Related
In my game I am creating some big class, which store references to smaller classes, which store some references too. And during gameplay I need to recreate this big class with all its dependencies by destroying it and making new one.
It's creation looks like that:
ABigClass::ABigClass()
{
UE_LOG(LogTemp, Warning, TEXT("BigClass created"));
SmallClass1 = NewObject<ASmallClass>();
SmallClass2 = NewObject<ASmallClass>();
......
}
And it works. Now I want to destroy and re create it by calling from some function
BigClass->~ABigClass();
BigClass= NewObject<ABigClass>();
which destroys BigClass and creates new one with new small classes, the problem is, that old small classes are still in memory, I can see it by logging their destructors.
So I try to make such destructor for Big Class
ABigClass::~ABigClass()
{
SmallClass1->~ASmallClass();
SmallClass2->~ASmallClass();
......
UE_LOG(LogTemp, Warning, TEXT("BigClass destroyed"));
}
SmallClass is inherited from other class, which have its own constructor and destructor, but I do not call it anywhere.
Sometimes it works, but mostly causes UE editor to crash when code compiled, or when game starsted/stopped.
Probably there is some more common way to do what I want to do? Or some validation which will prevent it from crash?
Please help.
Don't manually call a destructor. Replace
SmallClass1->~ASmallClass();
SmallClass2->~ASmallClass();
with either
delete SmallClass1;
SmallClass1 = nullptr;
or by nothing if those type are ref-counted by unreal in some fashion (likely).
Finally I found the way to do this.
First, I needed to get rid of all UPROPERTYs, related to classes, which I am going to destroy, except UPROPERTYs in a class, which controls their creation and destruction. If I need to expose these classes to blueprints somewhere else, it can be done with BlueprintCallable getter and setter functions.
Then I needed to calm down UE's garbage collector, which destroy objects on hot reload and on game shutdown in random order, ignoring my destructor hierarchy, which results in attempt to destroy already destroyed object and crash. So before doing something in destructor with other objects, I need to check if there is something to destroy, adding IsValidLowLevel() check.
And instead of delete keyword it is better to use DestructItem() function, it seems to be more garbage collector-friendly in many ways.
Also, I did not found a way to safely destroy objects, which are spawned in level. Probably because they are referenced somewhere else, but I dont know where, but since they are lowest level in my hierarchy, i can just Destroy() them in the world, and do not bother when exactly they will disappear from memory by garbage collector's will.
Anyway, I ended up with following code:
void AGameModeBase::ResetGame()
{
if (BigClass->IsValidLowLevel()) {
DestructItem(BigClass);
BigClass= nullptr;
BigClass= NewObject<ABigClass>();
UE_LOG(LogTemp, Warning, TEXT("Bigclass recreated"));
}
}
ABigClass::~ABigClass()
{
if (SmallClass1) {
if (SmallClass->IsValidLowLevel())
{
DestructItem(SmallClass1);
}
SmallClass1 = nullptr;
}
if (SmallClass2) {
if (SmallClass->IsValidLowLevel())
{
DestructItem(SmallClass2);
}
SmallClass2 = nullptr;
}
...
UE_LOG(LogTemp, Warning, TEXT("BigClass destroyed"));
}
ASmallClass::~ASmallClass()
{
for (ATinyClass* q : TinyClasses)
{
if (q->IsValidLowLevel())
{
q->Destroy();
}
}
TinyClasses = {};
UE_LOG(LogTemp, Warning, TEXT("SmallClass destroyed"));
}
And no crashes. Probably someone will find it useful in cases it is needed to clear game lever from hierarchy of objects without fully reloading it.
I am using the example below to implement a state design pattern.
https://sourcemaking.com/design_patterns/state/cpp/1
I don't wanna delete the *this pointer inside a member class because is not safe (I will call other member function after delete).
class ON: public State
{
public:
ON()
{
cout << " ON-ctor ";
};
~ON()
{
cout << " dtor-ON\n";
};
void off(Machine *m);
};
class OFF: public State
{
public:
OFF()
{
cout << " OFF-ctor ";
};
~OFF()
{
cout << " dtor-OFF\n";
};
void on(Machine *m)
{
cout << " going from OFF to ON";
m->setCurrent(new ON());
delete this; // <<< This line looks suspect and unsafe
}
};
void ON::off(Machine *m)
{
cout << " going from ON to OFF";
m->setCurrent(new OFF());
delete this; // <<< This line looks suspect and unsafe
}
Is there a better approach to implement the state design pattern? I thought about use Singletons, but I want avoid using singletons.
Best regards,
The design used in your link is:
a state machine keeps track of the pointer to the current state.
the current state is responsible to set the new state when a transition occcurs
it does so by creating the new state and instructing the machine to change the pointer
once this is done, the it comits suicide by deleteing itself.
As explained here, this can work if extra care is taken. Your code respects the prerequisites.
The alternative would be that the state machine deletes the current state in setCurrent(). But this is worse: as soon as the machine has deleted the object, this object no longer exists, so that when setCurrent() returns, you're de facto in the same (dangerous) situation as before (with delete this), with the important difference that this would not be obvious.
