Re: Being constructive [Was: Nit-picking]

[Mike Mattie]

On Sat, 12 Apr 2008 11:46:11 +0000
Alan Mackenzie <address@hidden> wrote:

> Hi, Eli!
> 
> On Sat Apr 12, 2008 at 01:30:27PM +0300, Eli Zaretskii wrote:
> 
> > [Moving this to emacs-devel, where it belongs.]
> 

[snip]

> I think it would be psychologically very uplifting for each of us to
> post, in a constructive non contentious fashion, what we are working
> on, what we are trying to achieve, and so on.  This was exactly what
> my previous post was meant to be.  Eli, could you possibly respond
> again to that last post with a summary of what _you_ are working on?
> We could develop a very positive constructive thread from this.  :-)
> 

I am working on a Parser Compiler that generates pure elisp parsers from
a macro DSL.

http://www.emacswiki.org/cgi-bin/emacs-en/ParserCompiler

There are many times when a regex has turned into a ad-hoc parser. My
goal is to make those small to medium size parsers compact and declarative.

I think it is very useful considering the Emacs design of using co-processes
for everything external. That's alot of small parsing jobs to hook up
some simple and not so simple tools. 

The architecture of the design is unique: the compiler itself is
programmable reducing the need for call-outs to handle the parts
of the grammar beyond the scope of the usual meta-operators.

There is often a claim that open-source chases head-lights but
this design for better or worse is unique. It's baking on
EmacsWiki while I evaluate it's properties.

For example if PEG and CFG are called grammar "classes" in the logical
sense it will be possible to mix those classes along with a grammar
specific class (user defined) with well defined behavior. That is
unique AFAIK (please cite related works if you know where this has
been done before).

When it is completed I will likely port it to mzScheme to compare
it against the performance of standard designs, and evaluate how
the design is expressed when I don't have to kludge abstraction 
facilities such as co-routines. I will always maintain the elisp 
version though because Emacs is always with me :)

It won't hit 0.1.0 until left-recursion is solved eliminating the
possibility of user defined infinite loops.

Currently it works well with right-recursive grammars, builds canonical
AST trees, and the code emitted looks like I wrote it by hand - the back-end
folds emitted elisp to take advantage of special forms that can be shared.

When it hit's 0.1.0 I'll post a mention again and submit it to ELPA.

It's my thank-you note to the Emacs community. If some-one wants to write
a counter thank-you note Elisp TCO is at the top of my elisp wish-list.

Cheers,
Mike Mattie

signature.asc
Description: PGP signature