NetNews Usenet Archive 1993 #1/NN_1993_1.iso/spool/comp/lang/lisp/3220

Internet Message Format | 1993-01-09 | 2.5 KB

Path: sparky!uunet!think.com!barmar
From: [e-mail] (Barry Margolin)
Newsgroups: comp.lang.lisp
Subject: Re: Improper lists
Date: 9 Jan 1993 19:41:40 GMT
Organization: Thinking Machines Corporation, Cambridge MA, USA
Lines: 42
Message-ID: <[e-mail]>
References: <[e-mail]>
NNTP-Posting-Host: gandalf.think.com

In article <[e-mail]> [e-mail] (William K. Foster) writes:
>This definition does not state whether an empty chain of conses is a
>list. An empty chain must be a list in order for nil to be a list.
>However, using the same reasoning, all atoms are lists, nil being a
>proper list and all other atoms being improper lists.

Given the definitions as they're currently written, you're correct: all
objects are lists (well, it *is* a "list processing" language :-).  I don't
think this was the original intent, though; I'm looking into getting the
definition of "dotted list" changed to require at least one cons (and refer
to this in the definition of "list").

Note also that there are two definitions of "list" in the glossary.  The
second definition is objects of type (OR CONS NULL), and atoms aren't lists
according to that.  The first definition describes the concept of lists,
while the second one describes the type list.  Since the first definition
is currently useless for discrimination, we may decide that any use of the
term for descrimination must intend the second definition.

>nth is defined to take a second argument which is a list.

This isn't a problem, even without the above clarifications.  The beginning
of the Conses chapter says:

  "Except as explicitly specified otherwise, any \term{standardized}
  \term{function} that takes a \term{parameter} that required to be a
  \term{list} should be prepared to signal an error \oftype{type-error} if
  the \term{value} received is a \term{dotted list}."

A problem with this is that it permits implementations to signal an error
when the argument is a dotted list, even if it doesn't need to access the
end of the list; for instance, (NTH '(A . B) 0) could signal an error (the
"should be prepared to signal" phrase allows an implementation to omit
checking, but it doesn't prevent checking).  Therefore, there's been a
proposal to change many of these functions to allow dotted list arguments.
In that case my clarifications would solve the problem of (NTH <atom> N)
being allowed.
--
Barry Margolin
System Manager, Thinking Machines Corp.

[e-mail]          {uunet,harvard}!think!barmar