Re: [MIT-Scheme-devel] fasl hook or something

[Joe Marshall]

The reason I want to intern is to have EQ-ness.  The wrapper trick
you suggest won't quite do it.

On Mon, May 10, 2010 at 2:11 PM, Matt Birkholz
<address@hidden> wrote:
>> From: Joe Marshall <address@hidden>
>> Date: Fri, 7 May 2010 11:43:00 -0700
>>
>> I want to intern some objects, but I want to be able to fasdump and
>> fasload them, too.  Is there any reasonable way (or unreasonable,
>> for that matter), to hook into the fasloader to intern things when
>> they are restored?
>
> A simple test can determine if your object was recently fasloaded, at
> which point you can re-intern (on-demand).  E.g.
>
>    (define-structure thing tag unique)
>
>    (define (thing.unique t)
>      (if (eq? unique-tag (thing-tag t))
>          (thing-unique t)
>          (let ((new (re-intern-unique (thing-unique t))))
>            (set-thing-unique! t new)
>            (set-thing-tag! t unique-tag)
>            new)))
>
>    ;; This cannot match any fasloaded string, until some smart-*ss
>    ;; adds fancy constant string sharing yabber to my KISS runtime... :-}
>    (define unique-tag "constant string, shared or not")
>
> The real trick is cutting your objects off from their methods.  These
> insidious devils are (one?) reason why this happens:
>
>    1 ]=> (fasdump (make-thing unique-tag 'stand-in) "thing.bin")
>
>    ;Dumping "thing.bin"...
>    ;Object cannot be dumped because it contains an environment: #[thing 19]
>    ;To continue, call RESTART with an option number:
>
> and why I have been heard to incant:
>
> (define-structure (alien-function
>                   ;; To be fasdump/loadable.
>                   (type vector) (named 'alien-function)
>                   ...)
>  ...)
>



-- 
~jrm