On Apr 11, 2009, at 12:52 PM, Bob Foster wrote: > Thanks. I didn't want to wander into a discussion of error recovery in > general - I'm not writing a paper, just wanted something practical to > use or implement - but ANTLR is probably the exception. > > I think I got the gist of it from the first page of the google search > 'error recovery ANTLR'. It does look decent, though I'd want to try > its > automatic resynching with a real grammar and test cases before making > any judgements.
ANTLR has an advantage over pure pegs in this case because ANTLR's decisions can commit before reading entire input. I.e., I use pure LL(k) unless k isn't sufficient. Then I use a cyclic DFA, LL(*), then I will backtrack. Even within a single decision, the predictor will backtrack only if necessary. Ter _______________________________________________ PEG mailing list PEG@lists.csail.mit.edu https://lists.csail.mit.edu/mailman/listinfo/peg
