An FSM with object state in Java - java

I'm implementing a Finite State Machine in java classes. I wasn't able to find another example with the states and events being their own objects (perhaps there's a reason to that?), and I'm not quite convinced with my solution, especially because I have to assign a state to the FSM (the actual manager) and then assign the FSM to the state (to notify a state change). Here's the code for the state manager, to which I assign a state, and spits it out if asked to:
public class FSM {
public void setCurrentState(FSMState newCurrentState) {
this.currentState = newCurrentState;
}
private FSMState currentState;
public FSMState getCurrentState() {
if (this.currentState == null)
System.out.println("No current state");
return this.currentState;
}
}
and here's the state, that uses a map to map event to output state, and in case of a transition, notifies the FSM class:
public class FSMState implements EventListener {
private FSM managingFSM;
private Map<Event,FSMState> transitions;
public FSMState(FSM managingFSM) {
this.transitions = new HashMap<Event, FSMState>();
this.managingFSM = managingFSM;
}
public void addEventTransition(Event event, FSMState outputState){
transitions.put(event, outputState);
event.registerListener(this);
}
#Override
public void eventOccured(Event e) {
FSMState newState = transitions.get(e);
this.managingFSM.setCurrentState(newState);
}
}
Is there a way, with a kind of similar solution, so that the state wouldn't have to notify the manager?

I'd say your implementation needs quite a few changes. Firstly, you probably want to create the states and their transitions before create the FSM itself. I would make a separate class to manage the initialisation and setup called FSMManager.
First change FSMState to be simply this:
public class FSMState {
private Map<Event,FSMState> transitions;
public FSMState() {
this.transitions = new HashMap<Event, FSMState>();
}
public void addEventTransition(Event event, FSMState outputState){
transitions.put(event, outputState);
}
public FSMState transition(Event e) {
if(transitions.containsKey(e)) {
return transitions.get(e);
}
else {
System.out.println("No transition found");
return this;
}
}
}
Then change the FSM class to look like this:
public class FSM implements EventListener {
private FSMState currentState;
public FSM(FSMState startState) {
currentState = startState;
}
public FSMState getCurrentState() {
if (this.currentState == null)
System.out.println("No current state");
return this.currentState;
}
#Override
public void eventOccured(Event e) {
currentState = currentState.transition(e);
}
}
Finally add the manager to initialise everything:
public class FSMManager {
private static final NUM_STATES = 5;
public static void main(String[] args) {
FSMState[] states = new FSMState[NUM_STATES];
for(FSMState state : states) {
state = new FSMState();
}
// Then add all the state transitions to all the states
states[0].addEventTransition(event1, states[1]);
states[1].addEventTransition(event1, states[2]);
states[1].addEventTransition(event2, states[4]);
// etc, etc
//Finally, create the FSM
FSM fsm = new FSM(state[0]);
//You will also have to register the FSM to listen for all the events
event1.registerListener(fsm);
event2.registerListener(fsm);
...
}
}
I don't know where you are getting all the Events from and how they are actually fired, but the above pattern is a good one to follow. This way when an event is fired, it will only be applied to the current state. The same events will produce different transitions in different states, as they would in a state machine.
This solution should help however it is not perfect as I do not know exactly how you are doing this, so if you have any questions/further problems comment and I will try to improve my answer.

Related

How do I know what changed in my Observable class?

