Slick Forums

Doesn't slick already use two threads so adding another 4 threads leaving 2 that will be waiting to run at some point in time. Then again the wait would be minimal.

Can anyone confirm this for me?

_________________
My Github - https://github.com/warmwaffles

Honestly, I think going parallel is a waste of time for most games. You're better off optimizing for single-thread, and perhaps put special purpose stuff into a separate thread, like pathfinding.

Sorry if I misinformed people about what I am doing. I was using Artemis server side. Most of the game isn't made yet.

I had a interval system that needed to be executed no matter what at exactly the right interval without waiting for itself to finish. I removed the system since there are alternatives.

I've got a question on the best way to implement something.

I've got a Transform component which has variables for x, y and angle. I also have a Physics component which contains a Box2D body. The Box2D body also contains x, y and angle variables.

Each update in the PhysicsSystem, I update the Transform component with the position and rotation of the Box2D body from the Physics component. This works well because it means other systems (like rendering and camera following) can just use the Transform component and don't have to know about the Physics component. This also allows entities that don't have a physics body to have a position.

The problem I have now is how to easily set the position of an entity. I could change the data in the Transform component, but this would be overwritten in the next frame by the PhysicsSystem. I need a way to notify the Physics component when the data in the Transform component is modified.

The simplest way would be to just test if the entity has a Physics component when setting a position and update both. This seems easy, but it relies on the developer remembering to update the position for both.

I don't think putting a method to set the position in the Entity class is possible (or correct) and components shouldn't contain logic.

I could add a ChangePosition component temporarily to the entity and have a ChangePositionSystem that updates both the Transform and Physics components of the entity. This seems really hacky..

Is there a better way of doing this?

Dunno the best way, I can't share mine. I had almost the same problem you have, and the way I am handling it right now is by abstracting the Transform data outside the Component (I named it Spatial).

So, a Transform component has a Spatial (an interface, no component) with methods like setPosition(x,y), getX(), getY(), setAngle(alpha), more. I have one implementation of the Spatial that uses a Box2d body inside and set stuff accordingly.

With this abstraction, I can call set() or get() to the Spatial object in all places without knowing if the implementation is Box2d or other.

Also, imagine you want entities with no box2d stuff, you can create a Spatial implementation that uses x,y,alpha directly and all your systems and other classes will not know that, because you keep using the interface.

Here are some examples, in my case Transform is called SpatialComponent, then we have the Spatial and finally two implementations, with SpatialPhysicsImpl and SpatialImpl.

Hope it was clear and could be of help.

_________________

It does not make sense to have such redundant information about position.

It's clear that you have two things in conflict with each other, Physics and Transform.

Question is, do you need Transform if you have Physics? Can't you just throw out Transform and just use Physics as it were Transform?

But then you might need to have a entity that is not part of the physics world?

Tricky. I don't have all answers really. This issue exists in other (entity/component) frameworks as well, so I can't really say it's related to the way Artemis is designed.

But my bet would be to somehow merge the two together. Perhaps just a Transform component, which can also have a Body (or not), and then a Transform system that detects if a Transform component has a Body or not to add it into a physics world.

I'm afraid there's really no clean solution if you wish to support both positional non-physics objects and positional physics objects.

I'd like to put in my input on a solution to Kazua's problem.

First of you can design your component however you want and it can have logic.

For example for each Physics component you have to check you could do a boolean check to see if that Physics component is enabled before you process it through your physics system.



This way your logic is in one place and both maintainable and readable.

I've mostly finished the framework, and am now working on the "starwarrior" game.

Artemis C# port is feature complete. Starwarrior was ported too(using bare XNA 4). Check it out:

https://github.com/thelinuxlich/artemis_CSharp
https://github.com/thelinuxlich/starwarrior_CSharp

We are also working on a GUI for it:
http://www.ploobs.com.br/forum/viewtopi ... p=793#p793

I think this engine deserves its own forum or discussion group(Google?)

Method generics don't need to include a argument for type.

The code I saw is:

in your EnemySpawnSystem. e.GetComponent should not need to declare Velocity twice as the type. Even if it is a ComponentType.

If I am mistaken about what you are trying to do then I am sorry ahead of time.

You are right, there were places where I forgot to change this stuff(I was converting StarWarrior together with Artemis), but this is old, I think you are reading a old revision

Hey.

Fun.

I do wonder if you guys are taking advantage of the language features in C# or C++.

Been a while since I did some C# or C++ programming, but perhaps things can be done very differently, I don't know.

I've been doing some microoptimizations and using Dictionary instead of Bag in some places. Other than that, I don't know if there are any more tricks on C# land

I fixed the Tag not being unregistered once its Entity is removed from the World, is it intentional to be unregistered manually?

Ah! I need to fix that

Why not removing entities from group(and also entitymanager) on World.deleteEntity(e)? Sometimes I have problems with a system deleting a entity and after that another system process a group that still has the deleted entity because world.loopStart() is not executed yet.