Jump to content

  • Log In with Google      Sign In   
  • Create Account


Designing Entity systems without global pointers to subsystems


Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.

  • You cannot reply to this topic
5 replies to this topic

#1 laylay   Members   -  Reputation: 310

Like
0Likes
Like

Posted 29 September 2012 - 10:12 PM

I have a few C++ design questions I'm unsure about.

For a while I've avoided any kind of entity system in my game but now the game is getting pretty large and has a lot of game objects that need to be created. When clients connect to the server, the server sends down every object on the server (players, projectiles, flags) This was messy without some kind of entity factory system because I'd have to know the order of game objects being sent down from the server.

So now I have a base entity class which has a ReadFromPacket method so I can just loop through the number of entities being sent and create each one. Now the problem is that some entities need different arguments in the constructor / draw / update methods because some entities need to use different subsystems (Renderer, Sound, Input) Should I just have global pointers to them subsystems so anything has access to them? Currently I'm just passing them around to things that need it but that doesn't play well with an entity system.

Sponsor:

#2 kd7tck   Members   -  Reputation: 715

Like
0Likes
Like

Posted 29 September 2012 - 11:14 PM

This is only a suggestion, so think what I say over a bit and modify it as needed.

Try using virtual templates with vectors as the means of passing variable number of arguments. Only assign the virtual functions you need to use for each object.

template<typename T>
virtual void Entity::draw(const std::vector<T>&);

Edited by kd7tck, 29 September 2012 - 11:14 PM.


#3 laylay   Members   -  Reputation: 310

Like
0Likes
Like

Posted 29 September 2012 - 11:18 PM

Not sure that would work because I'm looping through all the entities drawing each one. It wouldn't know what to pass in

Another problem I see with a entity based design is that sometimes entities need to be drawn before others like for opaque and translucent drawing. One way around that is to have 2 methods in the base class but that means you're calling a virtual method for things that may not need it.
Maybe I should be queuing up all drawing to be sorted and executed instead.

There's so many situations like that I don't know how to handle with entity based design but the way I'm handling game objects currently is way too much work. To add a new game object I need to write methods to iterate through them for drawing, updating and cleanup in my game state class but I have much more control over what they do.


I'm open to the idea of having global pointers to subsystems because I've seen multiple engines do it that way but I still can't persuade myself to do it. I've always been told it should be avoided.

Edited by laylay, 29 September 2012 - 11:33 PM.


#4 kd7tck   Members   -  Reputation: 715

Like
0Likes
Like

Posted 29 September 2012 - 11:38 PM

Not sure that would work because I'm looping through all the entities drawing each one. It wouldn't know what to pass in

Another problem I see with a entity based design is that sometimes entities need to be drawn before others like for opaque and translucent drawing. One way around that is to have 2 methods in the base class but that means you're calling a virtual method for things that may not need it.
Maybe I should be queuing up all drawing to be sorted and executed instead.

There's so many situations like that I don't know how to handle with entity based design but the way I'm handling game objects currently is way too much work. To add a new game object I need to write methods to iterate through them for drawing, updating and cleanup in my game state class but I have much more control over what they do.


I'm open to the idea of having global pointers to subsystems because I've seen multiple engines do it that way but I still can't persuade myself to do it. I've always been told it should be avoided.


Why have two methods, simply pass in an int ID with the vector. This id will tell the draw method how to draw the entity. As far as knowing what to pass in the vector, that is entirely up to you. You could set all virtual methods to return a bool, if false then that method was never needed by this entity. You are actually going to have to tinker with this for a while, this sort of thing takes a little planning and time. Cheating and using globals to save time, is only going to complicate things down the road.

#5 Kyall   Members   -  Reputation: 287

Like
1Likes
Like

Posted 30 September 2012 - 06:59 AM

// Updater template class
template <typename T>
class EntityUpdater {
public:

// Need to inherit to override this
virtual void Draw();

// Make this a singleton
static EntityUpdater<T> hInstance;
private:
std::vector<Entity> entities;

};

// Inherit it and override the update
class UpdateCat : public EntityUpdater<Cat> {

};


template <typename T>
void Register( void * entity ) {
   // If the singleton for the update of this entity type exists, then add this object	
   if( EntityUpdate<T>::hInstance ) {
	 // Add entity to the entities in EntityUpdate<T>
  }
}


You can then register each entity with Register<MyType> or you can use the template in a function pointer thing.

typedef void (*RegisterFunc)( void* ); 

struct TypeDefinition {
   RegisterFunc reg;
};


then later, and this is if you're using a reflection thing with your entity factory in the constructor for the type's reflection defition

cat.reg = Register<Cat>;

I might me missing an & there or something.
I say Code! You say Build! Code! Build! Code! Build! Can I get a woop-woop? Woop! Woop!

#6 render donkey   Members   -  Reputation: 138

Like
0Likes
Like

Posted 02 October 2012 - 03:32 AM

if your entities has different update and render method. you would like to divive them in to several subsystems.
each kind of your entities keep the subsystem's pointer it would like to use.

and.
each subsystems have their own queue to render and udpate. maybe they have the same base class. it's depend on your design.

maybe it wlll looks like this.


class SubSystemBase{};


class SubSystem0:public SubSystemBase{};
class SubSystem1:public SubSystemBase{};
....
class EnityBase{};

class EntityBaseClass0:publc EntityBase
{
SubSytem0* mSubSystem;
};
class EntityBaseClass1:public EnityBase
{
SubSytem1* mSubSystem;
};


it's only my suggestion. decide buy yourself.




Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.



PARTNERS