[antlr-interest] ast validation question

[Sohail Somani]

On Wed, 2006-02-22 at 22:11 -0500, Scott Amort wrote:
> Hello All,
> 
> I am writing a parser library for a relatively simple textual
> description language, and have progressed fairly well given my relative
> lack of familiarity with compiler tools like ANTLR.  This list and the
> sample code and grammars provided have been very helpful.
> 
> At this point, I have a functioning lexer and parser, and am now working
> on a tree walker for validation purposes.  I am hoping to get some
> opinions on how best to approach this - currently, I envision using the
> tree walking phase to first validate type, then data.  The language
> allows tags ( i.e. a '\' followed by text, e.g. \start or \stop).
> Currently the lexer identifies the alphanumeric portion after the
> backlash as a token and sends it on to the parser, which makes sure it
> is found in the correct spots.  Now, in the tree walker, I would like to
> make sure that this tag is valid.  Is there an ANTLR-specific way to
> accomplish this?  Or perhaps I should develop my own code to deal with
> this?  If so, what data structure do you think is best to store a fixed
> number of strings, and allow efficient matching or finding?

Couldn't you make these keywords and have them matched as in the parser?

Something like (probably doesnt compile):

class L extends  Lexer;
KEYWORDA:"FOO";
KEYWORDB:"BAR";
KEYWORDC:"NO";
SLASH:'\\';

class P extends Parser;
options
{
	importVocab=L;
}
{
	public static void main(String args[])
	{
		L l = new L(System.in);
		P p = new P(l);
		p.tagRule();
	}
}
tagRule:SLASH (KEYWORDA | KEYWORDB | KEYWORDC);

The key of course is fixed number of strings.