Edit:
If you feel unconfortable with this construct, you could opt for a variant:
class Machine
{
class State *current; // current state
class State *nextstate; // null, unless a state transition was requested
void check_transition(); // organise switch to nextstate if transition is needed
public:
Machine();
void setCurrent(State *s)
{
nextstate = s; // only sets the info about nextstate
}
void set_on();
void set_off();
};
In this scenario, the state-machine functions need to be slightly updated:
void Machine::set_on()
{
current->set_on(this); // as before
check_transition(); // organise the transition if the transition was requested
}
The state transition would then be managed by the state machine:
void Machine::check_transition() {
if (nextstate) {
swap (current, nextstate);
delete nextstate; // this contains the former current
nextstate = nullptr;
}
}
The machine organises the deletion of the unused states. Note that this approach would allow the use of shared_ptr instead of raw pointers.
Here a small online proof of concept.
By the way: the machine destructor should also delete the current and next state, in order to avoid leaking. And in any case the state destructor should be defined virtual
"... but I want avoid using singletons."
That's probably one of their rare valid use cases, since states should be stateless themselves (which doesn't mean they don't need an instance), and accessible concurrently regardless of the current state machine's context.
So that means you actually just need one instance of a state at a time.
The example shown at the mentioned website, also gives you an unnecessary performance hit, at the cost coming from new and delete whenever state changes, which could be avoided to have steady static instances for the states.
Actually you could consider to provide state instances as with the Flyweight Design Pattern, which essentially boils down to have Singleton state instances.
But well, it depends. UML state diagrams actually allow to have non-stateless states (as composite states with history attributes, or active states).
Have a look at my STTCL template library concept document, it explains some of the aspects, and design decisions I have used, to develop that template library, and how it can be used properly.
Be assured you I don't have any single delete this; in there ;-).
I'd say that example the website currently gives, is really badly designed and not generally recommendable1.
Using delete this isn't appropriate, dangerous and not acceptable, as you mentioned.
While using Singleton's is, if you have their valid use case at hand.
1) Sadly enough to notice this, since I've been using it as a "reference" for many design-patterns related questions here.
I have
struct MyWidget : QWidget {
// non-GUI related stuff:
int data;
int doSth();
};
I need to access a MyWidget instance from another thread (i.e. not the main thread). Is there any way to do that safely? I understand that I cannot access GUI related functions because some backends (e.g. MacOSX/Cocoa) don't support that. However, I only need to access data or doSth() in this example. But from what I have understand, there is simply no way to guarantee the lifetime of the object - i.e. if the parent window with that widget closes, the MyWidget instance gets deleted.
Or is there a way to guarantee the lifetime? I guess QSharedPointer doesn't work because the QWidget does its lifetime handling internally, depending on the parent widget. QPointer of course also doesn't help because it is only weak and there is no locking mechanism.
My current workaround is basically:
int widget_doSth(QPointer<MyWidget> w) {
int ret = -1;
execInMainThread_sync([&]() {
if(w)
ret = w->doSth();
});
return ret;
}
(execInMainThread_sync works by using QMetaMethod::invoke to call a method in the main thread.)
However, that workaround doesn't work anymore for some specific reason (I will explain later why, but that doesn't matter here). Basically, I am not able to execute something in the main thread at that point (for some complicated deadlock reasons).
Another workaround I'm currently thinking about is to add a global mutex which will guard the MyWidget destructor, and in the destructor, I'm cleaning up other weak references to the MyWidget. Then, elsewhere, when I need to ensure the lifetime, I just lock that mutex.
The reason why my current workaround doesn't work anymore (and that is still a simplified version of the real situation):
In MyWidget, the data is actually a PyObject*.
In the main thread, some Python code gets called. (It's not really possible to avoid any Python code calls at all in the main thread in my app.) That Python code ends up doing some import, which is guarded by some Python-import-mutex (Python doesn't allow parallel imports.)
In some other Python thread, some other import is called. That import now locks the Python-import-mutex. And while it's doing its thing, it does some GC cleanup at some point. That GC cleanup calls the traverse function of some object which holds that MyWidget. Thus, it must access the MyWidget. However, execInMainThread_sync (or equivalently working solutions) will deadlock because the main thread currently waits for the Python-import-lock.
Note: The Python global interpreter lock is not really the problem. Of course it gets unlocked before any execInMainThread_sync call. However, I cannot really check for any other potential Python/whatever locks. Esp. I am not allowed to just unlock the Python-import-lock -- it's there for a reason.
One solution you might think of is to really just avoid any Python code at all in the main thread. But that has a lot of drawbacks, e.g. it will be slow, complicated and ugly (the GUI basically only shows data from Python, so there need to be a huge proxy/wrapper around it all). And I think I still need to wait at some points for the Python data, so I just introduce the possible deadlock-situation at some other point.
Also, all the problems would just go away if I could access MyWidget safely from another thread. Introducing a global mutex is the much cleaner and shorter solution, compared to above.
