Jump to content

  • Log In with Google      Sign In   
  • Create Account


Designing a communication system for a game that uses object aggregation


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
6 replies to this topic

#1 agorglouk   Members   -  Reputation: 102

Like
0Likes
Like

Posted 13 June 2013 - 10:46 AM

I'm making a Component Based Game Engine, and i try to separate data from logic as much as possible by using this design:

 

                                                        Engine
                /                                            |                                                \
EntityManager                    ComponentManager                            SystemManager
(vector of entities)    (map of (entities - vector of components))      (vector of systems)

 

Where essentially all these managers are just holders of Entities, Components, and Systems accordingly.

A brief description of my implementation is:

 

Entity: a class that represents every object in the game (just and UUID).

 

Component : an abstract class that (its subclasses) will hold only data about a specific thing for an Entity, ex. PositionComponent will only have the x,y values of an Entity nothing more.

 

System: a abstract class that (its subclasses) will have all the logic to act on Enities, ex. PositionSystem will have in its Process() method passed by the Engine an Entity to operate on, and change its component values using some predefined logic.

 

My problem is, how do i make objects communicate a senario like this? For example how can i send a message from a System (that resides in the SystemManager) to the ComponentManager so i can get a specific component for the current Entity that the system is processing (i already have the methods in ComponentManagers to do this, thus my problem is what is the PROPER way design to do this)?

 

Or generally a game that is modeled like a Operating System (in layers from the inner kernel to the outer world) how can achieve communication between every layer? Are there any good design patterns that are useful for this purpose?

 

Any help appreciated.


Edited by agorglouk, 13 June 2013 - 10:47 AM.


Sponsor:

#2 phil_t   Crossbones+   -  Reputation: 3108

Like
1Likes
Like

Posted 13 June 2013 - 11:41 AM


My problem is, how do i make objects communicate a senario like this? For example how can i send a message from a System (that resides in the SystemManager) to the ComponentManager so i can get a specific component for the current Entity that the system is processing (i already have the methods in ComponentManagers to do this, thus my problem is what is the PROPER way design to do this)?

 

Ensure that each System has a reference to the ComponentManager, and expose public methods on the ComponentManager to retrieve a specific Component for a specific Entity.



#3 agorglouk   Members   -  Reputation: 102

Like
0Likes
Like

Posted 13 June 2013 - 02:21 PM

Yea , but i'm searching for an implementation that will isolate every thing in the engine, and your idea doesn't seem really structured to me (a little messy to communicate like this). I thought that the most appropriate communication would have been to send some command from the System to the SystemManager and the SystemMAnager to the Engine? and so the Engine to the ComponentManager to retrieve the component. I am not really sure but i think that building a game engine like an Operating System is the most expandable and generic idea in my mind to build on top lots of games. Any other opinions?



#4 Boreal Games   Members   -  Reputation: 773

Like
0Likes
Like

Posted 13 June 2013 - 07:49 PM

For something as simple as this, the obvious answer is often the right thing to do.  You need speed, so a direct lookup of the component is the best thing you can do.



#5 phil_t   Crossbones+   -  Reputation: 3108

Like
0Likes
Like

Posted 13 June 2013 - 10:33 PM


Yea , but i'm searching for an implementation that will isolate every thing in the engine, and your idea doesn't seem really structured to me (a little messy to communicate like this). I thought that the most appropriate communication would have been to send some command from the System to the SystemManager and the SystemMAnager to the Engine? and so the Engine to the ComponentManager to retrieve the component. I am not really sure but i think that building a game engine like an Operating System is the most expandable and generic idea in my mind to build on top lots of games. Any other opinions?

 

Like Boreal Games said, speed is important here. Presumably Systems are going to be batch processing 100s/1000s of entities/components, so they should have a quick way to enumerate through those entities' components.

 

That said, even if performance weren't an issue, without knowing more about your design goals I would still go with the simplest way.

 

If you want to create an expandable/generic engine, you need to come up with scenarios of how you'll want people to plug into it (what functionality is extensible) - only then can you come up with a good design. It sort of sounds from your original post that you want to be able to substitute every single piece of your engine. Is that really your design goal? How do you envision it being used?

 

There's no "right" way to design something - only a "right" way to design something given a set of goals/scenarios.


Edited by phil_t, 13 June 2013 - 10:45 PM.


#6 agorglouk   Members   -  Reputation: 102

Like
0Likes
Like

Posted 14 June 2013 - 09:18 AM

@phil_t:

 

Yes, this is exactly my goal to make an engine design that will be fully expandable and generic, being able to transform really easily to simple mini games (like 2D rpgs, platformers or anything really simple). I know that this is not a really good idea when it comes to performance issues but i'm just exploring the most optimal design possibilities when it comes to extensibility for a game engine. I'm just trying to implement the most ideal design scenario of Component Based Game Engines, sacrificing some performance, just to see if will worths the try to make a really abstract and complicated design for a game base, or if its better to implement a specific base for a specific game when it comes production time of multiple games. Probably i'm wrong, but i wont find out if i wont try to make this and fail/succeed. 

 

The main idea is, if this layer design scenario succeeded in the implementation of operating systems, why it shouldn't succeed in games?

(By saying layers i mean something like this:
layers in OSes: kernel - ... - outer world,
layers in Game Engine: Engine -  Managers - (Systems, Components,...) - ... - outer world  ))

Knowing that today's hardware is developing exponentially(so performance may not be a problem later) and the growth of data to be organised and 

managed too, why not explore less performance related and more organisational related designs?

 

And according to all the above how should i implement a layer to layer communication (the problem is mostly from outer layers (Systems) to inner layers (Engine))

so i can accomplish (or even try to approach) my purpose?



#7 phil_t   Crossbones+   -  Reputation: 3108

Like
0Likes
Like

Posted 14 June 2013 - 10:22 AM

I don't have any suggestions for your layering other than just dependency inversion (using interfaces to communicate between your layers),

 

But I will say that regardless of if you're making a 2D RPG, or a platformer, or whatever, you can probably re-use exactly the same EntityManager/SystemManager/ComponentManager setup (obviously you'll have different specific systems and components though). 

 


Knowing that today's hardware is developing exponentially(so performance may not be a problem later) and the growth of data to be organised and 
managed too, why not explore less performance related and more organisational related designs?

 

Well, RAM speed has not been increasing the same way processor speed has. Relatively slow access to memory has been why data-oriented design and focus on memory access patterns has become more important in recent years - and those design choices may sometimes conflict with traditional OOP design choices.






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