Slick Forums

Ah, nice find. I don't know how many are using Artemis, so I surely would like people to let me know.

I might set up a maven repo.

Haha thats mine ^^

I was using my own code until recently (when I first posted here). I've updated the project description with a link to Artemis. Sorry about the confusion.

Hoped to post a demo this week, but I realize now that there is some way to go from "shooting randomly spawning enemies" to game.

Btw. systems can communicate with each other. Check SystemManager.

The github code repo is automatically synced with the svn code repo. Automatic sync could fail of course, so if you want to use the github repo, do it as your own risk.

The idea to have a github code repo is to easily test things by forking the repo. We are already using it.

appel, have you considered working with git instead of svn? it could be a great gain for the ppl who want to test things over the engine and even wanted to make contributions (only proposing, don't want to discuss pros and cons of git or svn, at least not in this post)

_________________

As you have no dependencies in your repo, I really recommend you to upload the engine over maven central instead creating your own maven repo, ruben01 has already uploaded some projects to maven central so he could help you if you want to upload it. It would be tons of times better than having your own maven repo, and easier than someone could think.

Again, it is only a suggestion, if you think you don't want to upload it or maintain your maven repo is easier for you, then it is your choice, I will not argue against .

As an example, we have our maven natives plugin uploaded to central.

_________________

1)
How do you have to use this framework in terms of multiplayer. Could it be imaginable to store all the entities/components in a database like discussed in this article (http://t-machine.org/index.php/2009/10/ ... os-part-5/).

2)
I have still some question about the exact interpretation of an entity. Like for example is a (tiled) map an Entity or not? And an inventory is it an Entity or a component of a player entity.
I searched different definitions but there are so many (different) definition for an entity so I can't really catch the concept.
What is the meaning of an entity in terms of this approach?

It's a great framework and a really nice approach on this paradigm! And I can really feel the power of this approach.

1. In terms of multiplayer, I've started implementing both client and server versions. What I found is that the client and server do not necessarily share the same systems. The server is supposed to run the actual logic, but the client only extrapolation, or a simulation. The server can run without any rendering systems. You can see some idea shaping up in the tankz source code. http://gamadu.com/svn/tankz/

If you are going to do a multiplayer game, I'd start implementing it as a multiplayer game from start, and always have the client version connect to the server version, even if the player is only playing e.g. skirmish vs. the computer.

Sorry, I haven't imagined how to use a database with this framework, as that's really something I'm not really trying to do in my games. But there's really just three things you need to store, the id's of the entity (entityId and uniqueEntityId), the components for each entity, and maybe the state of the systems (if you store internal state for it).


2. The tiled map question is a good one. I would say it's not an entity, but it's really a system. E.g. TiledMapSystem, which would contain public methods that can be used in other systems and also rendering logic.
You could have something as a Entity, e.g. the 2d grid, for the tiled map, but I don't find it to be necessary. It's really up to you, you can have it as a entity+system or just a system.

Thanks for the answers. They are really pointing me in the good direction.



If you want to develop a game with a user database it could be imaginable to link a player/character entity to a user I think. The uniqueEntityId can be really helpful for that in terms of databases! I'm gonna try this out.



Ok so sometimes there is not really "one" good answer. I can choose myself what suits best. I have a lot of game elements where I wonder in which category they belong. But I suppose choosing will be easier with more experience with the framework!

You can see what I've done in Tankz in regards of CameraSystem, BoundarySystem, TerrainRenderSystem.

http://gamadu.com/svn/tankz/trunk/src/c ... ystem.java
http://gamadu.com/svn/tankz/trunk/src/c ... ystem.java
http://gamadu.com/svn/tankz/trunk/src/c ... ystem.java

TerrainRenderSystem uses CameraSystem to render the terrain correctly offset.
CameraSystem uses the BoundarySystem to ensure the camera doesn't go offmap.
BoundarySystem ensures certain entities don't go out of bounds.

How could this have been implemented with entities/components? I could have had a Boundary entity, Camera entity and Terrain entity with corresponding components. Then I would have tagged them and then make the systems process the tagged boundary, camera and terrain entities.

That's perhaps better if you need savegame feature, putting all the data the systems use into components so the systems are really stateless. But you can also store the state of the systems. But I found it to be really redundant.




In a previous framework I used to have systems as entities in the game world. So I would have had a TerrainEntity which would have had a Terrain component with public methods like retrieving a tile. I would have tagged that entity as "TERRAIN" and simply looked it up where I had to query the Terrain component if e.g. tile X,Y was passable or not. I found this way of doing things to be very strange, but it worked.

It's sometimes better to keep things a part, stuff that's not really entities should not be entities, and what you would consider as entities should be entities, e.g. bullets, creatures, explosions, etc. It helps keeping things cleaner and code more understandable.

Thanks for the explanation!

How many entities can be processed in one frame?

Not possible to answer this.

Have you tested it on any other specs?

I have, but don't have any accurate numbers to give you.

Obviously, performance varies between hardware.

My PC is nearly 4 years old.

It's really a question of being performant in this design architecture. You can make a different framework that runs just as fast, but it might not offer your the flexibility or architecture that Artemis does. So, comparing A with B is not reasonable if you don't consider the differences.

Finally squashed the last bug in moving over to Artemis, with the culprit of my woes in deleteEntity:


Double deleting an entity allowed the id to be used twice, resulting in anguish without words! I guess this problem is kind of specific to my code, but I was overcome with happiness when I found the solution.

Thanks for the speed! Is wonderful =) Hopefully I can complete my Slick contest contribution, so the power Artemis in a fully datadriven codebase can be showcased.

Squizzle: My code is far from optimized, but on the speed front I'm running hundreds of collision boxes, thousands of entities and ten-thousands of components without problem. Its easy to work with once you wrap your head around the concept, and implementing new functionality without breaking everything blows OOP so far out of the water its not even funny.