[antlr-interest] disambiguating sempred in a closure?

[Harald M. Müller]

Hi -

As far I can see, the sempreds ARE at the top of the (inner) rule. Below is
the code produced by ANTLR 2.7.5 for C# - I did not run it, but it looks
pretty correct.
And I would not go to ANTLR 3 with an 1100 line grammar, unless you plan
this as a major (and maybe failing) step in your project; I stopped working
on porting a 300 line grammar (lexer + parser) after I found that some
features from ANTLR 2.x are not (yet) present in ANTLR 3.x ...

Regards
Harald M.

	public void prod() //throws RecognitionException,
TokenStreamException
{
		
		returnAST = null;
		ASTPair currentAST = ASTPair.GetInstance();
		AST prod_AST = null;
		
		something();
		astFactory.addASTChild(currentAST, returnAST);
		{    // ( ... )*
			for (;;)
			{
				if (((LA(1)==KERMIT||LA(1)==FOZZIE))&&(
muppet ))		<<< go into this branch if muppet is true
				{
					{
						switch ( LA(1) )
						{
						case KERMIT:
						{
							AST tmp1_AST = null;
							tmp1_AST =
astFactory.create(LT(1));
	
astFactory.addASTChild(currentAST, tmp1_AST);
							match(KERMIT);
							break;
						}
						case FOZZIE:
						{
							AST tmp2_AST = null;
							tmp2_AST =
astFactory.create(LT(1));
	
astFactory.addASTChild(currentAST, tmp2_AST);
							match(FOZZIE);
							break;
						}
						default:
						{
							throw new
NoViableAltException(LT(1), getFilename());
						}
						 }
					}
				}
				else if
(((LA(1)==LEONARDO||LA(1)==DONATELLO))&&( turtle )) {		<<< go into
this branch if turtle is true
					{
						switch ( LA(1) )
						{
						case LEONARDO:
						{
							AST tmp3_AST = null;
							tmp3_AST =
astFactory.create(LT(1));
	
astFactory.addASTChild(currentAST, tmp3_AST);
							match(LEONARDO);
							break;
						}
						case DONATELLO:
						{
							AST tmp4_AST = null;
							tmp4_AST =
astFactory.create(LT(1));
	
astFactory.addASTChild(currentAST, tmp4_AST);
							match(DONATELLO);
							break;
						}
						default:
						{
							throw new
NoViableAltException(LT(1), getFilename());
						}
						 }
					}
				}
				else
<<< exit the loop if none of the above (lookaheads and flags) are true
				{
					goto _loop5_breakloop;
				}
				
			}
_loop5_breakloop:			;
		}    // ( ... )*
		prod_AST = currentAST.root;
		returnAST = prod_AST;
		ASTPair.PutInstance(currentAST);
	}
	

> -----Original Message-----
> From: antlr-interest-bounces at antlr.org 
> [mailto:antlr-interest-bounces at antlr.org] On Behalf Of J Chapman Flack
> Sent: Friday, December 21, 2007 4:52 PM
> To: antlr-interest at antlr.org
> Subject: [antlr-interest] disambiguating sempred in a closure?
> 
> 
> Hi,
> 
> I have backed myself into a corner where I need semantic 
> information to decide whether to take an exit branch:
> 
> prod : something
>         ( options { greedy=true; }
>         : { muppet }? ( KERMIT | FOZZIE )
>         | { turtle }? ( LEONARDO | DONATELLO )
>         )*
>       ;
> 
> so that even if the lookahead is KERMIT or FOZZIE I should 
> take the exit branch if I'm not looking for muppets. (Before 
> you ask, no, I don't know this statically; actions may alter 
> these flags on any iteration. Such is the language I'm trying 
> to parse.)
> 
> antlr 2.7.7 treats these sempreds as validating because they 
> are not at the top of the production, so they throw 
> exceptions instead of exiting the closure. I haven't tried to 
> port my grammar to antlr 3 yet.
> 
> 1.  Is there a way to refactor this production in antlr 2.7.7
>      to get the behavior I want?
> 
> 2.  Is it easier in antlr 3?  Is it enough easier that I should
>      bite the bullet and port the 1100 line grammar rather than
>      fussing with it in 2.7.7?
> 
> Thanks,
> -Chap
>