The Machine Objects class library allows the creation of state machines based on the "State" design pattern in plain C++. It extends the pattern with the option to create hierarchical state machines, making it possible to convert the popular UML statechart notation to working code in a straightforward way. Other features are entry and exit actions, state histories and state variables.
The "just show me code" link to an example state machine: Microwave
Homepage is at http://ehiti.de/machine_objects
Feedback goes to macho@ehiti.de
Copyright 2005 by Eduard Hiti.
You are encouraged to provide any changes, extensions and corrections for this software to the author for inclusion into future versions.In my experience as software developer I have found the State design pattern to be very useful. It enables implementing the important concept of state machines with common programming language features. By using only basic language mechanisms it is easy to apply in real-life software development.
Another important property that stems from this simplicity is orthogonality: the pattern can be combined with other design elements, patterns and idioms in arbitrary ways.
In contrast stand tool supported approaches to state machine creation (of which there is no shortage). Based on code generators and graphical editors, they tend to generate incomprehensible code as product and forfeit orthogonality by necessarily being outside the domain of the programming language.
Unfortunately the "State" pattern is limited in scope because it does not allow for hierarchical state machines. This is regrettable because flat state machines tend to become unwieldy when getting more elaborate, for the sheer number of states they produce.
Hierarchical state machines as defined by the statechart notation alleviate this problem by providing an additional structural element with the grouping of states into hierarchies, thereby facilitating the sharing of common logic between states.
The "State" pattern in its original form is not capable of modeling state hierarchies. The Machine Objects class library (in short Macho) extends the concept with this possibility, while keeping the properties of simplicity (where possible) and tool independence from its inspiration.
The class library as such does not need to be installed. Just include the header file Macho.hpp and add the file Macho.cpp to your make file or project definition.
Prerequisite however is a C++ compiler with sane support for templates (like gcc 2.95+ or MSVC7+). For the obsolete but inexplicably popular compiler MSVC6 a special version of the library is provided.
Included are the example state machines HelloWorld, Example, Microwave and Test. To make the examples run just compile them in the directory they are in, for example:
# GCC g++ -o microwave Microwave.cpp Macho.cpp # MSVC7 cl /EHsc Microwave.cpp Macho.cpp
The following descriptions assume some knowledge of the statechart notation. For more information see Wikipedia .
Starting point is the "State" design pattern. The essence of the pattern is to represent states by classes. State transitions are performed by instantiating objects of these classes.
From this perspective the constructors and destructors of state classes can be seen as taking on the role of entry and exit actions. Object attributes then constitute state variables. Events are dispatched by calling methods on state objects which implement the guards, actions and transitions of states.
For states to compose a hierarchy, substates must be able to take over the event handling logic of superstates, selectively redefining it where necessary. There exists a mechanism in C++ allowing redefinition of behaviour on the level of methods: polymorphism through class inheritance.
Modeling the substate/superstate relation with class inheritance is problematic however: the use of constructors and destructors as entry and exit actions is not possible anymore, neither is keeping state variables in objects.
The reason is that base classes are constituent parts of deriving classes, meaning object construction or destruction will trigger all involved class constructors or destructors and initialize or destroy all data members.
This runs counter to the semantics of entry/exit actions and state variables, where a transition between sibling substates should not trigger superstate entry/exit actions nor destroy superstate state variables.
Our solution to this problem is to use explicit methods for state entry and exit, being called in the correct sequence on state transitions. State variables are kept in separate state specific data structures which have a life cycle consistent with the hierarchy of states.
Before diving into implementation details we first define some terminology to be used from here on.
By state class we mean the class defined to represent a particular state. A state machine is a partially ordered set of states having a common top state class. A state machine instance maintains an element of that set as current state to which events can be dispatched. A state machine's event protocol is the set of events that are understood by the machine.
Furthermore extracts from the simple state machine defined in the
file Example.cpp will be used
for illustration. It is a good idea to have a look at this file now if
you are reading this for the first time. It contains a state machine
with four states: a top state, a superstate Super, two
substates StateA and StateB, and a small test
run. Each state has entry/exit actions, the top state and
StateA have state variables and the event protocol consist
of two events, event1 and event2.
A top state is defined by the macro
TOPSTATE:
TOPSTATE(Top) {
...
};
Every state machine must have a top state. The top state's interface defines the machine's event protocol: only the public virtual methods of the top state can be event handlers.
public.
