Re: Fwd: my current project: new lalr module

[Nala Ginrut <nalaginrut@gmail.com>]

On Wed, 2015-06-03 at 20:40 -0700, Matt Wette wrote:
> (Accidentally sent to guile-devel.)
> 
> Having fun with all the great code, and documentation, in guile 2. 
> 
> My current project is a lalr module. Last December I wanted to start
> working on an interpreter for modelica in guile.  Guile has a lalr
> parser generator (system base lalr).   When I started looking at it, I
> was not crazy about the input language. I looked to see if a better
> interface could be pasted on, but the code impenetrable for me.  So I
> started coding up my own LALR parser generator.  I was hoping it would
> take a month of nights/weekends but ended up taking me close to four
> months.  For now I am calling my implementation "lalr1".

cool! it'd be a great tool for our multi-lang plan. ;-)
> 
> A major difference between my lalr1 and lalr in guile is that the
> guile
> version is a translation of the C-coded bison tool and mine is a
> from-scratch implementation in Scheme based on algorithms from the
> Dragon Book.  (The bison implementation uses the DeRemer-Pennello
> algorithms, yacc does not.  I'd guess the bison-based implementation
> will generate a parser generator faster, but I think that is a minor
> advantage.)
> 
> Some features of lalr1:
> 0) It's implemented with syntax-rules and syntax-case (versus
> define-macro for lalr).
> 
oh~nice

> 1) There is no specification of terminals.  Terminals are detected in
> rules (via fenders) as quoted symbols or characters.
>     Example of terminals:  'number, #\+

hmm...and strings? Do you have plan to let users choose whether to use
terminals-inference? ;-)
> 

> 6) There is a macro to generate lalr1 input language from a
> lalr-parser specification.
> 
Yes, that would be great!

Happy hacking!

> 
> 
> 
> 
>