Well thinking of things, using the class file and all of the other such this is going to really depend on many things and will not be that light where it could easily be made into RatufaCoat. I think what I need here is a minimizer of sorts that takes an input class file and turns it into a tinier format for easier loading and usage. Then I just use these minimized classes and such for running things. At least initially I do not have to go crazy in what I am doing, literally just making a mini format. I can read the class file although it would be a bit more complex of a format to parse especially considering the constant pool and all the variable length data. I can keep this in the classfile stuff, but I can have a convert to minified format which uses the same format as the class file except with alternative constant pools. It would be a proprietary format used for SquirrelJME. I mean all of the stuff uses the constant pool. In this case I would really just be using an optimized constant pool for things. I think this will be the first case, at least something that can output a lighter format for this reason. I do not need to go that far for things. Of course minified classes would be completely not compatible with standard JVMs but I could just reuse the format and such. But to signify mini classes, I can have a different major and minor version that is not normaly class file compatible but SquirrelJME knows about it. At least according to the version used. Of course what I think I need first is something which can output the same input classes and such.
ClassFile class is great for things, but it is not something that
would result in a performant format for execution since it is expected that
the JVM will do all the optimization itself. So I think even using this
format is just going to needlessly complicate and slow things down. I
definitely do though want a pre-mapped constant pool reference thing so that
when a class is loaded those things are resolved as needed, at least after
the base class structure has been loaded.
I do know what I need exactly to run a virtual machine and the current format just adds too much information for that. At least initialy for code I do not need to perform much translation for the byte code, it will just be for optimizing things essentially so it can be more easily used by the VM. Of course the end result as a guess should end up being faster with most of the cruft and checks in the format removed.
I should make a test for the actual class file parsing and such just to be sure that it works properly.