It’s too bad we can’t talk with him on this board since he’s gone…
Browsing the source tree I see there are many, many things thare not used because they’re not known or things done twice (ex : Angle and Distance Interpolators, implemented the first time by David Yazel and the second time by David Wallace Croft as behaviors), e.g. the stripifier, a whole log system, a system of “properties”, an occluding system (is this really used by default ?), an integrated Profiler !!!, and gosh what’s that Octree thing, is it used at all ?
All of that is great but these features have fall in the dark while we could be all using it…
So if someone can contact David Yazel he could explain us what he was just about to do when he exited. So we can do it ourselves, heh ?
I think it would be a good idea to look through the whole code and clean it up. When I worked on the View class two days ago, I found many many unused variables and some kind of strange CamaraMode class (no offense, Will), which could actually be replaced by en Enum. There are some methods returning an Enumeration, where an iterator or direct access to the Vector/HashMap/etc. whould be better. Enumeration is deprecated. There are so many things to be cleaned up. And while we clean it up, we can understand some parts. Maybe eliminate not used/needed classes, rewrite badly grown parts to have it better.
Just to answer your question Amos, no I don’t know David Yazel. So I can’t help in that point. Sorry.
Yes, i know how this one came in. But why didn’t you use normal int for the constants? The private String fields filled by the constructor are never used. Do you have a problem with it, whan I modify it to be a real Enum?
But I think, a class returning a Vector or HashMap or whatever by a method is much more useful than one returning just an enumeration, because you safe coding (only one method infront of at least two or three) and you can make use of
Enumeration is the Iterator pendant. Both offer the functionality to loop over all elements once. As soon as hasNext/hasMoreElements return false, the are useless (except for a pontential deletion of the last element by Iterator’s remove method). On the other hand, Iterable allows you to retrieve an unlimited number of Iterators. if every element in a first instance is accessed during a loop, you can get another Iterator and so on. And as long you don’t make use of the remove method, it should be fine to have more the one Iterator at a time…
Enumeration is not depricated but the javadoc says:
[quote]NOTE: The functionality of this interface is duplicated by the Iterator interface. In addition, Iterator adds an optional remove operation, and has shorter method names. New implementations should consider using Iterator in preference to Enumeration.
[/quote]
Xith3d has a lot of unused classes. I made an experience of removing about 60%-70% of the code (from package dependency analysis with Eclipse) and i was able to still run the basic examples.
My advice is that this would be a great time do fully rewrite this engine. There are great ideas in it. It’s possible to improve it by passing iterators to the Renderer algorithms to make rendering independent of the scene graph data structure and also having the possibility to make a truly generic geom and texture loaders.
It’s also a great opurtunity to create full UML documentation for the core of the engine redesign it entirely. It’s not possile to work on an engine without documenting it’s core very well and then build upon it.
“If some others (including me) were willing to do so, would you be involved ?”
Sure, im looking for a new engine with the characteristics i described above. Im finishing my math api. It’s taking some time because im doing univ exams at the moment but that would be a starting contribution. A solid math api would be very important for a game engine foundation.
After that I would be interested on working on a generic loader that supports skeletal anim and shaders and doing a very simple scripting language similar to Oblivion scriping language. I already have a design for the scripting language. If anyone is interested i can put it on a thread.
PS: Do you guys want to keep the name of the engine. Xith doesn’t sound very pro. ;D What about xeon (or zeon?) ? It’s the name of the human colony in the matrix movies if i spelled it right.
By the way if you guys aren’t too convinced about the benefits of separating algorithms from a data structure check out what they did with the boost graph api (c++ stuff).
Applying a similar design to a game engine would mean the renderer would not touch the scenegraph structure directly but use an iterator instead or a more complex accessor class. This is also used in the design of JDSL. This would mean the O(M+N) redution in terms in complexity where M would be different scenes graphs (vrml, java3d, a basic scene graph, etc), N different renderes (software trough java2d cool for applets, jogl, other. etc… ) and a much easier time for loaders to build a scenegraph.
Another interesting point to separate the game engine and allow for extension would be a task manager api for the engene that would drive the renderer.
I like this idea of puting important algorithms into function objects to manage them. Im working on function objects for my math api that will be used later for the scripting language. It could be applied to the renderer. In the same way a set can use a separate sorting algorithms the render could take advantage of their own separate sorting algorithms to deal with primitives (atoms?) that are read from a scene graph indirectly using an accessor. Another great point to apply extension.
I tend to agree with this. Parametrized classes is the weakest form of extension but Java supports templates and a lot of other new things that come before xith was made.
