[antlr-interest] question on greedy matching

[Andrei Alexandrescu]

Hello,


I have a question on how to solve a particular greedy/non-greedy 
matching with ANTLR.

Consider the following language design for a sort of a try statement:

try statement ( catch (ID ID) '{' statement '}' )+

Clearly not an award-winning design, because it has the "dangling catch" 
ambiguity:

try
     try
         foo();
     catch (E1 e) { bar(); }
catch (E2 e) { baz(); }

There's a need for a meta-rule to decide where the last catch comes. So 
here it is:

"One catch always binds to the closest try statement. The others always 
bind to the outermost try statement possible."

I'm not sure I was clear enough, but the intent of the rule above binds 
the first catch in the example to the inner try, and the second catch to 
the outer try, as the indentation suggests.

Moreover, in the example:

try
     try
         foo();
     catch (E1 e) { bar(); }
catch (E2 e) { baz(); }
catch (E3 e) { baa(); }

again the rule makes the statements bind as the indentation suggests.

So, the question is, what would be the cleanest way to express that in 
ANTLR (preferrably without semantic predicates)? If I understand 
correctly, it's not an issue of setting the greedy option, because the 
same try construct must be parsed greedily or not, depending on whether 
the parent is a try statement as well.


Thanks,

Andrei