Tim Peters wrote: > [Bryan Olson] >> Unfortunately, the stuff about NFA's is wrong. Friedl's awful >> book > > Strongly disagree: [...]
I know I'm disagreeing with a lot of smart people in panning the book. >> What Python uses is search-and-backtrack. Unfortunately such >> engines don't have much theory behind them, and it's hard to >> reason generally about what they do. > > Yup X 3, and the last is precisely why Friedl's book is valuable for > people using "NFA" implementations: Friedl does a good job of > explaining when and why you're likely to trigger atrocious runtime > performance, and gives practical general techniques for avoiding those > problems. None of that has anything to do with CompSci regexps > either, but is vital knowledge for people hoping to make happy > non-trivial use of Python/Perl/etc regexps. I read the first edition, minus some engine-specific stuff. General techniques are what I want, and I didn't see them in the book. He had things to do and not do, but it didn't add up to building (ir-)regular expressions with reliably tractable behavior. -- --Bryan -- http://mail.python.org/mailman/listinfo/python-list