Strange incf behavior in Emacs 22. Intentional?

Strange incf behavior in Emacs 22. Intentional?

[spamfilteraccount]

Here are two functions:

(defun testfun1 ()
  (interactive)
  (setq a '(0 0))
  (print a))

(defun testfun2 ()
  (interactive)
  (incf (car a))
  (print a))


I call the first one, a is (0 0).
I call the second one, a is (1 0).
I call the second one, a is (2 0).
I call the first one, a is (2 0).

"a" won't get the value (0 0) again until I redefine testfun1.

I suspect it is caused be some kind of internal optimization. Is it
intentional? It cost me a great deal of head scratching to find out
why my program doesn't work.

Re: Strange incf behavior in Emacs 22. Intentional?

[Thien-Thi Nguyen]

() "spamfilteraccount@gmail.com" <spamfilteraccount@gmail.com>
() Thu, 28 Jun 2007 05:41:31 -0700

   why my program doesn't work.

expression:
'(0 0)

differs from expression:
(list 0 0)

but how/why?

thi

Re: Strange incf behavior in Emacs 22. Intentional?

[Lennart Borgman (gmail)]

spamfilteraccount@gmail.com wrote:
> Here are two functions:
> 
> (defun testfun1 ()
>   (interactive)
>   (setq a '(0 0))
>   (print a))
> 
> (defun testfun2 ()
>   (interactive)
>   (incf (car a))
>   (print a))
> 
> 
> I call the first one, a is (0 0).
> I call the second one, a is (1 0).
> I call the second one, a is (2 0).
> I call the first one, a is (2 0).
> 
> "a" won't get the value (0 0) again until I redefine testfun1.
> 
> I suspect it is caused be some kind of internal optimization. Is it
> intentional? It cost me a great deal of head scratching to find out
> why my program doesn't work.


Are you using Emacs 22? Could you then please file a bug report?

Re: Strange incf behavior in Emacs 22. Intentional?

[Juanma Barranquero]

On 6/28/07, spamfilteraccount@gmail.com

> (defun testfun1 ()
>   (interactive)
>   (setq a '(0 0))
>   (print a))

The expresion '(0 0) builds a list when you're defining testfun1. That
list is being modified by testfun2().

If you use

  (setq a (list 0 0))

you build a new list every time that testfun1 is executed.

Take a look at the Emacs Lisp Introduction and the Emacs Lisp
Reference; they're bundled with Emacs 22 (C-h I to access the Info
Directory).

             Juanma

Re: Strange incf behavior in Emacs 22. Intentional?

[David Kastrup]

"Lennart Borgman (gmail)" <lennart.borgman@gmail.com> writes:

> spamfilteraccount@gmail.com wrote:
>> Here are two functions:
>>
>> (defun testfun1 ()
>>   (interactive)
>>   (setq a '(0 0))
>>   (print a))
>>
>> (defun testfun2 ()
>>   (interactive)
>>   (incf (car a))
>>   (print a))
>>
>>
>> I call the first one, a is (0 0).
>> I call the second one, a is (1 0).
>> I call the second one, a is (2 0).
>> I call the first one, a is (2 0).
>>
>> "a" won't get the value (0 0) again until I redefine testfun1.
>>
>> I suspect it is caused be some kind of internal optimization. Is it
>> intentional? It cost me a great deal of head scratching to find out
>> why my program doesn't work.
>
>
> Are you using Emacs 22? Could you then please file a bug report?

There is no bug, and no optimization involved.  This is expected and
correct behavior (anything else would be _wrong_).  A quoted list is
not regenerated when a function is entered new.  A quoted list is
evaluated to a cons cell at _compile_ time.  The contents of this cons
cell can be changed with setcar (or, in this case, incf).

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: Strange incf behavior in Emacs 22. Intentional?

