"
+    #   State.new(machine, :parked, :value => nil).description                # => "parked (nil)"
+    #   State.new(machine, :parked, :value => 1).description                  # => "parked (1)"
+    #   State.new(machine, :parked, :value => lambda {Time.now}).description  # => "parked (*)
+    # 
+    # Configuration options:
+    # * <tt>:human_name</tt> - Whether to use this state's human name in the
+    #   description or just the internal name
+    def description(options = {})
+      label = options[:human_name] ? human_name : name
+      description = label ? label.to_s : label.inspect
+      description << " (#{@value.is_a?(Proc) ? '*' : @value.inspect})" unless name.to_s == @value.to_s
+      description
+    end
+    
+    # The value that represents this state.  This will optionally evaluate the
+    # original block if it's a lambda block.  Otherwise, the static value is
+    # returned.
+    # 
+    # For example,
+    # 
+    #   State.new(machine, :parked, :value => 1).value                        # => 1
+    #   State.new(machine, :parked, :value => lambda {Time.now}).value        # => Tue Jan 01 00:00:00 UTC 2008
+    #   State.new(machine, :parked, :value => lambda {Time.now}).value(false) # => <Proc:0xb6ea7ca0@...>
+    def value(eval = true)
+      if @value.is_a?(Proc) && eval
+        if cache_value?
+          @value = @value.call
+          machine.states.update(self)
+          @value
+        else
+          @value.call
+        end
+      else
+        @value
+      end
+    end
+    
+    # Determines whether this state matches the given value.  If no matcher is
+    # configured, then this will check whether the values are equivalent.
+    # Otherwise, the matcher will determine the result.
+    # 
+    # For example,
+    # 
+    #   # Without a matcher
+    #   state = State.new(machine, :parked, :value => 1)
+    #   state.matches?(1)           # => true
+    #   state.matches?(2)           # => false
+    #   
+    #   # With a matcher
+    #   state = State.new(machine, :parked, :value => lambda {Time.now}, :if => lambda {|value| !value.nil?})
+    #   state.matches?(nil)         # => false
+    #   state.matches?(Time.now)    # => true
+    def matches?(other_value)
+      matcher ? matcher.call(other_value) : other_value == value
+    end
+    
+    # Defines a context for the state which will be enabled on instances of
+    # the owner class when the machine is in this state.
+    # 
+    # This can be called multiple times.  Each time a new context is created,
+    # a new module will be included in the owner class.
+    def context(&block)
+      # Include the context
+      context = @context
+      machine.owner_class.class_eval { include context }
+      
+      # Evaluate the method definitions and track which ones were added
+      old_methods = context_methods
+      context.class_eval(&block)
+      new_methods = context_methods.to_a.select {|(name, method)| old_methods[name] != method}
+      
+      # Alias new methods so that the only execute when the object is in this state
+      new_methods.each do |(method_name, method)|
+        context_name = context_name_for(method_name)
+        context.class_eval <<-end_eval, __FILE__, __LINE__ + 1
+          alias_method :"#{context_name}", :#{method_name}
+          def #{method_name}(*args, &block)
+            state = self.class.state_machine(#{machine.name.inspect}).states.fetch(#{name.inspect})
+            options = {:method_missing => lambda {super(*args, &block)}, :method_name => #{method_name.inspect}}
+            state.call(self, :"#{context_name}", *(args + [options]), &block)
+          end
+        end_eval
+      end
+      
+      true
+    end
+    
+    # The list of methods that have been defined in this state's context
+    def context_methods
+      @context.instance_methods.inject({}) do |methods, name|
+        methods.merge(name.to_sym => @context.instance_method(name))
+      end
+    end
+    
+    # Calls a method defined in this state's context on the given object.  All
+    # arguments and any block will be passed into the method defined.
+    # 
+    # If the method has never been defined for this state, then a NoMethodError
+    # will be raised.