The top state is representative for the whole state machine. All other states of a machine are direct or indirect substates of the top state.
typedef alias
TOP available to all states of the machine.
Substates are defined by the SUBSTATE macro:
SUBSTATE(Super, Top) {
...
};
This statement defines the class Super as substate of class
Top. The macro parameters are the substate's name and
the name of its superstate. The superstate can be the top state or any
other substate.
typedef alias SUPER will point to the
superstate within the substate class.
Why are macros used for Machine Objects? The reason being
convenience only, you'll most likely agree with their use if you look at
the definitions of the macros TOPSTATE and
SUBSTATE:
#define TOPSTATE(TOP) \
struct TOP : public Macho::Link< TOP, Macho::TopBase< TOP > >
#define SUBSTATE(STATE, SUPERSTATE) \
struct STATE : public Macho::Link< STATE, SUPERSTATE >
The code is more readable and actually less error-prone (notice the
multiple use of the same macro parameter in TOPSTATE and
SUBSTATE) by using macro expansion.
Another macro is STATE(S) which MUST be invoked in every
state body:
SUBSTATE(Super, Top) {
STATE(Super)
...
};
The macro parameter is the state's name again. With this macro some definitions are provided in the class body (a constructor for instance). For more details look up the macro's definition.
A state's (optional) entry and exit actions are defined this way:
TOPSTATE(Top) { ... private: void entry(); void exit(); void init(); }; void Top::entry() { ... } void Top::exit() { ... } void Top::init() { ... }
The methods entry and exit of a state are
called upon transitioning into or out of it. The sequence of calls for a
state transition is as follows:
First the exit action of the current state is called and then those of its superstates (in bottom up order), up to and excluding the first superstate that is also superstate of the new state.
Then entry actions of superstates of the new state are called, top down from the first superstate that is not also a superstate of the previous state. Finally the entry action of the new state is called.
Entries and exits may NOT initiate new state transitions (doing otherwise will cause an assertion).
The method init defines a special kind of entry action:
Upon state transition, entry actions of the new state and its superstates
are triggered; init however is called only on the one state
the transition actually goes to (after its entry action is performed).
With init initial transitions of states can be
implemented, where a superstate immediately enters a substate on
entry (like the top state in Example). This implies that init
actions ARE in fact allowed to initiate state transitions (but only to
substates). Transitioning to a state in init will trigger that
new state's init method in turn.
Constructors or destructors can not be used for state classes: if you need to make initializations, do them in entry actions or box constructors.
exit,
entry and init into the private
section of your classes, at least for the top state, so that they are
not part of the event protocol.
State variables allow states to accumulate information from the events they have received in the past. This information is maintained in the scope of the associated state: the variables of a superstate are accessible to that state or any of its substates. Substates have full access to state variables of their superstates.
This is how state variables are defined:
TOPSTATE(Top) { struct Box { Box() : data(0) {} long data; }; STATE(Top) ... };
State variables are contained in a data type named Box
nested into the state class (the type name Box is
mandatory).
STATE macro.
The box type must be default constructable (means: has a default constructor). Apart from this requirement the box type can be any regular C++ type. It could even be a simple typedef:
TOPSTATE(Top) { typedef int Box; STATE(Top) ... };
A box is created before its state is entered (before
the call to entry) and by default destroyed after the state
is left (after the call to exit). By marking a state class
with the PERSISTENT macro, you can override this default
behaviour and have boxes survive state transitions:
SUBSTATE(StateA, Top) struct Box { ... }; STATE(StateA) PERSISTENT() ... };
Persistent boxes are created once at first entry of their state and exist for as long as the state machine instance itself exists.
A state's box is accessed by calling the method box,
which returns a reference to the state's box object:
void StateA::event1(int i) { ... cout << box().data; ... }
Superstate boxes are available by qualifying the box
method with the superstate's name:
void StateB::event2(long l) { ... cout << TOP::box().data; ... }
In this example the top state's box is accessed.
As mentioned the event protocol of a state machine is defined by its top state public interface:
TOPSTATE(Top) {
...
virtual void event1(int i) {}
virtual void event2(long l) {}
...
};
The Example state machine understands the events
event1 and event2: events are named like their
event handler methods.
Event handlers are simple C++ methods and may have arbitrary parameters:
SUBSTATE(StateA, Super) { ... void event1(int i); ... }; void StateA::event1(int i) { ... }
This example defines the event handler for event1 in
StateA. The method should implement all the logic to handle the
event.