You can use the signal/slot mechanism, but it can be tedious, if the number of GUI controls is large. I'd recommend a single signal and slot to control the gui. Send over a struct with all the info needed for updating the GUI.
void SomeWidget::updateGUISlot(struct Info const& info)
{
firstControl->setText(info.text);
secondControl->setValue(info.value);
}
You don't need to worry about emitting signals, if the recipient is deleted. This detail is handled by Qt. Alternatively, you can wait for your threads to exit, after exiting the GUI threads event loop. You'll need to register the struct with Qt.
EDIT:
From what I've read from your extended question, you're problems are related to communication between threads. Try pipes, (POSIX) message queues, sockets or POSIX signals instead of Qt signals for inter-thread communication.
Personally I don't like designs where GUI stuff (ie: A widget) has non-GUI related stuff... I think you should separate these two from each other. Qt needs to keep the GUI objects always on the main thread, but anything else (QObject derived) can be moved to a thread (QObject::moveToThread).
It seems that what you're explaining has nothing at all to do with widgets, Qt, or anything like that. It's a problem inherent to Python and its threading and the lock structure that doesn't make sense if you're multithreading. Python basically presumes that any object can be accessed from any thread. You'd have the same problem using any other toolkit. There may be a way of telling Python not to do that - I don't know enough about the cpython implementation's details, but that's where you'd need to look.
That GC cleanup calls the traverse function of some object which holds that MyWidget
That's your problem. You must ensure that such cross-thread GC cleanup can't happen. I have no idea how you'd go about it :(
My worry is that you've quietly and subtly shot yourself in the foot by using Python, in spite of everyone claiming that only C/C++ lets you do it at such a grand scale.
My solution:
struct MyWidget : QWidget {
// some non-GUI related stuff:
int someData;
virtual void doSth();
// We reset that in the destructor. When you hold its mutex-lock,
// the ref is either NULL or a valid pointer to this MyWidget.
struct LockedRef {
boost::mutex mutex;
MyWidget* ptr;
LockedRef(MyWidget& w) : ptr(&w) {}
void reset() {
boost::mutex::scoped_lock lock(mutex);
ptr = NULL;
}
};
boost::shared_ptr<LockedRef> selfRef;
struct WeakRef;
struct ScopedRef {
boost::shared_ptr<LockedRef> _ref;
MyWidget* ptr;
bool lock;
ScopedRef(WeakRef& ref);
~ScopedRef();
operator bool() { return ptr; }
MyWidget* operator->() { return ptr; }
};
struct WeakRef {
typedef boost::weak_ptr<LockedRef> Ref;
Ref ref;
WeakRef() {}
WeakRef(MyWidget& w) { ref = w.selfRef; }
ScopedRef scoped() { return ScopedRef(*this); }
};
MyWidget();
~MyWidget();
};
MyWidget::ScopedRef::ScopedRef(WeakRef& ref) : ptr(NULL), lock(true) {
_ref = ref.ref.lock();
if(_ref) {
lock = (QThread::currentThread() == qApp->thread());
if(lock) _ref->mutex.lock();
ptr = _ref->ptr;
}
}
MyWidget::ScopedRef::~ScopedRef() {
if(_ref && lock)
_ref->mutex.unlock();
}
MyWidget::~QtBaseWidget() {
selfRef->reset();
selfRef.reset();
}
MyWidget::MyWidget() {
selfRef = boost::shared_ptr<LockedRef>(new LockedRef(*this));
}
Now, everywhere I need to pass around a MyWidget pointer, I'm using:
MyWidget::WeakRef widget;
And I can use it from another thread like this:
MyWidget::ScopedRef widgetRef(widget);
if(widgetRef)
widgetRef->doSth();
This is safe. As long as ScopedRef exists, MyWidget cannot be deleted. It will block in its destructor. Or it is already deleted and ScopedRef::ptr == NULL.
I think a good way of implementing a state machine is to use the singleton pattern.
For example it can look like this:
class A
{
private:
friend class State;
State* _state;
void change_state(State* state) { _state = state; }
};
class State
{
virtual void action(A *a) = 0;
private:
void change_state(A *a, State *state) { a->change_state(state); }
};
class StateA : public State
{
public:
static State* get_instance()
{
static State *state = new StateA;
return state;
}
virtual void action(A *a) { change_state(a, StateB::get_instance(); }
};
class StateB : public State
{
public:
...
virtual void action(A *a) { change_state(a, StateA::get_instance(); }
};
My question is: I have read many articles about that the singleton pattern is so evil. Implementing this without a singleton pattern you have to call new everytime you change state, so for those who dont like singleton, how would you implement the state machine pattern?
I don't think that the singleton pattern is appropriate here. Singletons are good for representing abstract entities or physical objects for which there really is only one copy. To steal an example from Java, there is only one runtime environment in which a particular instance of the program executes. Singletons are good for representing these objects because they give the entire program the ability to name and reference it while preserving encapsulation and allowing for multiple possible backends.
