Comments on: Parsing framework status http://blog.reverberate.org/2007/06/04/16/ parsing, performance, minimalism with C99 Thu, 26 Jan 2012 04:15:43 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: josh http://blog.reverberate.org/2007/06/04/16/comment-page-1/#comment-677 josh Sat, 09 Jun 2007 17:19:24 +0000 http://blog.reverberate.org/2007/06/04/16/#comment-677 Hey Kaushik, thanks for writing. While it's true that an arbitrary NFA could have epsilon transitions out of the final state, it is an invariant of the NFA construction I used that no final state has <i>any</i> transitions out of it. So the algorithms in nfa_construct.lua will work for any NFA that was generated using that construction. Perhaps I should document this assumption. Hey Kaushik, thanks for writing. While it’s true that an arbitrary NFA could have epsilon transitions out of the final state, it is an invariant of the NFA construction I used that no final state has any transitions out of it. So the algorithms in nfa_construct.lua will work for any NFA that was generated using that construction. Perhaps I should document this assumption.

]]> By: Kaushik Sridharan http://blog.reverberate.org/2007/06/04/16/comment-page-1/#comment-676 Kaushik Sridharan Sat, 09 Jun 2007 09:29:41 +0000 http://blog.reverberate.org/2007/06/04/16/#comment-676 I don't really know Lua so I apologize if I am reading this completely wrongly, but in the concat function in nfa_construct.lua it looks like any existing epsilon transitions from the final state of nfa1 are overwritten when you append nfa2 to it. I guess you are assuming that nfa1 has no e-transitions from the final state, but I am not sure if that is a valid assumption to make. You could have reached the final state but not yet have consumed all the input; an e-transition back to an earlier state might still get you to the final state at the end of the input. Likewise for the alt and rep functions, although you do a table.insert in alt for new_nfa because you know it has multiple e-transitions from the start state. Again, I may be completely off track here; it has been a while since I looked at NFAs and I don't know anything about your project. Cheers! -K I don’t really know Lua so I apologize if I am reading this completely wrongly, but in the concat function in nfa_construct.lua it looks like any existing epsilon transitions from the final state of nfa1 are overwritten when you append nfa2 to it.

I guess you are assuming that nfa1 has no e-transitions from the final state, but I am not sure if that is a valid assumption to make. You could have reached the final state but not yet have consumed all the input; an e-transition back to an earlier state might still get you to the final state at the end of the input.

Likewise for the alt and rep functions, although you do a table.insert in alt for new_nfa because you know it has multiple e-transitions from the start state.

Again, I may be completely off track here; it has been a while since I looked at NFAs and I don’t know anything about your project.

Cheers!

-K

]]> By: josh http://blog.reverberate.org/2007/06/04/16/comment-page-1/#comment-672 josh Wed, 06 Jun 2007 06:14:30 +0000 http://blog.reverberate.org/2007/06/04/16/#comment-672 That's quite awesome. Thanks for Graphviz! It's been really helpful. That’s quite awesome. Thanks for Graphviz! It’s been really helpful.

]]> By: Stephen North http://blog.reverberate.org/2007/06/04/16/comment-page-1/#comment-671 Stephen North Tue, 05 Jun 2007 21:46:48 +0000 http://blog.reverberate.org/2007/06/04/16/#comment-671 Ever see this? It's cool. http://osteele.com/tools/reanimator/ Ever see this? It’s cool.
http://osteele.com/tools/reanimator/

]]>