Re: Module name mangling

[Dirk Herrmann]

On Tue, 23 Jan 2001, Martin Grabmueller wrote:

> I have reworked the patch a bit, but I am not sure whether the current
> solution is better.  I am still doing bit-twiddling, because (as
> Marius already pointed out), the result of the conversion has to be
> two characters.
> 
> But on a second thought: maybe we only have to care about characters
> whose hex representation have two digits? ...
> 
> Oh, another problem with `string->number': the resulting hex digits
> are all lowercase, which is not common practice for URL-encoding
> (don't know about the standard, but web browsers normally use
> upper-case characters).

Although Martin's solution works perfectly doing url encoding for
filenames, I now feel that such a patch would mean overdoing things: We
only have to care about a very limited set of characters, namely those
which can not be used within filenames on some systems.  For all other
characters, we should not care whether they are typically hex-encoded in
urls or not, but rather try to keep filenames _readable_ as far as
possible.  For example, why should a '+' character be encoded if it is
allowed within filenames on all file systems known?  (Just an example, I
don't know if '+' is really allowed in filenames everywhere...)

Thus, I suggest a simple solution as the following one.

----------------------------- BEGIN ------------------------------
;;; Support for filesystem friendly encoded module names

(define (filename-encode s)

  (define char-encoding-alist
    '((#\% . "%25") (#\* . "%2A")))

  (define (char-encode c)
    (cond ((assv c char-encoding-alist) => cdr)
          (else (string c))))

  (apply string-append   
         (map char-encode
              (string->list s))))

(define (symbol->encoded-filename sym)
  (filename-encode (symbol->string sym)))
------------------------------ END -------------------------------

Instead of an url encoding, we could also choose a different encoding if
preferred:

  (define char-encoding-alist  
    '((#\% . "%percent%") (#\* . "%star%")))

This would, for example, result in "and-let%star%.scm" instead of
"and-let%2A.scm", which IMO is easier to understand.

Best regards,
Dirk Herrmann