Is this a bug of Emacs-Lisp?

Is this a bug of Emacs-Lisp?

[Zhang Wei]

(defun dummy () '(1 . 2))

(dummy)
 => (1 . 2)

(setcdr (dummy) 3)

(dummy)
 => (1 . 3)

Modify the return value of dummy changed it's defination. Is this
a bug of Elisp? If it's not. How does this happen?


-- 
Zhang Wei or Brep
<brep@smth.org>

Re: Is this a bug of Emacs-Lisp?

[Giorgos Keramidas]

On 2006-02-17 13:47, Zhang Wei <id.brep@gmail.com> wrote:
> (defun dummy () '(1 . 2))
>
> (dummy)
>  => (1 . 2)
>
> (setcdr (dummy) 3)
>
> (dummy)
>  => (1 . 3)
>
> Modify the return value of dummy changed it's defination. Is this
> a bug of Elisp? If it's not. How does this happen?

You are modifying a `literal' cons cell.  It's not mandatory for the
LISP reader to create a new cons cell every time the (dummy) function
runs.

If you really want to create a new cons cell every time (dummy) runs,
you'll have to explicitly do that in the function:

    (defun dummy ()
      (copy-tree '(1 . 2)))
    => dummy

    (setcdr (dummy) 3)
    => 3

    (dummy)
    => '(1 . 2)

RE: Is this a bug of Emacs-Lisp?

[Drew Adams]

    If you really want to create a new cons cell every time (dummy) runs,
    you'll have to explicitly do that in the function:

        (defun dummy () (copy-tree '(1 . 2)))

Sure, but that's really doing things the hard way! If you want a new cons
cell, then function `cons' is your man - no need to beat around the bush:

 (defun dummy () (cons 1 . 2))

RE: Is this a bug of Emacs-Lisp?

[Drew Adams]

I wrote (defun dummy () (cons 1 . 2))
but of course meant (defun dummy () (cons 1 2))

Re: Is this a bug of Emacs-Lisp?

[Andreas Schwab]

"Drew Adams" <drew.adams@oracle.com> writes:

> Sure, but that's really doing things the hard way! If you want a new cons
> cell, then function `cons' is your man - no need to beat around the bush:
>
>  (defun dummy () (cons 1 . 2))

Except that you want to drop the period.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Is this a bug of Emacs-Lisp?

[Andreas Schwab]

Zhang Wei <id.brep@gmail.com> writes:

> (defun dummy () '(1 . 2))
>
> (dummy)
>  => (1 . 2)
>
> (setcdr (dummy) 3)
>
> (dummy)
>  => (1 . 3)
>
> Modify the return value of dummy changed it's defination. Is this
> a bug of Elisp? If it's not. How does this happen?

You can write FORTRAN in any language. :-) If you are destructively
modifying a constant, you'll get such results.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Is this a bug of Emacs-Lisp?

[David Kastrup]

Zhang Wei <id.brep@gmail.com> writes:

> (defun dummy () '(1 . 2))
>
> (dummy)
>  => (1 . 2)
>
> (setcdr (dummy) 3)
>
> (dummy)
>  => (1 . 3)
>
> Modify the return value of dummy changed it's defination.

No, it didn't.

> Is this a bug of Elisp? If it's not. How does this happen?

'(1 . 2) is a cons cell.  Cons cells are _always_ accessed by
reference, like arrays and strings.

So what is compiled into the function is a reference to a cons cell,
and this reference stays the same.  setcdr however changes the cons
cell itself.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: Is this a bug of Emacs-Lisp?

[Luc Teirlinck]

Zhang Wei wrote:

   (defun dummy () '(1 . 2))

   (dummy)
    => (1 . 2)

   (setcdr (dummy) 3)

   (dummy)
    => (1 . 3)

   Modify the return value of dummy changed it's defination. Is this
   a bug of Elisp?

No.  You get the same result in Common Lisp if you replace, in the
above, setcdr with rplacd.

