[antlr-interest] Tree building limitation in tree grammars?

[John B. Brodie]

Greetings!

On 11/28/2011 08:08 AM, franck102 wrote:
> My parser grammar can generate a number of subtrees with similar structure,
> and just a different root type. In the tree grammar I am trying to just pass
> such a subtree along without any changes, but that is turning out to be much
> trickier than I expected, am I missing something?
> 
> Given this AST: ^( AND true false ) I want the tree grammar to generate the
> same AST subtree.
> 
> The following attempts are not working:
> 
> ^( ( AND | OR ) booleanExpression booleanExpression )  // no rewrite =>
> flattens the tree, 

i am surprised. flattening has not been my experience, altho i have not
used an alternative for the root node....

>                                                                            
> // I get a nil root and 3 children
> 
> ^( ( root=AND | root=OR ) booleanExpression booleanExpression )
> -> ( ( AND | OR ) booleanExpression booleanExpression )               //
> syntax error
> 
> ^( ( root=AND | root=OR ) booleanExpression booleanExpression )
> -> ( AND? OR? booleanExpression booleanExpression )                   //
> syntax error
> 
> ^( root=(AND | OR ) booleanExpression booleanExpression )
> -> ( $root booleanExpression booleanExpression )                 // runtime
> error, $root is empty
> 
> Am I going to have to split those rules??
> 
> Any suggestion appreciated!
> 
> 

3 untested suggestions (i happen to prefer the third, but i am weird;-):

^( (root=AND|root=OR) e1=booleanExpression e2=booleanExpression )
   -> ^($root $e1 $e2)

or

^( binaryOperator booleanExpression booleanExpression ) // no rewrite

or

^( op=binaryOperator e1=booleanExpression e2=booleanExpression )
   -> ^($op $e1 $e2)

where

binaryOperator = AND | OR /* | others go here... */ ;