Given this, I disagree that the singleton is the best route to take for your state machine. If you do implement it as a singleton, you're saying that is always exactly one copy of that state machine. But what if I want to have two state machines running in parallel? Or no state machines at all? What if I want my own local state machine so I can experiment on it to see what happens to it? If your state machine is a singleton, I can't do any of these things because there really is only one state machine used by the entire program.
Now, depending on how you're using the state machine, perhaps it is appropriate. If the state machine controls the overall execution of the program, then it might be a good idea. For example, if you're developing a video game and want a state machine to control whether you're in the menu, or in a chat area, or playing the game, then it would be totally fine to have a singleton state machine because there is only one logical state of the program at any time. From your question, though, I can't deduce if this is the case.
As for how to implement the state machine without a singleton, you might want to make the state machine object allocate its own copy of every state and build up the transition table (if you need explicit state objects), or just have a giant switch statement and a single enumerated value controlling what state you're in. If you have a single instance of the state machine this is no less efficient than the current version, and if you have multiple instances it allows you to store local information in each state without polluting a global copy of the states that could be read by other parts of the program.
Your StateA, StateB classes have no data members. Presumably other states won't have modifiable data members either, since if they did then that state would be weirdly shared between different instances of A, that is different state machines running concurrently.
So your singletons have avoided half of the problem with the pattern (global mutable state). In fact with only a small change to your design, you could replace the state classes with functions; replace the pointers to their instances with function pointers; and replace the virtual call to action with a call through the current function pointer. If someone gives you a lot of hassle for using singletons, but you're confident that your design is correct, you could make this minor change and see if they notice that their "correction" has made no significant difference at all to the design.
The other half of the problem with singletons still wouldn't be fixed though, and that's fixed dependencies. With your singletons, it is not possible to mock StateB in order to test StateA in isolation, or to introduce flexibility when you want to introduce a new state machine to your library which is the same as the current one except that StateA goes to StateC instead of StateB. You may or may not consider that a problem. If you do, then rather than making each state a singleton, you need to make things more configurable.
For example, you could give each state some identifier (a string or perhaps a member of an enum), and for each identifier register a State* somewhere in class A. Then rather than flipping to the singleton instance of StateB, StateA could could flip to whatever state object is used to represent "state B" in this state machine. That could then be a test mock for certain instances. You would still call new once per state per machine, but not once per state change.
In effect, this is still the strategy pattern for class A as in your design. But rather than having a single strategy to move the state machine forward, and continually replace that as the state changes, we have one strategy per state the machine passes through, all with the same interface. Another option in C++, that will work for some uses but not others, is to use (a form of) policy-based design instead of strategies. Then each state is handled by a class (provided as a template argument) rather than an object (set at runtime). The behavior of your state machine is therefore fixed at compile time (as in your current design), but can be configured by changing template arguments rather than by somehow altering or replacing the class StateB. Then you don't have to call new at all - create a single instance of each state in the state machine, as a data member, use a pointer to one of those to represent the current state, and make a virtual call on it as before. Policy-based design doesn't usually need virtual calls, because usually the separate policies are completely independent, whereas here they implement a common interface and we select between them at runtime.
All of this assumes that A knows about a finite set of states. This may not be realistic (for example, A might represent an all-purpose programmable state machine that should accept an arbitrary number of arbitrary states). In that case, you need a way to build up your states: first create an instance of StateA and an instance of StateB. Since each state has one exit path, each state object should have one data member which is a pointer to the new state. So having created the states, set the StateA instances "next state" to the instance of StateB and vice-versa. Finally, set A's current state data member to the instance of StateA, and start it running. Note that when you do this, you are creating a cyclic graph of dependencies, so to avoid memory leaks you might have to take special resource-handling measures beyond reference-counting.
In your code, you're not associating a state with the state machine the state belongs to (assuming that class A is the state machine). This information is passed in to the action method. So, if you had two instances of class A (i.e. two state machines) then you could end up having a state update the wrong state machine.
If you're doing this to avoid repeated calls to new and delete for speed purposes, then this is probably a premature optimisation. A better solution, if you can show that using new and delete is too slow / causes other issues (memory fragmentation for example), is to define an operator new / delete in the State base class that allocates from its own memory pool.
Here's some pseudocode for how the state machine I'm currently using works:
class StateMachine
{
public:
SetState (State state) { next_state = state; }
ProcessMessage (Message message)
{
current_state->ProcessMessage (message);
if (next_state)
{
delete current_state;
current_state = next_state;
next_state = 0;
}
}
private:
State current_state, next_state;
}
class State
{
public:
State (StateMachine owner) { m_owner = owner; }
virtual ProcessMessage (Message message) = 0;
void *operator new (size_t size) // allocator
{
return memory from local memory pool
}
void operator delete (void *memory) // deallocator
{
put memory back into memory pool
}
protected:
StateMachine m_owner;
};
class StateA : State
{
public:
StateA (StateMachine owner) : State (owner) {}
ProcessMessage (Message message)
{
m_owner->SetState (new StateB (m_owner));
}
}
The memory pool could be an array of chunks of memory, each big enough to hold any State, with a pair of lists, one for the allocated blocks and one for the unallocated blocks. Allocating a block then becomes a process of removing a block from the unallocated list and adding it to the allocated list. Freeing is then the reverse process. I think the term 'free list' for this type of allocation strategy. It is very fast but has some wasted memory.
