> | My goal is to completely rewrite the address parser in lex/yacc; > >Is there a specific reason for that? yacc/lex parsers tend to be >bigger, slower, less tolerant, and pathetic in handling errors >(unless you do a lot of work.)
Well, note that it hasn't happened YET. Call it "an experiment"; I'd like to answer the question: is it even feasible? First, take a gander at sbr/mf.c, our current address parser. Ugh. Our address parser has been causing us a number of problems (exhibit A: anyone who puts a period in their name and doesn't quote it, which is technically permitted by the current grammar). So, it "works", sort of, but it needs some updates for updates that have occured along the way. But fixing the problems with it is hard, and the API is weird. If someone else wants to rewrite it or use something else to generate an address parser, then please feel free! I'm really concerned about long-term maintenance issues; it feels like it's m_getfld() all over again. As for speed .... well, that MIGHT have been a concern 30 years ago. Nowadays, I don't think it matters. --Ken _______________________________________________ Nmh-workers mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/nmh-workers
