Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!cbosgd!gatech!seismo!utah-cs!shebs
From: shebs@utah-cs.UUCP (Stanley Shebs)
Newsgroups: net.lang.lisp
Subject: Re: question for a Common Lisp guru
Message-ID: <3706@utah-cs.UUCP>
Date: Wed, 5-Mar-86 17:35:59 EST
Article-I.D.: utah-cs.3706
Posted: Wed Mar  5 17:35:59 1986
Date-Received: Sat, 8-Mar-86 04:32:02 EST
References: <1829@hammer.UUCP> <1142@mit-eddie.UUCP> <1840@hammer.UUCP>
Reply-To: shebs@utah-cs.UUCP (Stanley Shebs)
Distribution: net
Organization: University of Utah CS Dept
Lines: 41

In article <1840@hammer.UUCP> patcl@hammer.UUCP (Pat Clancy) writes:
>In article <1142@mit-eddie.UUCP> barmar@eddie.UUCP (Barry Margolin) writes:
>>>There does not seem to be any special form which could be
>>>used to implement such a macro [defun].
>>
>>It can expand into
>>	(setf (symbol-value ...
>
>Unfortunately, setf is itself a macro. So what is the final real
>form that that defun expands to?

Common Lisp doesn't specify it.  Remember, Common Lisp is "just" an
interchange standard - Steele's book does *not* describe any particular
implementation!  So in the case of defun, the implementation is free
to expand it into whatever implementation-dependent stuff it feels like
producing.  Pretty entertaining, no?  There *is* one suggested restriction
(the note on p.58), that implementations not expand macros into
implementation-dependent special forms.  Implementation-dependent
*functions* are OK tho, and in fact that's what you have to use.
(setf (symbol-function ...)) eventually expands into %set-symbol-function
or some such.  In the case of our Common Lisp (PCLS) which is built
on top of PSL, defun eventually turns into a call on the PSL function
PutD.

Almost all the setf functions are this way.  You will find GET, but not
PUT - that's because half the existing Lisp implementations use the argument
order (put symbol indicator value), and the other half do it as (put symbol
value indicator).  Common Lisp gets some additional compatibility by
not specifying it, and saying that "GET is setfable".

All this brings up an important question: what is the minimum set of
Common Lisp constructs needed to implement the rest of Common Lisp?
We've worked on this question, and decided it's very hard.  As the
above examples show, the absence of functions like PUT means that
you must have SETF of GET in the Common Lisp "kernel", which means
that you have to have SETF, which probably means that DEFINE-SETF-METHOD
has to be in this kernel, ad infinitum...  It's not easy to define
*any* subsets of Common Lisp!

>Pat Clancy
							stan shebs