The top state event handlers define the default behaviour for the whole state machine. If there is no meaningful implementation for an event handler at top level, the handler could either
assert(false)):
not handling the event will be a runtime error now.State transitions are made by calling the method setState
inside an event handler:
void StateA::event1(int i) { ... setState<StateB>(); }
The template parameter to setState is the new state to be
entered.
setState to in
inline event handlers.
The state transition takes place AFTER control flow leaves the event
handler. It is therefore possible to do meaningful work even after
calling setState (all involved objects are still alive). It
is however not allowed to call setState multiple times with
different states in a single event handler run: this will assert on
you!
To allow for parametrization of target states setState can
take up to six parameters of arbitrary type:
setState<StateB>("Some text", 42, true);
The arguments provided to setState are used to
invoke an init method of the target state with matching
signature:
void StateB::init(const char *, int, bool) { ... }
If no matching init method is provided a (template
mess) compile time error will result.
setState method uses type inference to determine the types of
its parameters. Since type inference is somewhat limited in C++ you must
sometimes disambiguate arguments by casting to avoid type errors.
The target of a state transition can also be provided as a runtime value:
setState(State<StateA>());
or alternatively
setState(StateA::alias());
These examples create objects of type Alias to represent the
state given as template parameter. Objects of this type can also be stored for
later use:
Alias state = State<StateA>();
...
setState(state);
The ability to pass around and store state aliases makes it possible to create reusable code by parametrizing state relationships.
State aliases can also be provided with arguments for initialization of the target state on transition:
Alias s = State<StateA>("Some text", 42, true);
For these arguments the same specifics as for setState above
apply.
To allow for runtime decisions on state relationships these can be inspected at runtime:
// Read like this: StateA "is child of" Super assert(StateA::isChild(Super::alias())); // Read like this: Super "is parent of" StateA assert(Super::isParent(StateA::alias()));
The same can be done with state aliases:
Alias stateA = StateA::alias(); Alias super = Super::alias(); assert(stateA.isChild(super)); assert(super.isParent(stateA));
State equality is determined as expected:
assert(Super::alias() == Super::alias()); assert(stateA != super);
It is possible for a superstate to remember a previously entered substate. This history state can be reentered later with a special state transition.
History for a state is enabled by invoking the macro
HISTORY or DEEPHISTORY in the state
definition:
SUBSTATE(Super, Top) { STATE(Super) HISTORY() ... };
Use of HISTORY selects the shallow history strategy,
whereby only direct substates of a superstate are remembered.
DEEPHISTORY remembers even the substates of substates.
PERSISTENT macro)!
The history of a state can be specified as target of a state transition:
setStateHistory<Super>();
This example will enter the history state of Super. Entry
actions of all involved states will be invoked, with a final call to
the parameterless init method of the actual history state. If
no history information is available (because the state has not been entered
yet or history was not enabled), Super itself is the target of
the transition.
setStateHistory because the exact target of the transition
is not determinable at compile time.
History information can also be inspected directly:
Alias s = Super::history(machine);
A call to a state's static method history with a machine
instance as argument will return the current history of the specified state
at the moment of the call for the specified machine instance. If the state
has no history information available an alias of the state itself is
returned.
machine().
Similar but slightly different is this special alias:
Alias h = StateHistory<Super>(machine());
This alias will keep pointing to the history of Super, not
just the history state at the moment of alias creation. This means that the
alias can represent different states in its life, depending on the changes
to the history of Super.
Consequently the statement
setStateHistory<Super>();
and this statement:
Alias state = StateHistory<Super>(machine());
...
setState(state);
are equivalent.
StateHistory alias becomes invalid when the
associated machine instance has shutdown.
A state machine instance is created by instantiating a machine object:
Macho::Machine<Example::Top> m();
The class Machine is a template class in namespace
Macho; its template parameter is the top state of the state
machine to be run.
The top state is immediately entered and initialized upon machine
creation (entry and init are called).
The Machine constructor takes as optional parameter a box
for the top state:
Macho::Machine<Example::Top> m(new Example::Top::Box(42));
This allows parametrization of a state machine on startup.
An alias can be used to specify a different start state:
Machine<Top> m(State<Example::StateA>());
The machine instance will enter the argument state with all appropriate entry and init actions being executed.
