Re: 'let' for functions ?

[Oliver Scholz <alkibiades@gmx.de>]

Ed L Cashin <ecashin@uga.edu> writes:
[...]
> As an exercise I tried in vain to create a simple implementation of
> labels that wouldn't require loading cl.  I couldn't do it, though,
> and had to get on with what I was supposed to be doing.  ;)

You need not fear the cl package. If you put 
`(eval-when-compile (require 'cl)' into your code, the compiled
package won't require cl.

However, since this is, as you say, a nice exercise, I give it a
shot. (Although I am supposed to be doing something different, too
...)

> I'd be very interested, though, to see such a simple implementation of
> labels if there's an elisp guru who would enjoy trying to find it.
> The hard part for me was recursion support.

I am not a Lisp-Guru, but here is my version anyways. I'd like to hear
whether there is any problem with this approach:

(defmacro my-simple-labels (spec &rest body)
  ;; We `fset' the symbols temporarily and restore the original state
  ;; after the body form was executed.
  (let ((symbol (make-symbol "symbol"))
	(definition (make-symbol "definition"))
	(sfunc (make-symbol "sfunc")))
    `(progn
       (let (,symbol)
	 (dolist (,definition ',spec)
	   (setq ,symbol (car ,definition))
	   ;; Store function definition, if necessary.
	   (when (fboundp ,symbol)
	     (put ,symbol 'my-simple-labels (symbol-function ,symbol)))
	   ;; New function definition.
	   (fset ,symbol (cons 'lambda (cdr ,definition)))))
       (unwind-protect (progn ,@body)
	 ;; Clean-up
	 (let (,sfunc)
	   (dolist (,symbol ',(mapcar 'car spec))
	     (setq ,sfunc (get ,symbol 'my-simple-labels))
	     (if ,sfunc
		 (fset ,symbol ,sfunc)
	       (fmakunbound ,symbol))))))))


    Oliver
-- 
7 Floréal an 211 de la Révolution
Liberté, Egalité, Fraternité!