[antlr-interest] Parsing with inverse matches

[Gavin Lambert]

At 11:20 23/11/2009, Vipul Delwadia wrote:
 >x	:	STRING+;
 >
 >fragment BACKSLASH
 >	:	'\\';
 >
 >NOTA:	BACKSLASH A;
 >
 >A	:	'a';
 >
 >STRING
 >	:	(~(A)|NOTA)+;
[...]
 > This works for the most part except when I try and match
 > just "\a", at which point I get a MismatchTokenException
 > (or sometimes a NoViableAltException).

The input "\a" by itself will produce a NOTA token, since that's 
the best-fit non-fragment rule.

Your parser is however not expecting that.  Almost certainly, both 
A and NOTA should be fragment rules.

(In general, you be suspicious any time a non-fragment lexer rule 
refers to any other non-fragment lexer rule.  It's not illegal, 
and sometimes it's essential, but unless you're really careful 
it's easy to break things by doing so.)


Also, as Kirby said, inside the STRING rule itself you have one 
alt which is a subset of the other.  Whenever you do this you 
should normally list the most general alt last, so NOTA should 
come before ~A.