On Mon, May 26, 2014 at 8:31 PM, Terence Parr <pa...@cs.usfca.edu> wrote: > Very interesting. Thanks for the clarifications. I can see how a misspelled > keyword could be better handled at character level, unless of course you mean > it to be an identifier. My intuition is that token level error handling would > provide more opportunities for clarity and natural recovery but I have > nothing to back this up. indicating which rules are like tokens would be a > good compromise maybe.
In Section 4 of http://arxiv.org/abs/1405.6646, on page 16, there is a discussion about this. I think it is similar to your idea. In a rule like: THEN <- 'then' !IDRest Skip When there is an error matching "theno" it seems that is better to say "Error matching 'then'" than "Error matching 'o'". Sérgio _______________________________________________ PEG mailing list PEG@lists.csail.mit.edu https://lists.csail.mit.edu/mailman/listinfo/peg