Upon destruction of the machine object the current sub- and superstates up to and including the top state are exited correctly and their boxes destroyed.
_shutdown as empty method.It is also possible to read the top state box of a running machine by
calling the box method of Machine:
cout << m.box().data;
box of class Machine returns a
const reference to the top state box. It is therefore not
possible to change box data (you should really do this from INSIDE your
machine)!
An alias of the machine's current state can be obtained by calling the
method currentState:
assert(m.currentState() == StateA::alias());
A state's history for a particular machine instance can be cleared by
calling the state's static method clearHistory with the
machine object as argument:
Super::clearHistory(m);
This statement resets history information for
Super inside machine m, without affecting substate
history however (use clearHistoryDeep for this).
Another static method of state classes allows testing for the current state of a machine object:
assert(StateA::isCurrent(m));
The method isCurrent returns true if the
given machine object is in the specified state or any of its substates at
that moment.
StateA::alias() == m.currentState()
checks if the machine is exactly in StateA.
To use these methods in event handlers you have to
retrieve the current state machine instance by calling the
machine method:
void StateA::event1(int i) { StateA::clearHistory(machine()); }
It's possible to create "snapshots" of machine objects:
Machine<Example::Top> machine;
Snapshot<Example::Top> snapshot(machine);
A snapshot object stores the entire configuration of a machine at the time the object is created. This includes box contents, history information and the current state of the machine instance.
The snapshot object can then be used to backtrack machine objects to the stored configuration.
To create snapshots ALL box types of a state machine must have a copy
constructor. Because of this additional requirement snapshot functionality must be
explicitly enabled by defining the symbol MACHO_SNAPSHOTS at
compile time:
g++ -D MACHO_SNAPSHOTS Example.cpp Macho.cpp
To restore a machine object to a previous configuration simply assign the snapshot object to it:
Machine<Example::Top> machine;
Snapshot<Example::Top> snapshot(machine);
machine = snapshot;
This will first shutdown the current machine instance (see this note though), consequently exiting all states and deleting all boxes, and then assigns the snapshot's configuration to the machine object. The snapshot object itself is unaffected by this operation.
_restore like this:
void _restore(Macho::_StateInfo & current) { setState(current); }
Destroying a snapshot object will not execute any exit or entry actions, but will delete all box objects contained by the snapshot, executing their destructors.
The simplest way to dispatch events (synchronously) to a state machine is by calling event handlers through the machine object's arrow operator, a technique commonly found with C++ smart pointers:
m->event1(42); m->event2(43);
Another possibility is to create explicit event objects that are then used as arguments to a
machine's dispatch method:
IEvent<Example::Top> * event = Event(&Example::Top::event1, 42); m.dispatch(event); IEvent<Example::Top> * event = Event(&Example::Top::event2, (long) 43); m.dispatch(event);
The Event function takes as arguments a pointer-to-member to
an event handler and all arguments needed to invoke that event handler. Of
course the event parameters must be consistent with the event handler's
signature.
Event is
limited to six, but this limit can easily be raised.
The result of an Event call is a pointer to an object
with the interface IEvent<T> created on the heap (with
T being the top state of the state machine to dispatch to).
This pointer can then be queued for later asynchronous dispatching:
typedef vector<IEvent<Example::Top> *> EventQueue;
EventQueue q;
q.push_back(Event(&Example::Top::event1, 42));
...
for (EventQueue::iterator it = q.begin(); it != q.end(); ++it)
m.dispatch(*it);
Event objects are by default deleted after being dispatched to a machine and cannot be reused afterwards. To avoid memory leaks you should delete undispatched event objects explicitly.
The Event function, too, uses type inference to determine the
types of event parameters. The same specifics as for
setState above apply.
For compilers not supporting "Koenig lookup" name
resolution you will have to call Event fully qualified:
Macho::Event(&Example::Top::event1, 42);
Event objects can also be dispatched inside an event handler:
void StateA::event1(int i) { ... dispatch(Event(&Top::event2, (long) 43)); }
The difference to calling the event handler directly is that the event object is dispatched after control flow has left the event handler, and then also not until a pending state transition has been performed. Therefore the following code fragment:
void StateA::event1(int i) { setState<StateB>(); dispatch(Event(&Top::event1, i)); }
is equivalent to:
void StateA::event1(int i) { dispatch(Event(&Top::event1, i)); setState<StateB>(); }
The consequence in both examples is that the event object will be dispatched in the context of a new state.
