[antlr-interest] AST construction: partial null trees

[Oliver Zeigermann]

I have not investigated why, but including partial null trees often
makes the whole generated ASTs null which makes debugging really hard.

Am I doing anything wrong, or could that indeed be handled more
gracefully by ANTLR?

As an example have a look at such a parser grammar rule creating an AST

quantorCall
	: quantor parameters -> ^( QUANTORCALL quantor parameters)
	;

which indeed would mess up my complete parsetree for reasons that are
not obvious at first glance.

However, even though parameters are not syntactically optional, they
can still return a null tree when the list of parameters is empty. If
so, the whole tree is messed up.

Of course, adding a '?' to the tree generation part of parameters fixes this:

quantorCall
	: quantor parameters -> ^( QUANTORCALL quantor parameters?)
	;

But should that really be necessary?

Or is it even considered bad style to return a null tree from a subrule?

Any input welcome!

Cheers

Oliver