[Lennart Borgman (gmail)]

David Kastrup wrote:
> "Lennart Borgman (gmail)" <lennart.borgman@gmail.com> writes:
> 
>> spamfilteraccount@gmail.com wrote:
>>> Here are two functions:
>>>
>>> (defun testfun1 ()
>>>   (interactive)
>>>   (setq a '(0 0))
>>>   (print a))
>>>
>>> (defun testfun2 ()
>>>   (interactive)
>>>   (incf (car a))
>>>   (print a))
>>>
>>>
>>> I call the first one, a is (0 0).
>>> I call the second one, a is (1 0).
>>> I call the second one, a is (2 0).
>>> I call the first one, a is (2 0).
>>>
>>> "a" won't get the value (0 0) again until I redefine testfun1.
>>>
>>> I suspect it is caused be some kind of internal optimization. Is it
>>> intentional? It cost me a great deal of head scratching to find out
>>> why my program doesn't work.
>>
>> Are you using Emacs 22? Could you then please file a bug report?
> 
> There is no bug, and no optimization involved.  This is expected and
> correct behavior (anything else would be _wrong_).  A quoted list is
> not regenerated when a function is entered new.  A quoted list is
> evaluated to a cons cell at _compile_ time.  The contents of this cons
> cell can be changed with setcar (or, in this case, incf).


Ah, yes. I forgot about this case, but Juanma explained this before here.

Re: Strange incf behavior in Emacs 22. Intentional?

[Stefan Monnier]

> A quoted list is evaluated to a cons cell at _compile_ time.

Actually, it's done at "read" time (aka "load" time).  I.e. when you load
the .el or .elc file, or when you hit C-x C-e on the (defun ...).

> The contents of this cons cell can be changed with setcar (or, in this
> case, incf).

And I consider this as a bug, actually, but a known one that's a bit
difficult to fix satisfactorily.


        Stefan

Re: Strange incf behavior in Emacs 22. Intentional?

[David Kastrup]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> A quoted list is evaluated to a cons cell at _compile_ time.
>
> Actually, it's done at "read" time (aka "load" time).  I.e. when you
> load the .el or .elc file, or when you hit C-x C-e on the (defun
> ...).

Sure.

>> The contents of this cons cell can be changed with setcar (or, in
>> this case, incf).
>
> And I consider this as a bug, actually, but a known one that's a bit
> difficult to fix satisfactorily.

Why would that be a bug?

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: Strange incf behavior in Emacs 22. Intentional?

[Stefan Monnier]

>>> A quoted list is evaluated to a cons cell at _compile_ time.
>> 
>> Actually, it's done at "read" time (aka "load" time).  I.e. when you
>> load the .el or .elc file, or when you hit C-x C-e on the (defun
>> ...).

> Sure.

>>> The contents of this cons cell can be changed with setcar (or, in
>>> this case, incf).
>> 
>> And I consider this as a bug, actually, but a known one that's a bit
>> difficult to fix satisfactorily.

> Why would that be a bug?

Because immediate literal values such as '(1 2) are conceptually part of the
code rather than part of the data, so setcar on them will change program
code which is against my religion.

Of course, this might be considered moot since my religion also
opposes setcar (and aset on strings).


        Stefan

Re: Strange incf behavior in Emacs 22. Intentional?

[Lennart Borgman (gmail)]

> Because immediate literal values such as '(1 2) are conceptually part of the
> code rather than part of the data, so setcar on them will change program
> code which is against my religion.
> 
> Of course, this might be considered moot since my religion also
> opposes setcar (and aset on strings).


All religions (as well as political parties) have and should have their 
contradictions. That is life.

I would say that the first part, changing '(1 2), is very confusing. 
Consider debugging for example.

Re: Strange incf behavior in Emacs 22. Intentional?

[David Kastrup]

"Lennart Borgman (gmail)" <lennart.borgman@gmail.com> writes:

>> Because immediate literal values such as '(1 2) are conceptually part of the
>> code rather than part of the data, so setcar on them will change program
>> code which is against my religion.
>>
>> Of course, this might be considered moot since my religion also
>> opposes setcar (and aset on strings).
>
>
> All religions (as well as political parties) have and should have
> their contradictions. That is life.
>
> I would say that the first part, changing '(1 2), is very
> confusing. Consider debugging for example.

"Doctor, it hurts when I do this."  "Don't do this then".

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum