[antlr-interest] How to best resolve ambiguity?

[Terence Parr]

On Dec 30, 2005, at 6:40 PM, Martin Traverso wrote:

> Ok, I did some testing, and antlr seems to be doing the natural  
> thing even without explicitly setting k=1. What threw me off was  
> that when I compiled the grammar I got the following errors:
>
> test.g:6:7: Decision can match input such as "'+'" using multiple  
> alternatives: 1, 2
> As a result, alternative(s) 2 were disabled for that input
> test.g:7:7: Decision can match input such as "'*'" using multiple  
> alternatives: 1, 2
> As a result, alternative(s) 2 were disabled for that input
>
> But looking at the generated code it doesn't seem like any alt was  
> disabled.

Actually, alt 2 is the exit branch for a single alt (...)* loop so  
it's essentially saying it won't ever exit upon '+'. :)

> On a related note, any idea why antlr complains with "2:16: The  
> following alternatives are unreachable: 2" for the following  
> grammar? A bug, maybe?
>
> grammar T;
> a    :  'x' ('+' a)? ;

For one, ANTLR must complain that there is no rule obviously followed  
by EOF.  In this case, though, I think it's saying that there is no  
input that will predict bypassing the (...)?, which of course there  
isn't due to lack of start rule.  It should become ('+' a) in the  
output and probably never terminate. :)

hooray for ANTLR v3's analysis engine. :) ;)

Ter