[antlr-interest] Re: Handling identifiers

[mafw14]

--- In antlr-interest at yahoogroups.com, mzukowski at y... wrote:
> Do identifiers get declared?  
> 
> You may need to handle this in the parser, the same way I handle 
typedefs in
> the gcc grammar http://www.codetransform.com/gcc.html with a 
semantic
> predicate.
> 
> Could you give some sample text?
> How do you know when to put an identifier into the literals hash 
table?

Thanks for your reply.  The identifiers are not declared within
the language but are predefined.  I do not want to include them
as keywords in the language but have a class that defines them
so the identifiers can be changed and that each identifier can
have attributes (such as data type).  The language is stronly
typed so I want to prevent an integer being compared with a
float etc.

I think I have partly figured it out now - I have modified
the rule for an identifier as follows:

IDENTIFIER
	options {testLiterals=true; }
	: t:LETTERS
	{
		if (Identifier.isNotIdentifier(t.getText()))
		{
			$setType(STRING_T);
		}
	}
	;

I wish to save the data type of each literal or identifier on
the AST.  Is there an easy way to extend the AST to store
additional attibutes?  BTW: This parser is only used as a
checker and does not generate any code.

A sample of the text of the language is:

(MSG_TYPE = XXX) AND (VALUE > 0.001) OR NOT (SOURCE IN ["S1", "S22"])

where MSG_TYPE and SOURCE are identifiers of type STRING_T and
VALUE is of type REAL_T.  XXX is a literal string (does not have to
be enclosed in quotes).  The square brackets enclose a set of
strings.

Thanks, Mark.



 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/