Question

I am designing a declarative language. In my language are definitions of “nodes” that can be reused later. Think of them like prototypes or templates. You can create copies of them as needed. I have a version of the interpreter that generates the appropriate Java objects during the definition phase. Up until recently I have been using a deep copy library to create copies of the Java representation of the nodes to create instances of the node. Due to the limitations of most of the deep copying libraries I have found and my lack of desire to write a copy constructor for every object in my system, I want to take a different approach. Part of my language specifies GUI Java Swing objects which cause issues with the deep copy libraries.

Is there a way I can save the AST subgraph for my object declaration and reprocess it when I want to create an instance of the node? I am asking the question in terms of ANTLR ASTs, but if you have a language implementation advice I welcome that too.

Answer 1

In an ANTLR parser/lexer grammar, you have access to a $ruleName.tree. This rule attribute is the reference to AST subtree generated during parsing. Using the @after field and the @members field, you can store the subtree. Let me show you.

@members{
private List<CommonTree> defs = new ArrayList<CommonTree>();

public List<CommonTree> getDefs(){
    return defs;
}

}

below in your grammar

ruleName 
@after{
    defs.add($ruleName.tree);
}
: IDENT ;

This will create a list of all the ruleName AST subtrees. These trees can be turned into CommonTreeNodeStream and used with your tree grammar code.

CommonTree defTree = Parser.getDefs().get(0); // just getting the first tree for example
CommonTreeNodeStream defStream = new CommonTreeNodeStream(defTree);
TreeWalker walker = new TreeWalker(defStream);
//assuming that I have a tree grammar that has been defined to return a list of nodes

See this question for more info on returning data from an ANTLR grammar rule.

List<Node> nodes = walker.topRule() 

This will allow you to create node instances by reprocessing saved AST subgraphs. You can create the TreeNodeStream and call the walker whenever you need to create an instance.