Small enhancement to options interface

Small enhancement to options interface

[Neil Jerram]

A strange thing about Guile's options interface is that xxx-options,
xxx-enable and xxx-disable all want a quoted option symbol as their
argument, whereas xxx-set! wants an _unquoted_ symbol.

What is worse, is that if you call xxx-set! with a _quoted_ symbol, as
in (debug-set! 'stack 500), it does nothing, but silently -- there is
no clear indication that what you asked for didn't work.

(It may be arguable that people should expect xxx-set! to take an
unquoted symbol by analogy with set! ...  but personally I've always
found the current behaviour confusing, and I frequently get it wrong.)

The patch below enhances xxx-set! so that it accepts both an unquoted
and a quoted symbol, and also simplifies some of the rather obfuscated
definition of define-option-interface.  With this enhancement, we keep
backwards compatibility, but we can document in future that the
options functions always take a quoted symbol. 

What do you think?

(Further thought - if this enhancement is good, would it be even
better, if optexp is not a symbol, just to let it be evaluated?  This
would make xxx-set! even more similar to xxx-enable and xxx-disable,
as the latter are procedures.)

        Neil

cd /home/neil/Guile/cvs/guile-core/ice-9/
diff -Naur /home/neil/Guile/cvs/guile-core/ice-9/boot-9.scm.old /home/neil/Guile/cvs/guile-core/ice-9/boot-9.scm
--- /home/neil/Guile/cvs/guile-core/ice-9/boot-9.scm.old	Tue Oct 29 00:09:13 2002
+++ /home/neil/Guile/cvs/guile-core/ice-9/boot-9.scm	Tue Oct 29 00:10:01 2002
@@ -2026,24 +2026,32 @@
 			 (begin (,interface (append (,interface)
 						    (list '(,'unquote name)
 							  (,'unquote exp))))
-				(,interface)))))))
+				(,interface))))))
+
+	 (get-option-symbol (lambda (optexp)
+			      (cond ((symbol? optexp)
+				     optexp)
+				    ((and (list? optexp)
+					  (= (length optexp) 2)
+					  (eq? (car optexp) 'quote))
+				     (cadr optexp))
+				    (else
+				     (error "Option must be specified as a symbol or quoted symbol"))))))
     (procedure->memoizing-macro
      (lambda (exp env)
-       (cons 'begin
-	     (let* ((option-group (cadr exp))
-		    (interface (car option-group)))
-	       (append (map (lambda (name constructor)
-			      `(define ,name
-				 ,(constructor interface)))
-			    (cadr option-group)
-			    (list make-options
-				  make-enable
-				  make-disable))
-		       (map (lambda (name constructor)
-			      `(defmacro ,name
-				 ,@(constructor interface)))
-			    (caddr option-group)
-			    (list make-set!)))))))))
+       (let* ((option-group (cadr exp))
+	      (interface (car option-group))
+	      (options/enable/disable (cadr option-group)))
+	 `(begin
+	    (define ,(car options/enable/disable)
+	      ,(make-options interface))
+	    (define ,(cadr options/enable/disable)
+	      ,(make-enable interface))
+	    (define ,(caddr options/enable/disable)
+	      ,(make-disable interface))
+	    (defmacro ,(caaddr option-group) (opt val)
+	      `(,,(car options/enable/disable)
+		(list ',(,get-option-symbol opt) ,val)))))))))
 
 (define-option-interface
   (eval-options-interface

Diff finished at Tue Oct 29 00:10:11



_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://mail.gnu.org/mailman/listinfo/guile-devel

Re: Small enhancement to options interface

[Neil Jerram]

>>>>> "Neil" == Neil Jerram <neil@ossau.uklinux.net> writes:

    Neil> A strange thing about Guile's options interface is that
    Neil> xxx-options, xxx-enable and xxx-disable all want a quoted
    Neil> option symbol as their argument, whereas xxx-set! wants an
    Neil> _unquoted_ symbol. [...]

Further^2 thoughts on this: actually it's a can of worms.  Allowing
xxx-set! to take a quoted symbol makes it look like it does
evaluation (which we can't make it do because that would be
non-back-compatible), and then someone will get very confused by code
like this:

(define stack depth)
(debug-set! stack 400)     ; intending to set the 'depth option

I'll just apply the deobfuscation part of this patch, and hope that a
better approach to options in general comes along soon ...

        Neil



_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://mail.gnu.org/mailman/listinfo/guile-devel