My application is a state machine with some local variables in each node, and time-outs for arcs as well, Having really enjoyed Shriram Krishnamurthi’s automaton compile-to-closures,

http://cs.brown.edu/~sk/Publications/Papers/Published/sk-automata-macros/

As I’ve implemented the additional stuff (local vars, timeout etc), I’ve started to get pretty complicated macros. I wonder if I would be better off compiling the automaton definition

(fsm init
    [init : (c -> more)]
    [more : (a -> more)
          (d -> more)
          (r -> end)]
    [end : ])

for example, to a class - which I could then add methods to so I could query the variables in a node, check time-remaining-in-node etc… I know all of this can be done with closures (and a dispatch method), but it seems a bit more straightforward to do it with explicit objects/methods.

Second questions - given that (above) type of definition, are macros emitting class defns the way to go, or parser-tools (lex/yacc) a better bet?

Again, thanks for the help…

If you want to change to using classes with basically the same syntax, then using macros is still the way to go

Only use parsing and lexing if you need to handle syntax that doesn’t fit with s-expressions

Thanks @samth. The description of the Automata is s-expressions, so macros it is. Thanks again