The main intended use of this feature is indeed the "forwarding" of events to a new state better suited to handle a particular event.
There are similarities to the concept of "deferred" events in the UML specification, although these have conceptual problems (to be discussed in a later version of this document) which prohibit their implementation in the Machine Objects library.
Please note that only a single event can be dispatched inside an event handler. This is not a technical limitation, but rather an aesthetical: Event chains are hard to follow (since any event could change state) and make the state machine brittle to changes in structure.
Since event chains can be coalesced into a single new event having the same side effects (the state machine is deterministic after all), there seems to be no good reason to allow internal dispatching of more than one event.
For similar reasons it is also not possible to dispatch events in
entry, exit or init.
The following sections focus on topics not essential for effective use of the Machine Objects library and can therefore be omitted on first reading.
On occasion considerable amounts of logic are repeatedly being used in different parts of a state machine design, whenever similar processes are to be followed in different contexts (error handling for example).
To achieve a better factoring of this kind of common behaviour so called template states can be used. Template states allow the implementation of fragments of state machine logic that can be plugged into an actual state machine where appropriate.
A template state is defined similarly to ordinary states:
template<class T>
TSUBSTATE(Template, T) {
TSTATE(Template)
};
This example defines a template state named Template.
The difference to previous state definitions is in the use of the
TSUBSTATE and TSTATE macros instead of
SUBSTATE and STATE, the parameter to these macros
being again the name and the superstate of the template state.
As is apparent by the occurence of the template keyword, a
template state really is a C++ template class behind the curtains.
The single template parameter to be provided is called the anchor of the template state. The anchor must always be an ordinary, non-template state. It is the state where the template state will hook into the state machine, with the anchor being a direct or indirect superstate of the template state.
The following statement
setState<Template<StateA> >();
consequentally means that the state machine will enter an instance of the
template state Template with the anchor state StateA
being the direct superstate of that template instance, which therefore inherits all
event handlers of StateA.
Template states can themselves form hierarchies, when template states are substates of other template states:
template<class T>
TSUBSTATE(TemplateSub, Template<T>) {
TSTATE(TemplateSub)
};
In this example the second parameter to the TSUBSTATE macro is
another template state, making it the superstate of TemplateSub.
The result is a state machine fragment that can be reused repeatedly
in a regular state machine simply by providing an anchor state:
setState<TemplateSub<StateB> >();
This statement will put a state machine into state
TemplateSub<StateB>, which as defined is a substate of
Template<StateB>, which in turn is a substate of anchor
StateB. The template parameter to all the template states is
the non-template state that is the anchor to the whole state machine
fragment.
For more detailed information about Machine Objects please consult the implementation source code and included examples or contact the author.
The Machine Objects library supports the creation of complex state machines in C++ with the well known mechanisms of class inheritance and object polymorphism. But let's look again: there is something else going on. With state machines stripped of their predominant graphical notation, a different concept seems to shine through: type migration.
In implementation and behaviour it is conceivable to view Machine Objects as objects changing their type within a single-rooted multi-level inheritance tree at runtime.
This could be used as an approach to tackle a long standing problem in object-oriented programming: the combination of mutable values with static taxonomies. The issue thereby is that values of specific type could change at runtime in such ways that their type's invariant property does not hold anymore.
The classic example of this problem is the case of Rectangle/Square: a type Square can reasonably be viewed as a subtype of Rectangle. By changing a Square object's width independently from its height (an operation quite conceivable for the supertype Rectangle), the object looses its characteristic property of having a width equal to its height and effectively becomes a Rectangle.
Currently there are no convincing solutions in standard object-oriented programming languages for more difficult problems of this kind. Machine Objects cope with the Rectangle/Square case easily, and can be used for more complex situations.
| Q: | I'm getting weird compiler errors... |
| A: |
Check if you have
|
| Q: | How fast is this stuff? |
| A: | Benchmarking the nontrivial state machine Test (without printing to console) gives me around 1.000.000 state transitions per second on my 2.4 GHz P4 machine. |
| Q: | Does it work with the Visual Studio 6 compiler? |
| A: | You'll find a variant version of Machine Objects for this compiler in the "msvc6" directory. Please consult the contained "Readme" file for differences! |
machine.
clearHistoryDeep to state classes.