+    def call(object, method, *args, &block)
+      options = args.last.is_a?(Hash) ? args.pop : {}
+      options = {:method_name => method}.merge(options)
+      state = machine.states.match!(object)
+      
+      if state == self && object.respond_to?(method)
+        object.send(method, *args, &block)
+      elsif method_missing = options[:method_missing]
+        # Dispatch to the superclass since the object either isn't in this state
+        # or this state doesn't handle the method
+        begin
+          method_missing.call
+        rescue NoMethodError => ex
+          if ex.name.to_s == options[:method_name].to_s && ex.args == args
+            # No valid context for this method
+            raise InvalidContext.new(object, "State #{state.name.inspect} for #{machine.name.inspect} is not a valid context for calling ##{options[:method_name]}")
+          else
+            raise
+          end
+        end
+      end
+    end
+    
+    # Draws a representation of this state on the given machine.  This will
+    # create a new node on the graph with the following properties:
+    # * +label+ - The human-friendly description of the state.
+    # * +width+ - The width of the node.  Always 1.
+    # * +height+ - The height of the node.  Always 1.
+    # * +shape+ - The actual shape of the node.  If the state is a final
+    #   state, then "doublecircle", otherwise "ellipse".
+    # 
+    # Configuration options:
+    # * <tt>:human_name</tt> - Whether to use the state's human name for the
+    #   node's label that gets drawn on the graph
+    def draw(graph, options = {})
+      node = graph.add_nodes(name ? name.to_s : 'nil',
+        :label => description(options),
+        :width => '1',
+        :height => '1',
+        :shape => final? ? 'doublecircle' : 'ellipse'
+      )
+      
+      # Add open arrow for initial state
+      graph.add_edges(graph.add_nodes('starting_state', :shape => 'point'), node) if initial?
+      
+      true
+    end
+    
+    # Generates a nicely formatted description of this state's contents.
+    # 
+    # For example,
+    # 
+    #   state = StateMachine::State.new(machine, :parked, :value => 1, :initial => true)
+    #   state   # => #<StateMachine::State name=:parked value=1 initial=true context=[]>
+    def inspect
+      attributes = [[:name, name], [:value, @value], [:initial, initial?]]
+      "#<#{self.class} #{attributes.map {|attr, value| "#{attr}=#{value.inspect}"} * ' '}>"
+    end
+    
+    private
+      # Should the value be cached after it's evaluated for the first time?
+      def cache_value?
+        @cache
+      end
+      
+      # Adds a predicate method to the owner class so long as a name has
+      # actually been configured for the state
+      def add_predicate
+        # Checks whether the current value matches this state
+        machine.define_helper(:instance, "#{qualified_name}?") do |machine, object|
+          machine.states.matches?(object, name)
+        end
+      end
+      
+      # Generates the name of the method containing the actual implementation
+      def context_name_for(method)
+        :"__#{machine.name}_#{name}_#{method}_#{@context.object_id}__"
+      end
+  end
+end

data/lib/state_machine/state_collection.rb

@@ -0,0 +1,112 @@
+require 'state_machine/node_collection'
+
+module StateMachine
+  # Represents a collection of states in a state machine
+  class StateCollection < NodeCollection
+    def initialize(machine) #:nodoc:
+      super(machine, :index => [:name, :qualified_name, :value])
+    end
+    
+    # Determines whether the given object is in a specific state.  If the
+    # object's current value doesn't match the state, then this will return
+    # false, otherwise true.  If the given state is unknown, then an IndexError
+    # will be raised.
+    # 
+    # == Examples
+    # 
+    #   class Vehicle
+    #     state_machine :initial => :parked do
+    #       other_states :idling
+    #     end
+    #   end
+    #   
+    #   states = Vehicle.state_machine.states
+    #   vehicle = Vehicle.new               # => #<Vehicle:0xb7c464b0 @state="parked">
+    #   
+    #   states.matches?(vehicle, :parked)   # => true
+    #   states.matches?(vehicle, :idling)   # => false
+    #   states.matches?(vehicle, :invalid)  # => IndexError: :invalid is an invalid key for :name index
+    def matches?(object, name)
+      fetch(name).matches?(machine.read(object, :state))
+    end
+    
+    # Determines the current state of the given object as configured by this
+    # state machine.  This will attempt to find a known state that matches
+    # the value of the attribute on the object.
