Path: utzoo!attcan!uunet!ogicse!caesar.cs.montana.edu!uakari.primate.wisc.edu!zaphod.mps.ohio-state.edu!think!snorkelwacker!bloom-beacon!SCU.BITNET!RDAVIS
From: RDAVIS@SCU.BITNET
Newsgroups: comp.lang.scheme
Subject: (atom? '()) => #t
Message-ID: <9003131152.aa20965@mintaka.lcs.mit.edu>
Date: 13 Mar 90 14:21:00 GMT
Sender: daemon@athena.mit.edu (Mr Background)
Organization: The Internet
Lines: 61




In article <1990Mar7.194937.1421@sun.soe.clarkson.edu>, jk0@sun.soe.clarkson.edu
 (Jason Coughlin) writes:
> Can someone explain the rationale behind treating '() as an atom?  It
> seems to me that it should be a pair:  it's the empty *list* not the
> empty *atom*.
>

Jacob L. Cybulski replies:
> When you look at lists from a purely functional point of view then
> any list is the result of CONS, whereas () is a primitive object,
> thus an atom....

and goes on to define the functional implementation of PAIRS

Vincent Mannis writes:
>   But remember, a list isn't a pair.
EXACTLY!!! but then he goes on to say...
> It's zero or more pairs connected together via a cdr chain.
NO!!

I am truly amazed at the apparent confusion between the data structure
LIST and its representation via PAIRs. LIST is an inductively defined
structure:
        1) basis case: the-empty-list is a LIST
        2) constructive case: if L is a LIST, then the result of
        adding X to the front L is a LIST.

A very convenient REPRESENTATION of this data structure is defined
using the atom NIL for the-empty-list, and the PAIR (or SYMBOLIC
EXPRESSION) constructor CONS for the list constructor. A different
representation could be chosen, but a list is still a list.

Perhaps the (intellectual) confusion could be avoided by using
LIST-CONS, FIRST, and REST, instead of CONS, CAR and CDR. In terms of
implementation, one is unlikely to define and use a predicate IS-LIST?
as it would have to cdr-down the pair it is given to see if it started
with NIL.

Victor Mannis also points out, correctly:
>  In fact, a pair is an object to which the car
>  and cdr procedures may be applied; it has been considered unkosher for
>  many years to allow (car '()) and (cdr '())."

to which Andrew Shalit replies:
>  Only unkosher in Scheme, not in Common Lisp.

"unkosher" is a very subjective term. Those who consider taking car and
cdr of nil to be unkosher (myself included), consider it so in Scheme and
in Common Lisp (and UCI-, Mac-, Inter-, Rutgers-, Franz-, TLC-, PSL, 1.6,...
1.5). Even when cdr of an atom gets you its property list, it is not
a nice thing to do - relying on implementation details. But all of this
goes back to age old arguments of the merits of "pure" lisp vs incredibly
powerful hacking. I am reminded of Drew McDermott's comment at the 1980
LISP Conference that what he really liked about LISP was that "I can feel
the bits between my toes".  but I digress...


Ruth E. Davis