One approach which assumes that all state objects live along StateMachine could be like this one:
enum StateID
{
STATE_A,
STATE_B,
...
};
// state changes are triggered by events
enum EventID
{
EVENT_1,
EVENT_2,
...
};
// state manager (state machine)
class StateMachine
{
friend StateA;
friend StateB;
...
public:
StateMachine();
~StateMachine();
// state machine receives events from external environment
void Action(EventID eventID);
private:
// current state
State* m_pState;
// all states
StateA* m_pStateA;
StateB* m_pStateB;
...
void SetState(StateID stateID);
};
StateMachine::StateMachine()
{
// create all states
m_pStateA = new StateA(this, STATE_A);
m_pStateB = new StateB(this, STATE_B);
...
// set initial state
m_pState = m_pStateA;
}
StateMachine::~StateMachine()
{
delete m_pStateA;
delete m_pStateB;
...
}
void StateMachine::SetState(StateID stateID)
{
switch(stateID)
{
case STATE_A:
m_pState = m_pStateA;
break;
case STATE_B:
m_pState = m_pStateA;
break;
...
}
}
void StateMachine::Action(EventID eventID)
{
// received event is dispatched to current state for processing
m_pState->Action(eventID);
}
// abstract class
class State
{
public:
State(StateMachine* pStateMachine, StateID stateID);
virtual ~State();
virtual void Action(EventID eventID) = 0;
private:
StateMachine* m_pStateMachine;
StateID m_stateID;
};
class StateA : public State
{
public:
StateA(StateMachine* pStateMachine, StateID stateID);
void Action(EventID eventID);
};
StateA::StateA(StateMachine* pStateMachine, StateID stateID) :
State(pStateMachine, stateID) {...}
void StateA::Action(EventID eventID)
{
switch(eventID)
{
case EVENT_1:
m_pStateMachine->SetState(STATE_B);
break;
case EVENT_2:
m_pStateMachine->SetState(STATE_C);
break;
...
}
}
void StateB::Action(EventID eventID)
{
switch(eventID)
{
...
case EVENT_2:
m_pStateMachine->SetState(STATE_A);
break;
...
}
}
int main()
{
StateMachine sm;
// state machine is now in STATE_A
sm.Action(EVENT_1);
// state machine is now in STATE_B
sm.Action(EVENT_2);
// state machine is now in STATE_A
return 0;
}
In more complex solution StateMachine would have event queue and event loop which would wait for events from the queue and dispatch them to the current state. All time-consuming operations in StateX::Action(...) should run in separate (worker) thread in order to prevent blocking event loop.
A design approach I am considering is to create a state factory that is a singleton, so that more then one state machine can use the state objects produced by the factory.
But this thought has taken me to the idea of implementing my state factory with flyweight pattern and that's where I have stopped.
Basically, I need to research the advantages of implementing the state objects as flyweights and then the advantages of a flyweight design pattern.
I have heard of this state machines using this type of pattern, but not sure if it will work for my needs.
Anyway, I was doing some research and bumped into this post. Just thought I would share...
Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 4 years ago.
Improve this question
What is a good way of dealing with objects and having them talk to each other?
Up until now all my games hobby/student have been small so this problem was generally solved in a rather ugly way, which lead to tight integration and circular dependencies. Which was fine for the size of projects I was doing.
However my projects have been getting bigger in size and complexity and now I want to start re-using code, and making my head a simpler place.
The main problem I have is generally along the lines of Player needs to know about the Map and so does the Enemy, this has usually descended into setting lots of pointers and having lots of dependencies, and this becomes a mess quickly.
I have thought along the lines of a message style system. but I cant really see how this reduces the dependencies, as I would still be sending the pointers everywhere.
PS: I guess this has been discussed before, but I don't know what its called just the need I have.
EDIT: Below I describe a basic event messaging system I have used over and over. And it occurred to me that both school projects are open source and on the web. You can find the second version of this messaging system (and quite a bit more) at http://sourceforge.net/projects/bpfat/ .. Enjoy, and read below for a more thorough description of the system!
I've written a generic messaging system and introduced it into a handful of games that have been released on the PSP as well as some enterprise level application software. The point of the messaging system is to pass only the data around that is needed for processing a message or event, depending on the terminology you want to use, so that objects do not have to know about each other.
A quick rundown of the list of objects used to accomplish this is something along the lines of:
struct TEventMessage
{
int _iMessageID;
}
class IEventMessagingSystem
{
Post(int iMessageId);
Post(int iMessageId, float fData);
Post(int iMessageId, int iData);
// ...
