[antlr-interest] new DFA implementation for generated v3 recognizers

[Jeff Barnes]

--- Terence Parr <parrt at cs.usfca.edu> wrote:

> 
> The solution I have is optimal in time and very good
> in space for all  
> states with less than some threshold number of
> edges.  

<snip>

> I'd waste all the space in between.  This
> will waste 64k ints  
> for any state that does that, but naturally, I'll
> push that to a  
> "special" state.
> 

So you'll have varying degrees of optimity in the time
domain, with no optimity in the space domain...

How are you constructing the "special" states?

Regards,
Jeff

=========
Jeff Barnes
(206)245-6100


Few things are impossible to diligence and skill.
--- Samuel Johnson (Rasselas Chap. xii.)