Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!uunet!mcsun!ukc!edcastle!aiai!jeff
From: jeff@aiai.ed.ac.uk (Jeff Dalton)
Newsgroups: comp.lang.lisp
Subject: Re: updating random slots in (say) defstruct
Message-ID: <1571@skye.ed.ac.uk>
Date: 17 Jan 90 18:05:58 GMT
References: <18837@bellcore.bellcore.com> <25B37814.3312@marob.masa.com>
Reply-To: jeff@aiai.UUCP (Jeff Dalton)
Organization: AIAI, University of Edinburgh, Scotland
Lines: 54

In article <25B37814.3312@marob.masa.com> cowan@marob.masa.com (John Cowan) writes:
>In article <18837@bellcore.bellcore.com> pc@flash.bellcore.com writes:
>>Must i explicitly define my structure as say a list, to be able to
>>replace an arbitrary component?

>This is handled on page 96 of CLtL (1st ed.), which discusses setf.
>That is section 7.2, (setf), seventh bullet, in case paginations differ.
>It is legal to say (setf (apply #'fn args) newval).

>In this case, one could say something like:
>	(setf instance (make-fubar :one 'a :two 'b :three 'c))
>	(setf accessor #'fubar-one)	; note use of #' here
>	(setf (apply accessor (list instance)) newvalue)
>to change the :one slot of instance to newvalue.

What CLtL says is that SETF will work for a call to APPLY provided
that (1) APPLY's first argument has the form #'function-name and
(2) calls to function-name are recognized as places by SETF.

That is, to be able to do

   (setf (apply #'f ...apply-args...) y)

you have to be able to do

   (setf (f ...args...) y)

In your example above, ACCESSOR would not be recognized as a place
by SETF, nor do you write #'ACCESSOR.  So

   (setf (apply accessor ...) ...)

probably won't work.  On the other hand,

   (setf (apply #'fubar-one ...) ...)

looks like it ought to work.  Unfortunately it doesn't in any of the
Common Lisps I've tried.

>The book warns that this works only if the setf method for the function
>is of a normal kind, i.e. produces an update function that resembles the
>access function except for having a final argument added which is the new
>value.

That's probably it.  Presumably nothing says the SETF methods for
structure slot accessors have the required property.  Too bad if so.

>So the solution is somewhat implementation dependent, but should work
>on any "reasonable" implementation.  (I have not tested it past one
>implementation.)

Does (SETF (APPLY ACCESSOR ...) ...) really work somewhere?

-- Jeff