Re: [boost] [regex] any interest in finite automata-based regexengine?

From: John Maddock (john_at_[hidden])
Date: 2004-11-08 07:42:53

> For what it's worth, I think that's a really promising idea.  As is, the
> library is a black box.  Nobody can introduce a new syntax parser
> without access to the library internals; this is directly contrary to
> the way that the STL separates containers and algorithms.

If you want to be able to access the state machine internals for a 
particular implementation that's probably not too hard: in fact it's on my 
Todo list.  The hard part is defining an abstract state machine interace 
that can be used by any state machine implementation rather than just one 
specific one.

In fact I once tried to do this and gave up after about 6 months (wasn't 
really geting anywhere), so I guess you could say I have history which 
colors my perspective here :-/

I am in violent agreement that this is a good idea though, it's just getting 
it done that's hard,

John.

> The question is whether it's possible to define an interface, or a
> family of interface, for creating and executing NDFAs.  (Or maybe
> even DFAs in some simple cases?)  Certainly the fact that better
> algorithms are possible for regexes that don't have backreferences
> than for regexes that do is interesting; to me, at least, this suggests
> that we might want to provide both algorithms and let users choose
> whichever one is appropriate.
>
>      --Matt
> _______________________________________________
> Unsubscribe & other changes: 
> http://lists.boost.org/mailman/listinfo.cgi/boost
>