[antlr-interest] ANTLR v3: enum as identifier and as a keyword

[Terence Parr]

On Jun 30, 2005, at 10:15 AM, David Ewing wrote:
> On Jun 29, 2005, at 1:10 PM, Jörg Rech wrote:
>> Hi,
>>  maybe I got something wrong here but isn't the lexer called  
>> before the recognizer for the methods? So if I enter a method  
>> changes to enableEnum have no effect on the lexer?
>>
>> Or is the lexer just registered and called by the recognizer/ 
>> parser on demand?
>>
>
> It's called on demand. Another possibility would be to change  
> 'IDENT' to '(IDENT | "enum")' in a few key places elsewhere in the  
> grammar (e.g. variable declarators, expressions). That might be a  
> tad more compatible, but it also might add ambiguities to the  
> grammar. In any case, I haven't tried either, but I'd be interested  
> to hear if they work.

Now that hoisted semantic predicates are back in ANTLR, you can do  
this very easily.  I just built a new example (and pushed it to the  
site--examples-v3.tar.gz has been updated).

/** Demonstrates how semantic predicates get hoisted out of the rule in
*  which they are found and used in other decisions.  This grammar  
illustrates
*  how predicates can be used to distinguish between enum as a  
keyword and
*  an ID *dynamically*. :)
* Run "java org.antlr.Tool -dfa t.g" to generate DOT (graphviz)  
files.  See
* the T_dec-1.dot file to see the predicates in action.
*/
grammar T;

{
/** With this true, enum is seen as a keyword.  False, it's an  
identifier */
boolean enableEnum = false;
}

stat: identifier    {System.out.println("enum is an ID");}
     | enumAsKeyword {System.out.println("enum is a keyword");}
     ;

identifier
     : ID
     | enumAsID
     ;

enumAsKeyword : {enableEnum}? "enum" ;

enumAsID : {!enableEnum}? "enum" ;
...

The rule stat decision sees ID or ID syntactically, but with  
predicates, the decision looks like this:

-------------- next part --------------
A non-text attachment was scrubbed...
Name: pastedGraphic.pdf
Type: application/pdf
Size: 14150 bytes
Desc: not available
Url : http://www.antlr.org/pipermail/antlr-interest/attachments/20050630/c86da645/pastedGraphic-0001.pdf
-------------- next part --------------

The =>alt notation indicates which alt will be predicted.  The  
generated decision looks like:

             int LA1_0 = input.LA(1);
             if ( LA1_0==ID ) {
                 alt1=1;
             }
             else if ( LA1_0==7 ) {
                 int LA1_2 = input.LA(2);
                 if ( evalPredicate(enableEnum,"enableEnum") ) {
                     alt1=2;
                 }
                 else if ( evalPredicate(!enableEnum,"!enableEnum") ) {
                     alt1=1;
                 }
                 else {
                     ...error...
                 }
             }
             else
                    ...error...

Pretty slick, eh?  Hooray!

Ter
--
CS Professor & Grad Director, University of San Francisco
Creator, ANTLR Parser Generator, http://www.antlr.org
Cofounder, http://www.jguru.com