Re: LALR parser question

[Shark8 <onewingedshark@gmail.com>]

On Tuesday, April 30, 2013 11:15:00 AM UTC-6, Hibou57 (Yannick Duchêne) wrote:
> Le Tue, 30 Apr 2013 18:06:53 +0200, Shark8 a  
> 
> écrit:
> > hybrid  ascent-descent
> 
> That's a part of what I have in mind too :p But won't read the paper you  
> mention, because I don't want to be influenced by others ideas on that  
> topic, to avoid to unconsciously repeat what others already did.

:)
Yep, I haven't read it either but it looks quite interesting. And I can totally respect not wanting to copy others -- that was a main peeve of mine when in college, mentioning I was working on an OS and then being told "just get the Linux source."

> Since some time I'm thinking again I should try to do one applying the  
> idea I have in mind. The hybrid ascent-descent is indeed an option, while  
> I believe the ascent is the most natural.

I'm not sure about that; recursive decent is really natural for anyone who's done recursion -- See: http://compilers.iecc.com/crenshaw/

> I think the parser should read  
> as the humans read, and humans, the most of the time, start looking at the  
> tiny parts (the bottom) and goes bottom-up, and just occasionally go  
> top-down, and so just when there is a clear marker for this, like a  
> heading title in a text as an example and any other marker which only  
> appears at the top of a construct.

Really, I thought they read in a more top-down manner, with perhaps things front-loaded [i.e. forward-declared] as in sections and a TOC.

But that's why it's such an interesting field: there's a lot of ways to do it and a lot of them have different strengths and weaknesses.