Use xtext to replace io system

[monochromata]

No description provided.

[amharrison]

The current thinking is a slimmer format (than the xml). We will build both the general AST (EMF/Ecore) for all models, and then this new syntax (Xtext linked to the EMF). For a first pass, it will probably be easier to build the Xtext parser and then create a custom model-to-model transform to convert the new AST into the current system. This will permit us to continue to use the existing tooling, while we incrementally upgrade.

The format I'm proposing is like this:

semanticLabel(parameters...)[
 //symbolic contents
]
{
 //subsymbolic contents
}

example:

model(modelName)[
   module(org.jactr.core.declarative.ClassName)[] {RetrievalThreshold=0, EnabledPartialMatching=true}
  ....

   declarative()[
    chunktype(chunkTypeName, parent)[slot1=null, slot2=null]{}
    chunk(chunkName, chunkTypeName)[slot1=1, slot2=2]{CreationCycle=-1}
   ]{}

  procedural()[
   production(pName)[
    match(goal, chunkType)[
        slot1>1, slot2<2]
    match(retrieval, otherType)[slot1!=2]

    add(goal, chunkType)[]
    remove(retrieval)[]
    ]{Utility=1}
  ]{}
]
{
  EnabledPersistentExecution = true,
  EnableCycleSkipping = true
}

[monochromata]

I'm not sure if this issue might better fit jactr-eclipse instead of jactr. Anyways, I found a number of notes in org.jactr.eclipse.core that might be kept in mind when re-implementing the editors:

-Editor templates!!

• quick production, chunk, chunktype templates for both languages

-Quick Fix

• syntax errors

-Code completion

• for both languages!

-Coloring

• likely requires snagging the entire token stream [change to jactr.io]