+    # 
+    # == Examples
+    # 
+    #   class Vehicle
+    #     state_machine :initial => :parked do
+    #       other_states :idling
+    #     end
+    #   end
+    #   
+    #   states = Vehicle.state_machine.states
+    #   
+    #   vehicle = Vehicle.new         # => #<Vehicle:0xb7c464b0 @state="parked">
+    #   states.match(vehicle)         # => #<StateMachine::State name=:parked value="parked" initial=true>
+    #   
+    #   vehicle.state = 'idling'
+    #   states.match(vehicle)         # => #<StateMachine::State name=:idling value="idling" initial=true>
+    #   
+    #   vehicle.state = 'invalid'
+    #   states.match(vehicle)         # => nil
+    def match(object)
+      value = machine.read(object, :state)
+      self[value, :value] || detect {|state| state.matches?(value)}
+    end
+    
+    # Determines the current state of the given object as configured by this
+    # state machine.  If no state is found, then an ArgumentError will be
+    # raised.
+    # 
+    # == Examples
+    # 
+    #   class Vehicle
+    #     state_machine :initial => :parked do
+    #       other_states :idling
+    #     end
+    #   end
+    #   
+    #   states = Vehicle.state_machine.states
+    #   
+    #   vehicle = Vehicle.new         # => #<Vehicle:0xb7c464b0 @state="parked">
+    #   states.match!(vehicle)        # => #<StateMachine::State name=:parked value="parked" initial=true>
+    #   
+    #   vehicle.state = 'invalid'
+    #   states.match!(vehicle)        # => ArgumentError: "invalid" is not a known state value
+    def match!(object)
+      match(object) || raise(ArgumentError, "#{machine.read(object, :state).inspect} is not a known #{machine.name} value")
+    end
+    
+    # Gets the order in which states should be displayed based on where they
+    # were first referenced.  This will order states in the following priority:
+    # 
+    # 1. Initial state
+    # 2. Event transitions (:from, :except_from, :to, :except_to options)
+    # 3. States with behaviors
+    # 4. States referenced via +state+ or +other_states+
+    # 5. States referenced in callbacks
+    # 
+    # This order will determine how the GraphViz visualizations are rendered.
+    def by_priority
+      order = select {|state| state.initial}.map {|state| state.name}
+      
+      machine.events.each {|event| order += event.known_states}
+      order += select {|state| state.context_methods.any?}.map {|state| state.name}
+      order += keys(:name) - machine.callbacks.values.flatten.map {|callback| callback.known_states}.flatten
+      order += keys(:name)
+      
+      order.uniq!
+      order.map! {|name| self[name]}
+      order
+    end
+    
+    private
+      # Gets the value for the given attribute on the node
+      def value(node, attribute)
+        attribute == :value ? node.value(false) : super
+      end
+  end
+end

data/lib/state_machine/state_context.rb

@@ -0,0 +1,138 @@
+require 'state_machine/assertions'
+require 'state_machine/eval_helpers'
+
+module StateMachine
+  # A method was called in an invalid state context
+  class InvalidContext < Error
+  end
+  
+  # Represents a module which will get evaluated within the context of a state.
+  # 
+  # Class-level methods are proxied to the owner class, injecting a custom
+  # <tt>:if</tt> condition along with method.  This assumes that the method has
+  # support for a set of configuration options, including <tt>:if</tt>.  This
+  # condition will check that the object's state matches this context's state.
+  # 
+  # Instance-level methods are used to define state-driven behavior on the
+  # state's owner class.
+  # 
+  # == Examples
+  # 
+  #   class Vehicle
+  #     class << self
+  #       attr_accessor :validations
+  #       
+  #       def validate(options, &block)
+  #         validations << options
+  #       end
+  #     end
+  #     
+  #     self.validations = []
+  #     attr_accessor :state, :simulate
+  #     
+  #     def moving?
