Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uwm.edu!zaphod.mps.ohio-state.edu!think.com!barmar
From: barmar@think.com (Barry Margolin)
Newsgroups: comp.lang.lisp
Subject: Re: Question about INTERN
Message-ID: <1991Jan31.174839.20943@Think.COM>
Date: 31 Jan 91 17:48:39 GMT
References: <4037@skye.ed.ac.uk> <1991Jan30.225106.26561@Think.COM> <ALMS.91Jan31114032@ministry.cambridge.apple.com>
Sender: news@Think.COM
Organization: Thinking Machines Corporation, Cambridge MA, USA
Lines: 65

In article <ALMS.91Jan31114032@ministry.cambridge.apple.com> alms@cambridge.apple.com (Andrew L. M. Shalit) writes:
>In article <1991Jan30.225106.26561@Think.COM> barmar@think.com (Barry Margolin) writes:
>Two more suggestions: replace FORMAT with CONCATENATE.  This will
>ensure that you don't get screwed by *print-case*.  

Good point.  I've never really liked Common Lisp's CONCATENATE (my guess is
that 95% of the sequences that are concatenated are lists (which already
have APPEND) and strings (in my Symbolics-specific code I prefer
STRING-APPEND)).  Also, in this case it is clumsy because of the need to
use SYMBOL-NAME to convert the symbol to a string.

>						     Also, you can
>replace the EVAL with a FUNCALL (going by your "smaller hammer for a
>smaller job" theory).  

You're right for the GET-SLOT-VALUE case, but unfortunately not
SET-SLOT-VALUE (I realize you weren't talking about this case, I'm just
amplifying).  In this case, there's some enhanced readability from
implementing both GET and SET in similar fashion.

>			It might also be a good idea to pass a package
>argument to INTERN.  All these together give us:
>
>
>   (defun get-slot-value (obj slot)
>     (when (symbolp obj)
>       (setq obj (symbol-value obj)))
>     (funcall (intern (concatenate 'string
>                                   (string (type-of obj))
>                                   "-"
>                                   (string slot))
>                      (symbol-package slot))       ;just a guess!
>               obj))

I actually considered supplying the package argument.  Unfortunately, there
really is no *right* way to do this, which is why it's usually wrong to try
to construct function names using INTERN at runtime.  The package of the
slot argument to {GET,SET}-SLOT-VALUE is not necessarily the package that
the structure was defined in, especially when inheritance is involved.
Consider:

(defstruct personal-inventory
  car
  house)

(setq my-inv (make-personal-inventory :car :oldsmobile :house :condo))

(get-slot-value my-inv 'car)
Error: Undefined function LISP::PERSONAL-INVENTORY-CAR

or on a Symbolics

Error: Attempt to intern BARMAR in locked package COMMON-LISP.

This is because CAR is inherited from the LISP package.  It would probably
be safer to use (symbol-package (type-of obj)), but that can also have
similar problems (although not involving the LISP package, if the program
obeys the rules against redefining built-in Common Lisp symbols, as
described on p.260 of CLtL2).

--
Barry Margolin, Thinking Machines Corp.

barmar@think.com
{uunet,harvard}!think!barmar