So I'm using the Observer pattern in my app in order to get notified of changes in another class without having to look for them.
I have a Singleton class which extends Observable. Inside this class I have two CountDownTimer type variables. Eachs of these contains two methods: onTick() and onFinished().
Let's call those Timers A and B for the sake of simplicity.
Every time A.onTick(), A.onFinished(), B.onTick(), B.onFinished() are called, I must call notifyObservers() to notify my Observer that something has changed.
Until here everything works fine. The problem is that I know something has changed, but I don't know what exactly has changed. Depending on which one notified me, I must execute some code on the Observer side.
How do I know which of these methods notified me?
Use LiveData instead of Observable. LiveData is quite useful because not only it's observable but also it binds to your activity's lifecycle so you don't have to worry about handling it yourself.
Maybe this example will help you:
public class MyTimerWrapper {
public static MyTimerWrapper getInstance() {
// Your singleton logic
createTimers();
return instance;
}
private CountDownTimer timerA;
private CountDownTimer timerB;
private MutableLiveData<TimerEvent> timerALiveData = new MutableLiveData<TimerEvent>();
private MutableLiveData<TimerEvent> timerBLiveData = new MutableLiveData<TimerEvent>();
public LiveData<TimerEvent> startTimerA() {
timerA.start();
return timerALiveData;
}
public LiveData<TimerEvent> startTimerB() {
timerB.start();
return timerBLiveData;
}
private void createTimers() {
createTimerA();
createTimerB();
}
private void createTimerA() {
timerA = new CountDownTimer(30000, 1000) {
#Override
public void onTick(long millisUntilFinished) {
// If you're running on another thread
timerALiveData.postValue(TimerEvent.TICK);
// Otherwise
timerALiveData.setValue(TimerEvent.TICK);
}
#Override
public void onFinish() {
// If you're running on another thread
timerALiveData.postValue(TimerEvent.FINISH);
// Otherwise
timerALiveData.setValue(TimerEvent.FINISH);
}
}
}
private void createTimerB() {
// Same as createTimerA, but with timerB
}
}
public enum TimerEvent {
TICK,
FINISH
}
Now to observe that data in your activity:
MyTimerWrapper timerWrapper = MyTimerWrapper.getInstance();
timerWrapper.startTimerA().observe(this, new Observer {
#Override
public void onChanged(TimerEvent timerEvent) {
// Here you'll be able to see whether timerA is ticking or finished
}
})
You can create a custom EventType class and pass it to Observable.notifyObservers(Object arg):
public class EventType {
String eventType; //"onTick" or "onFinish"
TimerType timerType;
EventType(String eventType, TimerType timerType){
this.eventType = eventType;
this.timerType = timerType;
}
}
TimerType is an enum type:
public enum TimerType {
A,
B;
}
and create TimerA and TimerB classes extending CountDownTimer:
private class TimerA extends CountDownTimer {
final EventType onTickEvent = new EventType("onTick", TimerType.A);
final EventType onFinishEvent = new EventType("onFinish", TimerType.A);
#Override
public void onTick(long millisUntilFinished) {
notifyObservers(onTickEvent);
}
#Override
public void onFinish() {
notifyObservers(onFinishEvent)
}
}
The Observerwill receive the EventType instance via its update(Observable o, Object arg); in the arg argument

Java Override Method In Instantiated Object Information Transfer

Please note: I am new in this subject.
Suppose I have a class Event.
public class Event {
//constructors, etc.
public void pathFollowed(int location) {
//this method could be called at any time
}
}
And a class called EventManager.
public class EventManager {
private int managerLocation;
private ArrayList<Event> events;
public EventManager() {
events = new ArrayList<Event>();
}
public void addEvent(Event e) {
//THIS IS THE AREA OF INTEREST
events.add(e);
}
}
In the "area of interest" comment, is there any way of setting the value of managerLocation whenever the Event e calls upon pathFollowed(int location). My goal is that when any of the Events in the events arraylist calls pathFollowed(int location) that managerLocation would be set to "location" ("location" referring to the input in the pathfollowed method).
I was originally thinking of over-riding the pathFollowed method, but then I realized this can't be done because by the time the event gets to the addEvent method, it is already instantiated and can't be changed in this manner.
Thanks in advance.
Maybe some kind of listener pattern?
public class Event {
private List<PathListener> pls; //Or just one if you know you'll only need one
//constructors, etc.
public void pathFollowed(int location) {
//this method could be called at any time
for(PathListener pl : pls)
pl.notifyLocation(location);
}
public addPathListener(PathListener pl) {
pls.add(pl);
}
}
EventManager:
public class EventManager implements PathListener {
private int managerLocation;
private ArrayList<Event> events;
public EventManager() {
events = new ArrayList<Event>();
}
public void addEvent(Event e) {
e.addPathListener(this);
events.add(e);
}
#Override
public notifyLocation(int location) { //Of the PathListener interface
managerLocation = location;
}
}
This is just a kind-of-generic example, because I don't know what your purpose is, but maybe it will get you thinking.