   If it's not. How does this happen?

If a function returns a quoted cons, then after you change the cdr of
the cons, it still returns that (now changed) cons.

Using (defun dummy (cons 1 2)) will prevent this.

Sincerely,

Luc.

Re: Is this a bug of Emacs-Lisp?

[Luc Teirlinck]

Zhang Wei wrote:

   Modify the return value of dummy changed it's defination. Is this
   a bug of Elisp? If it's not. How does this happen?

>From my previous reply:

    If a function returns a quoted cons, then after you change the cdr of
    the cons, it still returns that (now changed) cons.

More precisely: you change the actual cons stored in the function cell
of the symbol `dummy'.  The function returns that stored cons each
time it is called.  To make it return a new cons each time, make it
return (cons 1 2).

Sincerely,

Luc.

Re: Is this a bug of Emacs-Lisp?

[Alan Mackenzie]

Hi!

On Fri, 17 Feb 2006, Zhang Wei wrote:

>(defun dummy () '(1 . 2))

>(dummy)
> => (1 . 2)

>(setcdr (dummy) 3)

>(dummy)
> => (1 . 3)

>Modify the return value of dummy changed it's defination. Is this
>a bug of Elisp? If it's not. How does this happen?

No, it's not a bug at all.  You haven't changed the definition of dummy
with the setcdr - you've merely altered a cons cell which happened to
have been produced some time earlier by dummy.

Please also note that emacs-devel@gnu.org isn't really the place for this
sort of question - it's a mailing list for discussing the development of
Emacs.  You'd do better by posting to the list <help-gnu-emacs@gnu.org>,
or the equivalent newsgroups, gnu.emacs.help.

>-- 
>Zhang Wei or Brep
><brep@smth.org>

-- 
Alan Mackenzie (Munich, Germany)

Re: Is this a bug of Emacs-Lisp?

[Kevin Rodgers]

Zhang Wei wrote:
> (defun dummy () '(1 . 2))
> 
> (dummy)
>  => (1 . 2)
> 
> (setcdr (dummy) 3)
> 
> (dummy)
>  => (1 . 3)
> 
> Modify the return value of dummy changed it's defination. Is this
> a bug of Elisp? If it's not. How does this happen?

It is not a bug.

After the defun, (symbol-function 'dummy) =>
(lambda nil (quote (1 . 2)))

That is, the function binding of the symbol dummy is a list of 3 elements:
lambda, nil, and (quote (1 . 2))

That last element is the form that is evaluated to produce the result,
which itself is a list of 2 elements: quote and (1 . 2)

Evaluating (quote x) returns x, which in this case is the (1 . 2) cons
cell.  It is not a copy of x, it is x itself, which in this case is a
member of the function binding's sublist.

setcdr is a destructive operation that modifies that object, and by
side-effect, the function definition.

You can even write a self-modifiying function, by having dummy call
setcdr (or any destructive function) on its own function binding.

-- 
Kevin Rodgers

Re: Is this a bug of Emacs-Lisp?

[Sascha Wilde]

Zhang Wei <id.brep@gmail.com> wrote:

> (defun dummy () '(1 . 2))
>
> (dummy)
>  => (1 . 2)
>
> (setcdr (dummy) 3)
>
> (dummy)
>  => (1 . 3)
>
> Modify the return value of dummy changed it's defination.

No.  The function dummy returns a cons cell -- all ways the _same_
cons cell, so if you change it, well it ends up changed.  :-)

What you might want is:

(defun dummy () (cons 1 2))

cheers
sascha
-- 
Sascha Wilde

Lisp is the red pill.
  -- John Fraser, comp.lang.lisp

Re: Is this a bug of Emacs-Lisp?

[Richard M. Stallman]

    Modify the return value of dummy changed it's defination. Is this
    a bug of Elisp?

It is not a bug.  You modified part of the data structure in the
function definition.