Question

Given the gramar

grammar T;

options
{
    k=4;
  language=CSharp3;     
  TokenLabelType=CommonToken; 
  output=AST;   
  ASTLabelType=CommonTree;  
}

tokens 
{
  LPAREN = '(';
  RPAREN = ')';
  LBRACK = '{';
  RBRACK = '}';
}

fragment
ID  :   ('a'..'z'|'A'..'Z'|'_')('a'..'z'|'A'..'Z'|'0'..'9'|'_')*;
WS : (' ' | '\t' | '\n' |'\r' )+ { $channel = Hidden; } ;

public program: CLASSDEF+ EOF!   ;

CLASSDEF: 'class' ID LBRACK
       RBRACK     ;

This produces a lexer and a parser which I use as follows

using System;
using Antlr.Runtime;
using Antlr.Runtime.Tree;

namespace compiler
{
  internal class Program2
  {
    public static void Main(string[]arg)
    {
      ANTLRStringStream Input = new ANTLRStringStream(@"class foo 
{ 
}");
      TLexer lex = new TLexer(Input);
      Console.WriteLine("errors:" + lex.NumberOfSyntaxErrors);
      CommonTokenStream tokens = new CommonTokenStream(lex);
      TParser parser = new TParser(tokens);

      var parsed = parser.program();
      Console.WriteLine("errors: " + parser.NumberOfSyntaxErrors);
      CommonTree tree = parsed.Tree;


      Console.WriteLine("type:" + tree.Type);
      Console.WriteLine("text:" + tree.Text);
      Console.WriteLine("children:" +tree.ChildCount);
      Console.WriteLine(tree.ToString());
      Console.WriteLine(tree.ToStringTree());

      Console.ReadKey();
    }
  }
}

When running this code I get 0 lex errors and 1 parse error

result

errors:0
errors: 1
type:0
text:{
}
children:0
<error: {
}>
<error: {
}>

Questions!

1. I thought ANTLR was supposed to give intelligent error messages, yet I fail to find out whats wrong

2. Am I missing code to improve on the error messages?

Answer 1

You made CLASSDEF a lexer rule (in other words: a single token), which is incorrect. When the lexer stumbles upon input like "class X", it cannot create a CLASSDEF token because there’s a space between "class" and "X" (and no, the WS token will not help you with this since CLASSDEF is a lexer rule).

In other words: make CLASSDEF a parser rule instead (and remove fragment from ID!):

grammar T;

options
{
  language=CSharp3;     
  output=AST; 
}

tokens 
{
  CLASS  = 'class';
  LPAREN = '(';
  RPAREN = ')';
  LBRACK = '{';
  RBRACK = '}';
}

public program
 : class_def+ EOF!
 ;

class_def
 : CLASS ID LBRACK RBRACK
 ;

ID
 : ('a'..'z'|'A'..'Z'|'_') ('a'..'z'|'A'..'Z'|'0'..'9'|'_')*
 ;

WS
 : (' ' | '\t' | '\n' |'\r' )+ { $channel = Hidden; } 
 ;

Now parsing input like "class foo { }" will produce the following parse: