Re: Determining LR(k)-ness

[Philip Aker]

On 2009-11-18, at 15:07:01, Hans Aberg wrote:

> On 18 Nov 2009, at 23:46, Philip Aker wrote:
>
> > I just used LR kind of loosely. So to be technically correct, I  
> > guess I should have said LALR(k)-ness.
>
> That does not help - it is just LALR(1).

Ok, I just plain have to admit a badly named question.

> > What I want to know is when the parser splits, even if it splits  
> > multiple times, is there any way that bison can report the maximum  
> > 'k' lookahead number reached by all of these split parsers.
> >
> > Like if there was a grammar where the parser was split once and was  
> > able to resolve the problem at that point, it would report  
> > "2" (because LALR(1) already has 1 token of lookahead).

> That is another property - the number of past branchings. Perhaps  
> you can write some actions computing it. But even in a non-GLR  
> setting, there is no implemented feature telling whether the  
> algorithm requires 0 or 1 lookahead tokens. This must be taken into  
> account when making mid-rule actions, for example.

IOW, there is no glr reporting available.

Thank anyways, I do appreciate the input one way or another,

Philip Aker
echo address@hidden@nl | tr address@hidden p-za-o.@

Democracy: Two wolves and a sheep voting on lunch.