Re: srfi-39 implementation

[Jose A. Ortega Ruiz]

Hi,

Please find attached a new version of SRFI-39.

Kevin Ryde <address@hidden> writes:

>
> Paperwork will be required if adopted.
>

Fine.

>
> Tests are vital, you'll never know if it works if you don't exercise
> all features.
>

Indeed.

>>   #:use-module (ice-9 syncase)
>
> I believe syncase is pretty big and not very fast.  define-macro or
> similar is probably a better idea.  (And with enough quasiquote might
> hide all helpers :).
>

OK. I've used define-macro this time. Nonetheless, I'm still exporting
the additional procedure with-parameters* because i've found it useful
in other contexts :).

>> (define make-parameter
>>   (case-lambda
>>     ((val) (make-parameter/helper val (lambda (x) x)))
>                                          ^^^^
> `identity' in boot-9.
>

OK. Of course.

>> (define get-fluid-tag (lambda () 'get-fluid)) ;; arbitrary unique (as per 
>> eq?) value
>
> make-symbol perhaps.
>

It seems 1.6.4 does not defines `make-symbol'. Since I haven`t
installed the cvs version yet, I've followed Paul's suggestion and
used (list 0).

>> (define (make-parameter/helper val conv)
>
> How does this stand in comparison to make-mutable-parameter in boot-9?
> Similar but not quite the same?
>

Yup. Basically the same, except that make-parameter/helper
(substituted by make-parameter in the new code) adds support for
'parameterize' and (in its current version) makes addtional error
checking.

>>          ((eq? (car new-value) get-fluid-tag) value)
>>          ((eq? (car new-value) get-conv-tag) conv)
>
> Perhaps primitive-make-property or something to associate these,
> instead of a function call.  (If I understand what it's trying to do.)
>

Although you are right in that primitive-make-property & co. can be
used, i think the case-based thing is slightly lighter and maybe even
faster, since no hashtable creation/lookup is involved. Or, at any
rate, the difference should be minimal, and the case-based version is
shorter and, IMHO, simpler (unless I'm not using the property
primitives correctly). I'm attaching a version using case
(srfi-39.scm) and a version using properties (srfi-39-alt.scm), so you
can pick the one you like best :)

>> (define (check a b a-val b-val)
>>   (if (not (eqv? (a) a-val)) (error "failure -- a" (a) a-val))
>>   (if (not (eqv? (b) b-val)) (error "failure -- b" (b) b-val)))
>
> No, you should use the test-suite/lib.scm framework.
>

OK. The attached srfi-39.test contains a test suite using this
framework.


Thanks a lot for your time and help.

Regards,
jao

-- 
"People sometimes ask me if it is a sin in the Church of Emacs to use vi.
 Using a free version of vi is not a sin; it's a penance". - Richard Stallman

srfi-39.scm
Description: srfi-39

srfi-39.test
Description: test suite

srfi-39-alt.scm
Description: Alternative impl using properties