Waiting for Sub Tasks to Finish in JavaFX

I am trying to create a simple game using JavaFX. The game consists in a main game and sub-games, which the player may have to play, depending on the result of the main game. In the end, the main game has to update it's state (p.e.: score) depending on the sub-games result.
I made a simplified and generalized version of how I have implemented the game logic:
Result play(Player p) {
Result r = p.play(this);
for(SubGame game : r.getSubGames())
{
p.play(game);
}
update(r);
return r;
}
This game works perfectly in the terminal, since it has a linear execution.
But using JavaFX (implemented in the Player), I cannot control the flow of the program properly, because of the game loop.
I have followed this tutorial to handle multiple screens for the main game and sub-games. The Player class can successfully change the screen to a sub-game, using the handler. But the update no longer waits for the sub-games to be played, and this function returns while the player is mid-game.
I am trying to keep the game logic separate from UI, so changes to the code showed above shouldn't be dependent on the interface framework.
Any help?
Use an event-driven approach, setting the values of observable properties, and responding when they change.
For example, you might encapsulate the state of the game with
public class GameState {
private ObservableList<SubGame> currentGames = FXCollections.observableArrayList();
public ObservableList<SubGame> getCurrentGames() {
return currentGames();
}
private ReadOnlyObjectWrapper<SubGame> currentGame = new ReadOnlyObjectProperty<>();
public ReadOnlyObjectProperty<SubGame> currentGameProperty() {
return currentGame.getReadOnlyProperty() ;
}
public final SubGame getCurrentGame() {
return currentGameProperty().get();
}
public GameState() {
// initialize sub game list...
}
public void nextGame() {
int index = currentGames.indexOf(currentGame.get());
if (index < currentGames.size() - 1) {
currentGame.set(currentGames.get(index + 1));
}
}
public void start() {
currentGame.set(currentGames().get(0));
}
public boolean hasMoreGames() {
return currentGames.indexOf(currentGame.get()) < currentGames.size() - 1 ;
}
}
And similarly you might have some observable state in your SubGame class:
public class SubGame {
private final BooleanProperty finished = new SimpleBooleanProperty();
public BooleanProperty finishedProperty() {
return finished ;
}
public final boolean isFinished() {
return finishedProperty().get();
}
public final void setFinished(boolean finished) {
finishedProperty().set(finished) ;
}
// ...
}
Now your game logic is just implemented with listeners:
void play(Player p) {
Result r = p.play(this);
GameState gameState = new GameState();
gameState.currentGameProperty().addListener((obs, oldGame, newGame) -> {
newGame.finishedProperty().addListener((obs, wasFinished, isNowFinished) -> {
if (isNowFinished) {
// maybe update score etc based on state of newGame...
if (gameState.hasMoreGames()) {
gameState.nextGame();
} else {
// logic here for "all games are finished...
}
}
});
});
gameState.start();
}
Obviously the details of how you implement this depend on your requirements etc, but this general approach should work for anything you need.

State pattern: Identify state class type from string

Background
I have a command that uses the state pattern. When the command state changes I am notified in the UI that for example class stageOneState is now the active state. Is it bad practice to check the state class type using a string as an identifier? Is this undoing the work of the state pattern?
What would be an alternative?
Example
if (notifiedState.type == "state1") {
// Update UI accroding to state1
} else ...
Example
Example from http://www.tutorialspoint.com/design_pattern/state_pattern.htm
public interface State {
public void doAction(Context context);
}
public class StartState implements State {
public void doAction(Context context) {
System.out.println("Player is in start state");
context.setState(this);
}
public String toString(){
return "Start State";
}
}
public class StopState implements State {
public void doAction(Context context) {
System.out.println("Player is in stop state");
context.setState(this);
}
public String toString(){
return "Stop State";
}
}
public class Context {
private State state;
public Context(){
state = null;
}
public void setState(State state){
this.state = state;
}
public State getState(){
return state;
}
}
public class StatePatternDemo {
public static void main(String[] args) {
Context context = new Context();
StartState startState = new StartState();
startState.doAction(context);
System.out.println(context.getState().toString());
StopState stopState = new StopState();
stopState.doAction(context);
System.out.println(context.getState().toString());
}
}
You undo the state pattern when you check the state of an object from the outside and then act differently depending on which state it is. If it seems more convenient, then probably state is the wrong tool for the job.
The sentence "command having a state" raises a red flag. Is this some ungodly mashup of state and command patterns?
There is a terrible error in the example: you are comparing strings using identity instead of value.
Generally, if you have any doubts about any pattern, just don't try to use it. They are poisonous animals, better studied from safe distance.
This pattern is sometimes called stringly typed (a pun on strongly typed).
The best approach is to use an enum.

