Re: Portin SSAX/SXML

Hey Oleg,

Thanks for your reply. I finally got around to properly packaging SSAX,
and wanted to share some notes.

On Sat, 2004-07-10 at 00:32 -0700, [email protected] wrote:
> > Guile's syntax-rules support sucks, though -- the module takes about a
> > second to load up.
> 
> Perhaps you would like to use SSAX/lib/SSAX-expanded.scm then
> (especially for simple.scm, which you enclosed).

I ended up doing this. Of course, that way you don't get
sxml:make-parser and related macros, which is a bit wrong. I'm not sure
what I'll do about this. It could be, as you say, that I should only
avoid r5rs macros in the simple wrapper, and load up the full macros for
(sxml ssax). Dunno.

Also as you suggested, the -expanded source file is filtered as it is
being opened. For the production code, it takes only the forms headed by
`define', and for the tests, it imports the (sxml ssax) module, then
loads up and evaluates the forms not headed by `define'. It is worth
noting that the tests can't run with just the exports given in Michael's
packages.scm; I had to poke into the internal stucture of the (sxml
ssax) module to get at the following symbols:

        ssax:read-NCName ssax:read-QName ssax:largest-unres-name
        ssax:Prefix-XML ssax:resolve-name ssax:scan-Misc
        ssax:assert-token ssax:handle-parsed-entity ssax:warn
        ssax:skip-pi ssax:S-chars ssax:skip-S ssax:ncname-starting-char?
        make-xml-token unesc-string name-compare

Michael, are those intentionally omitted from your interface?

> > The other guile-related bit that matters is input-parse.scm; next-token
> > is MUCH MUCH MUCH faster with `read-delimited' from guile's
> > (ice-9 rdelim).
> 
> I don't know about Guile's module system; some systems support
> selected importing of code

Odd. You can select certain things to import, you can rename symbols
programmatically, but you can't select certain things to _not_ import.
Well, since guile doesn't compile, and has first-class modules, you can
hack around it in other ways. I fixed it.

> > (define* (xml->sxml #:optional (port (current-input-port)))
> It should be possible then to define 'define-opt' to map to native
> Guile's form define*.

Ah yes, I did that with input-parse.scm. Here it is, for the record.

;; rewrite oleg's define-opt into define* style
(define-macro (define-opt bindings body . body-rest)
  (let* ((rev-bindings (reverse bindings))
         (opt-bindings
          (and (pair? rev-bindings) (pair? (car rev-bindings))
               (eq? 'optional (caar rev-bindings))
               (cdar rev-bindings))))
    (if opt-bindings
	`(define* ,(append (reverse (cons #:optional (cdr rev-bindings)))
			  opt-bindings)
	   ,body ,@body-rest)
	`(define* ,bindings ,body ,@body-rest))))

So, ssax 5.1 is in guile-lib now (or will be headed there soon). Thanks
for the comments, they helped a lot. And thanks for the code as well :)

Cheers,
-- 
Andy Wingo <[email protected]>
http://ambient.2y.net/wingo/


-------------------------------------------------------
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_id=4721&alloc_id=10040&op=click