Post(TMessageEvent * pMessage);
Post(int iMessageId, void * pData);
}
typedef float(*IEventMessagingSystem::Callback)(TEventMessage * pMessage);
class CEventMessagingSystem
{
Init ();
DNit ();
Exec (float fElapsedTime);
Post (TEventMessage * oMessage);
Register (int iMessageId, IEventMessagingSystem* pObject, FObjectCallback* fpMethod);
Unregister (int iMessageId, IEventMessagingSystem* pObject, FObjectCallback * fpMethod);
}
#define MSG_Startup (1)
#define MSG_Shutdown (2)
#define MSG_PlaySound (3)
#define MSG_HandlePlayerInput (4)
#define MSG_NetworkMessage (5)
#define MSG_PlayerDied (6)
#define MSG_BeginCombat (7)
#define MSG_EndCombat (8)
And now a bit of an explanation. The first object, TEventMessage, is the base object to represent data sent by the messaging system. By default it will always have the Id of the message being sent so if you want to make sure you have received a message you were expecting you can (Generally I only do that in debug).
Next up is the Interface class that gives a generic object for the messaging system to use for casting while doing callbacks. Additionally this also provides an 'easy to use' interface for Post()ing different data types to the messaging system.
After that we have our Callback typedef, Simply put it expects an object of the type of the interface class and will pass along a TEventMessage pointer... Optionally you can make the parameter const but I've used trickle up processing before for things like stack debugging and such of the messaging system.
Last and at the core is the CEventMessagingSystem object. This object contains an array of callback object stacks (or linked lists or queues or however you want to store the data). The callback objects, not shown above, need to maintain (and are uniquely defined by) a pointer to the object as well as the method to call on that object. When you Register() you add an entry on the object stack under the message id's array position. When you Unregister() you remove that entry.
That is basically it. Now this does have the stipulation that everything needs to know about the IEventMessagingSystem and the TEventMessage object... but this object should Not be changing that often and only passes the parts of information that are vital to the logic dictated by the event being called. This way a player doesn't need to know about the map or the enemy directly for sending events off to it. A managed object can call an API to a larger system also, without needing to know anything about it.
For example: When an enemy dies you want it to play a sound effect. Assuming you have a sound manager that inherits the IEventMessagingSystem interface, you would set up a callback for the messaging system that would accept a TEventMessagePlaySoundEffect or something of that ilk. The Sound Manager would then register this callback when sound effects are enabled (or unregister the callback when you want to mute all sound effects for easy on/off abilities). Next, you would have the enemy object also inherit from the IEventMessagingSystem, put together a TEventMessagePlaySoundEffect object (would need the MSG_PlaySound for its Message ID and then the ID of the sound effect to play, be it an int ID or the name of the sound effect) and simply call Post(&oEventMessagePlaySoundEffect).
Now this is just a very simple design with no implementation. If you have immediate execution then you have no need to buffer the TEventMessage objects (What I used mostly in console games). If you are in a multi-threaded environment then this is a very well defined way for objects and systems running in separate threads to talk to each other, but you will want to preserve the TEventMessage objects so the data is available when processing.
Another alteration is for objects that only ever need to Post() data, you can create a static set of methods in the IEventMessagingSystem so they do not have to inherit from them (That is used for ease of access and callback abilities, not -directly- needed for Post() calls).
For all the people who mention MVC, it is a very good pattern, but you can implement it in so many different manners and at different levels. The current project I am working on professionally is an MVC setup about 3 times over, there is the global MVC of the entire application and then design wise each M V and C also is a self-contained MVC pattern. So what I have tried to do here is explain how to make a C that is generic enough to handle just about any type of M without the need to get into a View...
For example, an object when it 'dies' might want to play a sound effect.. You would make a struct for the Sound System like TEventMessageSoundEffect that inherits from the TEventMessage and adds in a sound effect ID (Be it a preloaded Int, or the name of the sfx file, however they are tracked in your system). Then all the object just needs to put together a TEventMessageSoundEffect object with the appropriate Death noise and call Post(&oEventMessageSoundEffect); object.. Assuming the sound is not muted (what you would want to Unregister the Sound Managers.
EDIT: To clarify this a bit in regards to the comment below:
Any object to send or receive a message just needs to know about the IEventMessagingSystem interface, and this is the only object the EventMessagingSystem needs to know of all the other objects. This is what gives you the detachment. Any object who wants to receive a message simply Register(MSG, Object, Callback)s for it. Then when an object calls Post(MSG,Data) it sends that to the EventMessagingSystem via the interface it knows about, the EMS will then notify each registered object of the event. You could do a MSG_PlayerDied that other systems handle, or the player can call MSG_PlaySound, MSG_Respawn, etc to let things listening for those messages to act upon them. Think of the Post(MSG,Data) as an abstracted API to the different systems within a game engine.
Oh! One other thing that was pointed out to me. The system I describe above fits the Observer pattern in the other answer given. So if you want a more general description to make mine make a bit more sense, that is a short article that gives it a good description.
Hope this helps and Enjoy!
the generic solutions for communication between objects avoiding tight coupling:
Mediator pattern
Observer pattern
Here is a neat event system written for C++11 you can use. It uses templates and smart pointers as well as lambdas for the delegates. It's very flexible. Below you will also find an example. Email me at info#fortmax.se if you have questions about this.
