Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!dino!ux1.cso.uiuc.edu!cs.uiuc.edu!yonezawa
From: yonezawa@cs.uiuc.edu (Noritake Yonezawa)
Newsgroups: comp.lang.lisp
Subject: Re: Mirror speeds
Message-ID: <YONEZAWA.90Oct14005632@m.m.cs.uiuc.edu>
Date: 14 Oct 90 05:56:32 GMT
References: <47064@cornell.UUCP> <raja.655852887@copper>
Sender: news@ux1.cso.uiuc.edu (News)
Distribution: comp
Organization: University of Illinois at Urbana-Champaign
Lines: 27
In-Reply-To: raja@copper.ucs.indiana.edu's message of 13 Oct 90 21:21:27 GMT

In article <raja.655852887@copper> raja@copper.ucs.indiana.edu (Raja Sooriamurthi) writes:

> Here is a CPSed form of the mirror function. Now all calls to the
> auxillary function mir are properly tail recursive and a scheme
> implementation (and some CL implementations) would optimize the
> recursive calls into jumps.

Here is another form of the mirror function.
The auxillary functions are CPSed, but not tail recursive.
I avoided using 'append'.
Does anyone know better solution?

(defun mirror (x)
  (labels ((f1 (x c)
	     (if (atom x)
		 (funcall c x)
		 (f2 x nil c)))
	   (f2 (x y c)
	     (if (null x)
		 (funcall c y)
		 (f1 (car x) #'(lambda (z) (f2 (cdr x) (cons z y) c))))))
    (f1 x #'identity)))

--
Noritake Yonezawa [yonezawa@cs.uiuc.edu]
Department of Computer Science
University of Illinois at Urbana-Champaign