Re: What?

[Rainer Joswig <address@hidden>]

In article <003f01c246ec$6f8bbc10$be8cfd3e@wilde>,
 "Jason Trenouth" <jason.trenouth@bigfoot.com> wrote:

> BTW I noticed this embedded citation but I wasn't sure who said it
> (Andreas?):

This was my comment. ;-)

> > >>> I can understand that Dylan does no longer follow above ideas and
> > >>> tries to be compatible with the mainstream.
> 
> Anyway I just wanted to say that perhaps Dylan didn't go far enough for the
> "mainstream". From comments I've seen on code that compares Java to Dylan,
> "mainstream" folks don't like the 'flashing neon' of Dylan's use of
> punctuation symbols.

Don't forget that at Apple's time mainstream also meant
Object Pascal. Many applications on the Mac were written
in it, especially using MacApp. Then it was C++. Now
it is Objective C.

It is especially funny that Apple left mainstream by using
Objective C (and not C++ and not Java) for some of their
core applications. Now Objective C with its Smalltalkish
object system is was Dylan could have been. Sigh.

Again, Lisp adapts to this situation. Gary Byers
has written some nice integration code for OpenMCL. An
example - now you can use something like Objective C's
message passing syntax in Lisp code:

(defun create-text-attributes (&key (font (default-font))
                                    (line-break-mode :char)
                                    (color nil))
  (let* ((dict [[[(@class "NSMutableDictionary") "alloc"]
                 "initWithCapacity:" :unsigned (if color 3 2)]
                "retain"]))
    [dict "setObject:forKey:"
          :id (create-paragraph-style font line-break-mode)
          :id #@"NSParagraphStyle"]
    [dict "setObject:forKey:" :id font :id  #@"NSFont"]
    (when color
      [dict "setObject:forKey:" :id color :id #@"NSColor"])
    dict))

And create classes, of course:

(def-objc-class "lispeditordocument" "NSDocument"
  ((textview "textView") :id)
  filedata
  packagename
  echoarea)


<..>

> Of course, much of this is down to convention. If Dylan were to adopt more
> conventional conventions (!) for identifiers then the same code might look
> like this: :-)
> 
> define method sumStream ( stream :: Stream ) => ( s :: Integer )
>   let sum :: Integer = 0;
>   let n :: falseOr( String ) = #f;
>   while ( n := readLine( stream, onEndOfStream: #f ) )
>     sum := sum + stringToInteger( n );
>   end;
>   sum
> end method;

I always wondered whether it would be defineMethod instead of
define method ??

defineMethod sumStream ( stream :: Stream ) => ( s :: Integer )
  let sum :: Integer = 0;
  let n :: falseOr( String ) = #f;
  while ( n := readLine( stream, onEndOfStream: #f ) )
    sum := sum + stringToInteger( n );
  end;
  sum
endMethod;

> Still more than the Java example, but at least the punctuation is no longer
> in identifiers. (Yes I know Dylan is case-insensitive...)

It could be case-insensitive, but "case-preserving".
 
> I'm not seriously suggesting any change, but I just wanted to say that Dylan
> fell short of wrapping Lisp in a "mainstream" syntax.

I'm not really sure there is only one mainstream syntax - C is
very influential, though.