What these classes gives you is a way to send events with arbitrary data attached to them and an easy way to directly bind functions that accept already converted argument types that the system casts and checks for correct conversion prior to calling your delegate.
Basically, every event is derived from IEventData class (you can call it IEvent if you want). Each "frame" you call ProcessEvents() at which point the event system loops through all the delegates and calls the delegates that have been supplied by other systems that have subscribed to each event type. Anyone can pick which events they would like to subscribe to, as each event type has a unique ID. You can also use lambdas to subscribe to events like this: AddListener(MyEvent::ID(), [&](shared_ptr ev){
do your thing }..
Anyway, here is the class with all the implementation:
#pragma once
#include <list>
#include <memory>
#include <map>
#include <vector>
#include <functional>
class IEventData {
public:
typedef size_t id_t;
virtual id_t GetID() = 0;
};
typedef std::shared_ptr<IEventData> IEventDataPtr;
typedef std::function<void(IEventDataPtr&)> EventDelegate;
class IEventManager {
public:
virtual bool AddListener(IEventData::id_t id, EventDelegate proc) = 0;
virtual bool RemoveListener(IEventData::id_t id, EventDelegate proc) = 0;
virtual void QueueEvent(IEventDataPtr ev) = 0;
virtual void ProcessEvents() = 0;
};
#define DECLARE_EVENT(type) \
static IEventData::id_t ID(){ \
return reinterpret_cast<IEventData::id_t>(&ID); \
} \
IEventData::id_t GetID() override { \
return ID(); \
}\
class EventManager : public IEventManager {
public:
typedef std::list<EventDelegate> EventDelegateList;
~EventManager(){
}
//! Adds a listener to the event. The listener should invalidate itself when it needs to be removed.
virtual bool AddListener(IEventData::id_t id, EventDelegate proc) override;
//! Removes the specified delegate from the list
virtual bool RemoveListener(IEventData::id_t id, EventDelegate proc) override;
//! Queues an event to be processed during the next update
virtual void QueueEvent(IEventDataPtr ev) override;
//! Processes all events
virtual void ProcessEvents() override;
private:
std::list<std::shared_ptr<IEventData>> mEventQueue;
std::map<IEventData::id_t, EventDelegateList> mEventListeners;
};
//! Helper class that automatically handles removal of individual event listeners registered using OnEvent() member function upon destruction of an object derived from this class.
class EventListener {
public:
//! Template function that also converts the event into the right data type before calling the event listener.
template<class T>
bool OnEvent(std::function<void(std::shared_ptr<T>)> proc){
return OnEvent(T::ID(), [&, proc](IEventDataPtr data){
auto ev = std::dynamic_pointer_cast<T>(data);
if(ev) proc(ev);
});
}
protected:
typedef std::pair<IEventData::id_t, EventDelegate> _EvPair;
EventListener(std::weak_ptr<IEventManager> mgr):_els_mEventManager(mgr){
}
virtual ~EventListener(){
if(_els_mEventManager.expired()) return;
auto em = _els_mEventManager.lock();
for(auto i : _els_mLocalEvents){
em->RemoveListener(i.first, i.second);
}
}
bool OnEvent(IEventData::id_t id, EventDelegate proc){
if(_els_mEventManager.expired()) return false;
auto em = _els_mEventManager.lock();
if(em->AddListener(id, proc)){
_els_mLocalEvents.push_back(_EvPair(id, proc));
}
}
private:
std::weak_ptr<IEventManager> _els_mEventManager;
std::vector<_EvPair> _els_mLocalEvents;
//std::vector<_DynEvPair> mDynamicLocalEvents;
};
And the Cpp file:
#include "Events.hpp"
using namespace std;
bool EventManager::AddListener(IEventData::id_t id, EventDelegate proc){
auto i = mEventListeners.find(id);
if(i == mEventListeners.end()){
mEventListeners[id] = list<EventDelegate>();
}
auto &list = mEventListeners[id];
for(auto i = list.begin(); i != list.end(); i++){
EventDelegate &func = *i;
if(func.target<EventDelegate>() == proc.target<EventDelegate>())
return false;
}
list.push_back(proc);
}
bool EventManager::RemoveListener(IEventData::id_t id, EventDelegate proc){
auto j = mEventListeners.find(id);
if(j == mEventListeners.end()) return false;
auto &list = j->second;
for(auto i = list.begin(); i != list.end(); ++i){
EventDelegate &func = *i;
if(func.target<EventDelegate>() == proc.target<EventDelegate>()) {
list.erase(i);
return true;
}
}
return false;
}
void EventManager::QueueEvent(IEventDataPtr ev) {
mEventQueue.push_back(ev);
}
void EventManager::ProcessEvents(){
size_t count = mEventQueue.size();
for(auto it = mEventQueue.begin(); it != mEventQueue.end(); ++it){
printf("Processing event..\n");
if(!count) break;
auto &i = *it;
auto listeners = mEventListeners.find(i->GetID());
if(listeners != mEventListeners.end()){
// Call listeners
for(auto l : listeners->second){
l(i);
}
}
// remove event
it = mEventQueue.erase(it);
count--;
}
}
I use an EventListener class for the sake of convenience as base class for any class that would like to listen to events. If you derive your listening class from this class and supply it with your event manager, you can use the very convenient function OnEvent(..) to register your events. And the base class will automatically unsubscribe your derived class from all events when it is destroyed. This is very convenient since forgetting to remove a delegate from event manager when your class is destroyed will almost certainly cause your program to crash.
