[antlr-interest] Huge amounts of follow sets

[Andreas Meyer]

Dear ANTLR developers!

I have a huge grammar with lots of ambiguities (its a language with 600 
keywords, that can also be used as identifiers ...). It worked fine with 
ANTLR 2.7, but after upgrading it to 3.1, I finally get an error when 
compiling the generated parser: method size exceeds 64k. This is due to 
the number of follow-sets being generated and initialized ( > 6000). One 
workaround would be: move assignments from class-level to methods (up to 
some number of assignments per method), and then call these functions 
from one "master" function. Are there any other alternatives, other than 
(probably) making the grammar less ambiguous? Thanks for any hints!

Best Regards,
Andreas Meyer
(smartshift.de)