On Thu, Nov 04, 2010 at 11:16:47AM -0700, John Clements wrote: > For now, this goes in the FYI category: I have a student that's implemented a > simple error-recording database. We've now installed it on our lab machines, > and you can see some early results here: > > http://li21-127.members.linode.com:8020/errrecorder > > The goal of the project is to allow students & others to provide helpful > hints that will guide people toward solutions to their problems. As you can > see, this early version has way-too-coarse categories: we're looking into > using measures of string similarity to group similar-looking error messages. > > John > Also FYI: The writers of the CDC Algol 68 compiler used an LR parser. When the time came to do parsing error recovery, they initially used some heuristics. However, every time a user complained about a bad error recovery, they figured out what error recovery he wanted, found the spot in the FSM where it had been done badly, and inserted an hand-picked error recovery.
It ended up with very good error recovery. -- hendrik _________________________________________________ For list-related administrative tasks: http://lists.racket-lang.org/listinfo/users