An engine needs a good api for collections. Java util apis just do the bare essential and don’t provide very flexible implementations and essential data structures like graphs which are needed for the A* algorithm and pathfinding. The JDSL api is very good, made by a group of univ researchers for their book.
I would be very curious about your math lib, since the idea of separating data structures and algorithms is the main idea behind the math-lib I’m working on. see here for more information. P.S. I’m not using Xith and probably I’m not planning to do so, but I’d be interested in collaboration.
I also implemented character animation system, which can deform a mesh using a skeleton AND allows morh-targets to be applied to it (I use those for mimics and lip-synchronization). You can download a video here. Further, deformation can be either done on the CPU or the GPU, where the techniques are used, I described in my article ‘Real-Time Character Animation on the GPU’ published in ShaderX^4. I also plan to open source this, if there will be a collaboration with my math-lib… - I don’t want to open source all my work if nothing comes back.
Finally, I’m glad to say that a paper I wrote last time was accepted for the IVA’06, which deals with a system for authoring agents behavior. It uses an XML-based format, into which every schripting language supported by the the javax.script framwork can be embedded.
Why not change the engine name but then it wouldn’t be an evolution and we should start a really seperate project. (If you feel like it, go request it on dev.java.net so that you’re sure it get approved within 6 monhts ;D ;D )
I have always been
Seems really interesting !
I agree.
Please do : there’s really many chances for your software to be improved if your open-source it (and there’s no sense to open-source it in 4 years just when the technology is widespread, like Id Software is doing). If your API is really interesting (and what I saw makes me say it is) then some people will use it and contributes back.
…
Now these projects seems really cool to me and if it’s started I want to contribute. But :
I use Xith for my current game and I have no plan to change (takes too much time to port)
I have not as much math knowledge as I wanted… however as long as I understand your work I can do documentation, which is really important indeed. I can also provide art for demos if needed.
[quote]Further, deformation can be either done on the CPU or the GPU, where the techniques are used, I described in my article ‘Real-Time Character Animation on the GPU’ published in ShaderX^4. I also plan to open source this, if there will be a collaboration with my math-lib… - I don’t want to open source all my work if nothing comes back.
[/quote]
That would be nice. I have checked your math lib and will have a closer look at it. When i complete my own math lib it will be GPL too so you can just take anything you want and add to your own. I will post it in these forums when i have something as well as any other contribution like the scripting language im planning.
[quote="<MagicSpark.org [ BlueSky ]>,post:16,topic:27707"]
Why don’t use groovy or beanshell ?
[/quote]
I will make that scripting language article. I will explain it there.
Thats ok. Both projects are open-source so we can all benefit with this. I don’t have any intention of charging any money code. It’s much more interesting to charge money for content than software when creating a game. IMO.
That reminds me of another important issue. Is it possible to create a project at dev.java.net for a collection of textures, models and other public domain art sets? It’s realy a pain to have to do your own art set just when you want a simple demo to test a scene fo the engine. I think that a public domain collection of textures, 3d models, animations, sounds, music, fonts would solve this problem. Maybe even have a forum just for this so that people can post their free art and submit it to a cvs repo.
(Note : you cited me as yourself : see it’s quote author=zingbat ^^)
We began something like that at MagicSpark, but hey our artists just started (~6 months) modelling ;D often I rant at them cause their models are too high-poly and normals are wrong, and all that. It’s hosted on a forum so not very convenient. A dev.java.net project would be much better (with subversion).
Should I ask for it or will you do that (if I do I can grant you Owner access right after it get accepted).
[quote="<MagicSpark.org [ BlueSky ]>,post:18,topic:27707"]
Should I ask for it or will you do that (if I do I can grant you Owner access right after it get accepted).
[/quote]
I never managed one of these projects so i would prefer someone else did it. How much cvs space can we use for a free art repository like this? Textures can easly eat up to 1Gb of space.
My best source for quality textures is this:
NVidia texture vol. 1
and
This is also a good source for testures and models:
Perhaps it would be better to copy all the textures in the linked sites and reorganize them using our own directory structure and naming conventions. We can never know when one of these sites will be down or closed permanently when we need something.
I never managed one of these projects so i would prefer someone else did it. How much cvs space can we use for a free art repository like this? Textures can easly eat up to 1Gb of space.
My best source for quality textures is this:
NVidia texture vol. 1
and
This is also a good source for testures and models:
Perhaps it would be better to copy all the textures in the linked sites and reorganize them using our own directory structure and naming conventions. We can never know when one of these sites will be down or closed permanently when we need something.
[/quote]
I agree with you yet we should ask them permission if they’re not in public domain / creative commons.
