[antlr-interest] automatic tree grammar generation

[Loring Craymer]

Ter--

The basic methodology is "refactor until the transformation is obvious".  For this one, the sequence is:

a :  b  ( '+' x = b  -> ^('+' $a $x) )* ;

goes to

a :   b  ( '+' x = b  -> ^('+' $a $x) )+ 
        |  b
    ;

after which you factor out

c : b  ( '+' x = b  -> ^('+' $a $x) )+ ;

which becomes

c :  c1 = a  '+' x = b -> ^('+' $c1 $x) ;

and substituting back into a gives

a :  ^( '+' a b )
        | b
    ;

Cheers!

--Loring



----- Original Message ----
> From: Terence Parr <parrt at cs.usfca.edu>
> To: Kamal Verma <kverma at yahoo.com>
> Cc: Loring Craymer <lgcraymer at yahoo.com>; Bernardo Elayda <belayda at gmail.com>; antlr-interest at antlr.org
> Sent: Monday, May 19, 2008 5:30:21 PM
> Subject: Re: [antlr-interest] automatic tree grammar generation
> 
> 
> On May 19, 2008, at 4:49 PM, Kamal Verma wrote:
> 
> > Wonderful, when can we have it.
> >
> > Few months ago I had similar need and I almost thought of same  
> > thing. It will be wonderful if have it as a command line option on  
> > antlr. Like "-treeg ". How about string templating the  
> > 'AST' of 'antlr' grammar. ( inside the tool )
> >
> > It will also help with understanding of generated AST and debug and  
> > develop the grammar.
> >
> > I will love to hear Terence thoughts on this.
> 
> Not always easy for -> rewrites even when they use $a vars.
> 
> a : (b->b) ('+' x=b -> ^('+' $a $x) )* ;
> 
> not sure how i'd figure that out off top of my head.
> 
> A basic "get me started" kind of thing would be pretty easy I guess. :)
> 
> Ter