+  #       self.class.validations.all? {|validation| validation[:if].call(self)}
+  #     end
+  #   end
+  # 
+  # In the above class, a simple set of validation behaviors have been defined.
+  # Each validation consists of a configuration like so:
+  # 
+  #   Vehicle.validate :unless => :simulate
+  #   Vehicle.validate :if => lambda {|vehicle| ...}
+  # 
+  # In order to scope validations to a particular state context, the class-level
+  # +validate+ method can be invoked like so:
+  # 
+  #   machine = StateMachine::Machine.new(Vehicle)
+  #   context = StateMachine::StateContext.new(machine.state(:first_gear))
+  #   context.validate(:unless => :simulate)
+  #   
+  #   vehicle = Vehicle.new     # => #<Vehicle:0xb7ce491c @simulate=nil, @state=nil>
+  #   vehicle.moving?           # => false
+  #   
+  #   vehicle.state = 'first_gear'
+  #   vehicle.moving?           # => true
+  #   
+  #   vehicle.simulate = true
+  #   vehicle.moving?           # => false
+  class StateContext < Module
+    include Assertions
+    include EvalHelpers
+    
+    # The state machine for which this context's state is defined
+    attr_reader :machine
+    
+    # The state that must be present in an object for this context to be active
+    attr_reader :state
+    
+    # Creates a new context for the given state
+    def initialize(state)
+      @state = state
+      @machine = state.machine
+      
+      state_name = state.name
+      machine_name = machine.name
+      @condition = lambda {|object| object.class.state_machine(machine_name).states.matches?(object, state_name)}
+    end
+    
+    # Creates a new transition that determines what to change the current state
+    # to when an event fires from this state.
+    # 
+    # Since this transition is being defined within a state context, you do
+    # *not* need to specify the <tt>:from</tt> option for the transition.  For
+    # example:
+    # 
+    #   state_machine do
+    #     state :parked do
+    #       transition :to => :idling, :on => [:ignite, :shift_up]                          # Transitions to :idling
+    #       transition :from => [:idling, :parked], :on => :park, :unless => :seatbelt_on?  # Transitions to :parked if seatbelt is off
+    #     end
+    #   end
+    # 
+    # See StateMachine::Machine#transition for a description of the possible
+    # configurations for defining transitions.
+    def transition(options)
+      assert_valid_keys(options, :from, :to, :on, :if, :unless)
+      raise ArgumentError, 'Must specify :on event' unless options[:on]
+      raise ArgumentError, 'Must specify either :to or :from state' unless !options[:to] ^ !options[:from]
+      
+      machine.transition(options.merge(options[:to] ? {:from => state.name} : {:to => state.name}))
+    end
+    
+    # Hooks in condition-merging to methods that don't exist in this module
+    def method_missing(*args, &block)
+      # Get the configuration
+      if args.last.is_a?(Hash)
+        options = args.last
+      else
+        args << options = {}
+      end
+      
+      # Get any existing condition that may need to be merged
+      if_condition = options.delete(:if)
+      unless_condition = options.delete(:unless)
+      
+      # Provide scope access to configuration in case the block is evaluated
+      # within the object instance
+      proxy = self
+      proxy_condition = @condition
+      
+      # Replace the configuration condition with the one configured for this
+      # proxy, merging together any existing conditions
+      options[:if] = lambda do |*condition_args|
+        # Block may be executed within the context of the actual object, so
+        # it'll either be the first argument or the executing context
+        object = condition_args.first || self
+        
+        proxy.evaluate_method(object, proxy_condition) &&
+        Array(if_condition).all? {|condition| proxy.evaluate_method(object, condition)} &&
+        !Array(unless_condition).any? {|condition| proxy.evaluate_method(object, condition)}
+      end
+      
+      # Evaluate the method on the owner class with the condition proxied
+      # through
+      machine.owner_class.send(*args, &block)
+    end
+  end
+end