[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
2 problems
I'm not sure why either of these are "problems" except as you may have
depended upon undocumented behaviour which changed:
Date: Mon, 23 May 88 10:12:44 EDT
From: bds@mitre-bedford.ARPA
1) Suppose I have the following:
(defvar *a*)
(defvar *b*)
(defun eq-test ()
(let ((a '(a b c d))
(b '(a b c d)))
(setq *a* a)
(setq *b* b)))
After running eq-test, it turns out that *a* and *b* are eq!
This, by the way, occurs with arrays, strings, etc. The only
way to avoid this is to make liberal use of copy-* in these
functions.
I don't believe there's anything in common-lisp which prohibits
constant-folding. If you explicitly want new structure to be consed
each time, you should make that fact explicit:
(defun eq-test ()
(let ((a (list 'a 'b 'c 'd))
(b (copy-list '(a b c d))))
(setq *a* a)
(setq *b* b)))
Otherwise there's can't see why it shouldn't improve the performance of
your code.
2) Suppose I have a structure named rts which is stored as a
named vector. Given an instance of this structure, rts-1,
if I then do (setq rts-2 (copy-rts rts-1)), I find that although
every slot of rts-2 is eq to the corresponding slot in rts-1,
rts-1 and rts-2 are not equal !
Moreover, given that the nominal-equipment-list slot of rts is a list
of equipment structures, I find that
(eq (rts-nominal-equipment-list rts-1)
(rts-nominal-equipment-list rts-2))
returns t, but when I do
(push t (rts-nominal-equipment-list rts-1))
rts-1 is modified but rts-2 is not!
Look at page 270 of the Silver book. Push is equivalent to
(setf (rts-nominal-equipment-list rts-1) (cons t (rts-nominal-equipment-list rts-1)))
It's a macro.
Look at this example:
(setf foo '(1 2)) => (1 2)
(setf bar foo) => (1 2)
(eq bar foo) => t
(push 3 foo) => (3 1 2)
(eq bar foo) => nil
(eq bar (cdr foo))=> t
Perhaps what you want is something like the following (only works on
lists):
(defun bd-push (elt list)
(without-interrupts
(setf (cdr list) (cons (car list) (cdr list))
(car list) elt))
list)