Re: [MIT-Scheme-devel] Patches for current Swank

[Taylor R Campbell]

Thanks for the patches -- I finally merged them today.  There are a
few little issues which it would be nice to fix.  First, can you try
to keep the code under 80 columns wide?

   From: Peter Feigl <address@hidden>
   Date: Thu, 3 May 2012 09:31:24 +0200
   Subject: [PATCH 2/6] Add autodoc for procedures

   ...

   +                     (cons symbol (read-from-string (string-trim 
(with-output-to-string
   +                                                                   (lambda 
() (pa binding))))))

Instead of using PA to get a string and then reading it back in, it
would be better to adapt the implementation of PA to get a list.
Maybe we should have

(define (procedure-parameter-description procedure)
  ...)

in global.scm and write PA in terms of that.

   +        (let ((answer (if ans (with-output-to-string (lambda () (write 
ans))) ':not-available)))

You can just use WRITE-TO-STRING here.

   From: Peter Feigl <address@hidden>
   Date: Thu, 3 May 2012 09:45:45 +0200
   Subject: [PATCH 5/6] Fix basic debugging problems.

   +         (dynamic-wind
   +             (lambda ()

I don't know who was responsible for choosing this horrid indentation
for DYNAMIC-WIND in GNU Emacs, but you should put (put 'dynamic-wind
'scheme-indent-function nil) in your .emacs file to fix it...

   +(define *index*)

   -(define (emacs-rex socket sexp pstring)
   -  (fluid-let ((*buffer-pstring* pstring))
   +(define (emacs-rex socket sexp pstring id)
   +  (fluid-let ((*buffer-pstring* pstring)
   +          (*index* id))
        (eval (cons* (car sexp) socket (map quote-special (cdr sexp)))
             swank-env)))

Can you add some comments to explain what this index is, and perhaps
give it a more descriptive name?  It's not clear that EMACS-REX is the
right place to put it, either -- it looks to me as though it is an
SLDB-specific thing and as such should be handled only by SLDB code,
not by the generic Swank communication.

   +     (format #f "~a in package ~a~a of type ~a.~%~%" (string-upcase 
(symbol->string symbol))

   +(define (get-object-type-name obj)
   +  (cond ((boolean? obj) "boolean")
   +    ((string? obj) "string")
   +    ((char? obj) "char")
   +    ...

Is there a reason you define this instead of just using
USER-OBJECT-TYPE for the whole thing?