Question

I’m having a trouble about rewrite rule to convert from parsing tree into AST tree in antlr.

Here’s my antlr code:

grammar MyGrammar;

options {
  output= AST;
  ASTLabelType=CommonTree;
  backtrack = true;
}


tokens {
    NP;
    NOUN;
    ADJ;
}

//NOUN PHRASE
np  :    ( (adj)*  n+ (adj)*  -> ^(ADJ adj)*  ^(NOUN n)+ ^(ADJ adj)* )
    ;


adj : 'adj1'|'adj2';
n   : 'noun1';

When I input “adj1 noun1 adj2” , the result of parse tree like this:

But the AST tree after rewrite rule seem not exactly like the parse tree, the adj is double and not in order, like this:

So my question is how can I rewrite rule to have a result like the parsing tree above?

Answer 1

Your noun phrase rule collects all the adjectives and copies them to both sides of the nouns because ANTLR can’t automatically distinguish between one group of matched adjs and another.

Here is a break-down of the np rule:

np  :    ( 
           (adj)*  //collect some adjectives
             n+ 
           (adj)*  //collect some more adjectives 
               -> ^(ADJ adj)*  //all adjectives written
                  ^(NOUN n)+   //all nouns written
                  ^(ADJ adj)*  //all adjectives written again
         )
    ;

One way to separate the two groups is to collect them into their own respective lists. Here’s an example, applied to rule np:

np  :    ( 
           (before+=adj)*  //collect some adjectives into "before"
             n+ 
           (after+=adj)*  //collect some adjectives into "after"
               -> ^(ADJ $before)*  //"before" adjectives written
                  ^(NOUN n)+   //all nouns copied
                  ^(ADJ $after)*  //"after" adjectives written
         )
    ;

This way ANTLR knows which adjs to write out before and after the ns.