[antlr-interest] sematic analysis ?

[Kaleb Pederson]

Venkat,

One way to do what you've mentioned is to create a tree walker which
walks the AST.  Then, within the AST, you'll need to do a couple of
things.

First, you need to populate a symbol table.  That symbol table will
store the different objects/variables that are declared.  In your
example, it would store a and b, their values, if known, and their
types.

Second, you need to do an analysis of the structure while using the
symbol table.  Here's a fragment from one of my tree walking grammars:

plusMinusExpression returns [Type type]
	:	^(PLUS lhs=baseExpression rhs=baseExpression)
		{
			typeChecker.assertIsNumericType($lhs.type);
			typeChecker.assertIsNumericType($rhs.type);
			typeChecker.assertEqualTypes($lhs.type, $rhs.type);
			$type = $lhs.type;
		}
		/* ... */
		;

The above calls on my typeChecker to verify that both types are
numeric (I don't use + for concatenation) and I assert that the types
are equal (since I don't have any automatic casting from int, to
double, etc.). All of my type checking happens in a single pass of a
tree walker, so it's nicely componentized.

I hope that helps.

--
Kaleb Pederson

http://kalebpederson.com
http://twitter.com/kalebpederson


On Fri, Mar 19, 2010 at 9:46 AM, venkat medhaj <hvmedhaj at gmail.com> wrote:
> Hi,
>
> I am kinda new to antlr and I wonder how to perform symantic analysis using
> the tool. I was able to generate the AST output for a small grammar and now
> I want to check for the type check errors. For ex: if I have a statement
> that looks something like
>
> int a + b , I just have to make sure that both are of type int.
>
> or if I define a twice in my input, then thats a redefine error. Can some
> one give me some insight into this ?
>
> Thanks,
> -Venkat.
>
> List: http://www.antlr.org/mailman/listinfo/antlr-interest
> Unsubscribe: http://www.antlr.org/mailman/options/antlr-interest/your-email-address
>