That looks really cool, awesome to be able to generate static FSM code. I use my FSM (and started to use Jake Gordon's) when I need a lightweight state variable, a step up from a collection of bools when it starts to get messy with lots of ands and ors.
I mostly like this. The only gripe I have with it is the stringly-typed way of triggering events. It seems like you might be able to better leverage the type system to ensure that spelling mistakes or fat-fingerings don't cause a runtime error sometime in the future. If the .Event() method took an object that satisfied the "Trigger" (or something) interface, instead of just a string, you might be able to catch those kinds of errors during compile-time instead.
Thanks for the feedback! It's a good suggestion, and having something else than strings as events is something that I have analyzed. Currently different handlers depends on string operations as selectors, for example "before_<event>". Any ideas on how to move that to the static type world (preferably without using the reflect package)?
Zed Shaw wrote a great essay about state charts in Ragel:
http://www.zedshaw.com/essays/ragel_state_charts.html
The official docs:
http://www.complang.org/ragel/ragel-guide-6.8.pdf