A neat way to get a unique type id for an event by simply declaring a static function in the class and then casting it's address into an int. Since every class will have this method on different addresses, it can be used for unique identification of class events. You can also cast typename() to an int to get a unique id if you want. There are different ways to do this.
So here is an example on how to use this:
#include <functional>
#include <memory>
#include <stdio.h>
#include <list>
#include <map>
#include "Events.hpp"
#include "Events.cpp"
using namespace std;
class DisplayTextEvent : public IEventData {
public:
DECLARE_EVENT(DisplayTextEvent);
DisplayTextEvent(const string &text){
mStr = text;
}
~DisplayTextEvent(){
printf("Deleted event data\n");
}
const string &GetText(){
return mStr;
}
private:
string mStr;
};
class Emitter {
public:
Emitter(shared_ptr<IEventManager> em){
mEmgr = em;
}
void EmitEvent(){
mEmgr->QueueEvent(shared_ptr<IEventData>(
new DisplayTextEvent("Hello World!")));
}
private:
shared_ptr<IEventManager> mEmgr;
};
class Receiver : public EventListener{
public:
Receiver(shared_ptr<IEventManager> em) : EventListener(em){
mEmgr = em;
OnEvent<DisplayTextEvent>([&](shared_ptr<DisplayTextEvent> data){
printf("It's working: %s\n", data->GetText().c_str());
});
}
~Receiver(){
mEmgr->RemoveListener(DisplayTextEvent::ID(), std::bind(&Receiver::OnExampleEvent, this, placeholders::_1));
}
void OnExampleEvent(IEventDataPtr &data){
auto ev = dynamic_pointer_cast<DisplayTextEvent>(data);
if(!ev) return;
printf("Received event: %s\n", ev->GetText().c_str());
}
private:
shared_ptr<IEventManager> mEmgr;
};
int main(){
auto emgr = shared_ptr<IEventManager>(new EventManager());
Emitter emit(emgr);
{
Receiver receive(emgr);
emit.EmitEvent();
emgr->ProcessEvents();
}
emit.EmitEvent();
emgr->ProcessEvents();
emgr = 0;
return 0;
}
This probably does not only apply to game classes but to classes in the general sense. the MVC (model-view-controller) pattern together with your suggested message pump is all you need.
"Enemy" and "Player" will probably fit into the Model part of MVC, it does not matter much, but the rule of thumb is have all models and views interact via the controller. So, you would want to keep references (better than pointers) to (almost) all other class instances from this 'controller' class, let's name it ControlDispatcher. Add a message pump to it (varies depending on what platform you are coding for), instantiate it firstly (before any other classes and have the other objects part of it) or lastly (and have the other objects stored as references in ControlDispatcher).
Of course, the ControlDispatcher class will probably have to be split down further into more specialized controllers just to keep the code per file at around 700-800 lines (this is the limit for me at least) and it may even have more threads pumping and processing messages depending on your needs.
Cheers
Be careful with "a message style system", it probably depends on implementation, but usually you would loose static type checking, and can then make some errors very difficult to debug. Note that calling object's methods it is already a message-like system.
Probably you are simply missing some levels of abstraction, for example for navigation a Player could use a Navigator instead of knowing all about the Map itself. You also say that this has usually descended into setting lots of pointers, what are those pointers? Probably, you are giving them to a wrong abstraction?.. Making objects know about others directly, without going through interfaces and intermediates, is a straight way to getting a tightly coupled design.
Messaging is definitely a great way to go, but messaging systems can have a lot of differences. If you want to keep your classes nice and clean, write them to be ignorant of a messaging system and instead have them take dependencies on something simple like a 'ILocationService' which can then be implemented to publish/request information from things like the Map class. While you'll end up with more classes, they'll be small, simple and encourage clean design.
Messaging is about more than just decoupling, it also lets you move towards a more asynchronous, concurrent and reactive architecture. Patterns of Enterprise Integration by Gregor Hophe is a great book that talks about good messaging patterns. Erlang OTP or Scala's implementation of the Actor Pattern have provided me with a lot of guidance.
#kellogs suggestion of MVC is valid, and used in a few games, though its much more common in web apps and frameworks. It might be overkill and too much for this.
I would rethink your design, why does the Player need to talk to Enemies? Couldn't they both inherit from an Actor class? Why do Actors need to talk to the Map?
As I read what I wrote it starts to fit into an MVC framework...I have obviously done too much rails work lately. However, I would be willing to bet, they only need to know things like, they are colliding with another Actor, and they have a position, which should be relative to the Map anyhow.
Here is an implementation of Asteroids that I worked on. You're game may be, and probably is, complex.