java events and listeners, bad implementation?

I am currently implementing custom events and listeners according to the code posted below. I have been told that this is a very dirty implementation and that this needs to be changed. However, i am very new to java and android and do not see what is wrong with the current implementation. The way i have it below works and seems to be doing everything i needed it too. I was wondering if some people could please take a look at my code and make some suggestions on what i should change and what i am doing wrong. Taking my example and modifying it so that i can see what your talking about would be greatly appreciated.
Thanks in advance!
/* SmartApp.java */
public class SmartApp extends Activity
{
private ConnectDevice cD = new ConnectDevice();
private DataRobot dR = new DataRobot();
private DataBuilder dB = new DataBuilder();
private DataSender dS = new DataSender();
public void onCreate(Bundle savedInstanceState) {
super.onCreate(savedInstanceState);
setContentView(R.layout.intro);
cD.addDataReceivedListener(new DataReceivedListener() {
#Override
public void dataReceivedReceived(DataReceivedEvent event) {
// TODO Auto-generated method stub
dR.analyzeData(event.getData());
}
});
dR.addDataAnalyzedListener(new DataAnalyzedListener() {
#Override
public void dataAnalyzedReceived(DataAnalyzedEvent event) {
// TODO Auto-generated method stub
dB.submitData(event.getData());
}
});
dB.addDataBuilderListener(new DataBuilderListener() {
#Override
public void dataBuilderReceived(DataBuilderEvent event) {
// TODO Auto-generated method stub
dS.sendData(event.getData());
}
});
}
}
/* ConnectDevice.java
* This class is implementing runnable because i have a thread running that is checking
* the contents of a socket. Irrelevant to events. */
public class ConnectDevice implements Runnable {
private List _listeners = new ArrayList();
private String data;
/* Constructor */
public ConnectDevice() {// does some socket stuff here, irrelevant to the events}
public void run() {// does some socket stuff here, irrelevant to the events}
public synchronized void addDataReceivedListener(DataReceivedListener listener) {
_listeners.add(listener);
}
public synchronized void removeDataReceivedListener(DataReceivedListener listener) {
_listeners.remove(listener);
}
private synchronized void fireDataReceivedEvent(String temp) {
DataReceivedEvent dRE = new DataReceivedEvent(this, temp);
Iterator listeners = _listeners.iterator();
while(listeners.hasNext()) {
((DataReceivedListener)listeners.next()).dataReceivedReceived(dRE);
}
}
public interface DataReceivedListener {
public void dataReceivedReceived(DataReceivedEvent event);
}
}
/* DataRobot.java */
public class DataRobot {
/* This class is for analyzing the data */
private List _listeners = new ArrayList();
private String data;
public boolean analyzeData(String temp) {
/* Analyze the data
* This function analyzes the data, as explained in the OP
* This function fires the analyzed data event when finished
* analyzing the data.
*/
data = temp;
fireDataAnalyzedEvent(data); // this fires the dataanalyzedevent
return true; //for now this will always return true
}
public synchronized void addDataAnalyzedListener(DataAnalyzedListener listener) {
_listeners.add(listener);
}
public synchronized void removeDataAnalyzedListener(DataAnalyzedListener listener) {
_listeners.remove(listener);
}
private synchronized void fireDataAnalyzedEvent(String temp) {
DataAnalyzedEvent dRE = new DataAnalyzedEvent(this, temp);
Iterator listeners = _listeners.iterator();
while(listeners.hasNext()) {
((DataAnalyzedListener)listeners.next()).dataAnalyzedReceived(dRE);
}
}
public interface DataAnalyzedListener {
public void dataAnalyzedReceived(DataAnalyzedEvent event);
}
}
/* DataBuilder.java */
public class DataBuilder {
private List _listeners = new ArrayList();
private String data;
public boolean submitData(String temp) {
/* Builds the data
* This function builds the data, as explained in the OP
* This function fires the databuilder data event when finished
* building the data.
*/
data = temp;
fireDataBuilderEvent(data); //firing the databuilder event when finished
return true;
}
public synchronized void addDataBuilderListener(DataBuilderListener listener) {
_listeners.add(listener);
}
public synchronized void removeDataBuilderListener(DataBuilderListener listener) {
_listeners.remove(listener);
}
private synchronized void fireDataBuilderEvent(String temp) {
DataBuilderEvent dRE = new DataBuilderEvent(this, temp);
Iterator listeners = _listeners.iterator();
while(listeners.hasNext()) {
((DataBuilderListener)listeners.next()).dataBuilderReceived(dRE);
}
}
public interface DataBuilderListener {
public void dataBuilderReceived(DataBuilderEvent event);
}
}
/* DataSender.java */
/* this class has no event, because it is done firing events at this point */
public class DataSender {
private String data;
public boolean sendData(String temp) {
data = temp;
return true;
}
}
Below here are the event objects for each event. I Have each of this defined in a separate file, not sure if that is good procedure or not.
/* DataReceivedEvent.java */
public class DataReceivedEvent extends EventObject{
private String data;
public DataReceivedEvent(Object source, String temp) {
super(source);
// TODO Auto-generated constructor stub
data = temp;
}
public String getData() {
// this function is just an accessor function
return data;
}
}
/* DataAnalyzedEvent.java */
public class DataAnalyzedEvent extends EventObject{
private String data;
public DataAnalyzedEvent(Object source, String temp) {
super(source);
// TODO Auto-generated constructor stub
data = temp;
}
public String getData() {
// this function is just an accessor function
return data;
}
}
/* DataBuilderEvent.java */
public class DataBuilderEvent extends EventObject {
private String data;
public DataBuilderEvent(Object source, String temp) {
super(source);
// TODO Auto-generated constructor stub
data = temp;
}
public String getData() {
// this function is just an accessor function
return data;
}
}
I would not say it is a "very dirty implementation". Using callbacks/observers/listeners is a good practice in my opinion.
When I write Android applications I like to layer it such that the "application" is plain old Java with no Android imports and could theoretically be used in a Swing app, a Java EE-based web site, etc. The "Android" part is strictly user interface.
What I use callbacks for is to allow the Android code to register interest in events that take place in the application. For example, in a Blackjack game, an Activity might call game.getDealer().playHand() to tell the application to perform the dealer hand play logic. As that logic executes in the application, events are fired like cardDrawn(card), cardFlipped(card), handTotalChanged(handTotal), etc. The Android part of the app listens to these and redraws things on the screen accordingly (but it knows nothing about Blackjack).
I actually just have my activities implement interfaces like CardListener, HandListener, etc. so they can receive the event directly (unlike how you do it), but your style isn't necessarily a bad way.
I agree with #SingleShot in theory, for the parts of your Android application that can be Android-agnostic, and so long as the overhead introduced by all the indirection layers does not slow the app down too much. IMHO, in many apps, there is relatively little that fits this description.
In another post, you proposed your above solution for one activity to communicate to another activity. In Android, activities aren't just some Java objects you can toss around willy-nilly. They are managed by the OS and have particular lifecycles. While the observer/observable pattern is quite delightful in some places, it is unsuitable where the observer/observable connection will create garbage collection problems. In particular, one activity cannot, and should not, be trying to hold some sort of listener interface on another activity.
Similarly, a clean observer/observable pattern may break down in the face of databases, threads, services, and other bits of Android reality.
So, in pure Java code, isolated from Android, what you have is probably OK. However, do not go around recommending it as solutions for Android-specific problems unless you know it will work for those Android-specific problems. And, when you start trying to make your code work in an Android app, please do not be shocked if you run into problems trying to make your textbook pattern implementation work within the constraints placed upon Android apps.

Categories

Resources