why does `variable-at-point' return 0?

why does `variable-at-point' return 0?

[Drew Adams]

Just curious.  Why doesn't `variable-at-point' return nil if there is no
variable at point?

All I can think of is that for some reason someone wanted to test using
`symbolp' instead of testing using `not' (or `and').  But I cannot imagine why.

Checking how `variable-at-point' is actually used in the source code did not
enlighten me.  Each use just checks whether the value is `symbolp' and goes on
from there.  An actual value of zero is never used AFAICT, except to serve as a
non-symbol.

There is no special need to distinguish nil from any variable - in Emacs Lisp
you cannot give the symbol nil a symbol-value other than nil.

It is true that (boundp nil) is true, so `boundp' is not a test of variableness.
But I don't see why it wouldn't be just as easy to test whether the value is nil
instead of testing whether it is a symbol.  What am I missing?

RE: why does `variable-at-point' return 0?

[Drew Adams]

ping

> Just curious.  Why doesn't `variable-at-point' return nil if 
> there is no variable at point?
> 
> All I can think of is that for some reason someone wanted to 
> test using `symbolp' instead of testing using `not' (or `and').
> But I cannot imagine why.
> 
> Checking how `variable-at-point' is actually used in the 
> source code did not enlighten me.  Each use just checks whether
> the value is `symbolp' and goes on from there.  An actual value
> of zero is never used AFAICT, except to serve as a non-symbol.
> 
> There is no special need to distinguish nil from any variable - in 
> Emacs Lisp you cannot give the symbol nil a symbol-value other than nil.
> 
> It is true that (boundp nil) is true, so `boundp' is not a 
> test of variableness.  But I don't see why it wouldn't be just as
> easy to test whether the value is nil instead of testing whether it
> is a symbol.  What am I missing?

Re: why does `variable-at-point' return 0?

[Štěpán Němec]

On Mon, 16 Apr 2012 20:07:47 +0200
Drew Adams wrote:

> ping
>
>> Just curious.  Why doesn't `variable-at-point' return nil if 
>> there is no variable at point?

[...]

>> What am I missing?

Dunno if you're missing anything, but you're certainly not the only one
wondering. FWIW, for my purposes I've been using the function below ever
since I discovered `variable-at-point':

(defun .variable-at-point ()
  "Sanitized version of the utterly [CENSORED] `variable-at-point'.
\(The latter returns 0 as the failure value. Very useful,
indeed.)"
  (let ((v (variable-at-point)))
    (unless (eq v 0) v)))

-- 
Štěpán

RE: why does `variable-at-point' return 0?

[Drew Adams]

> >> What am I missing?
> 
> Dunno if you're missing anything, but you're certainly not 
> the only one wondering. FWIW, for my purposes I've been
> using the function below ever since I discovered `variable-at-point':
> ...
>   (let ((v (variable-at-point)))
>     (unless (eq v 0) v)))

Thanks for confirming that I might not be missing anything.

Maybe Emacs Dev would consider putting this in line with other at-point
functions.

How about deprecating `variable-at-point' and defining a new function
`var-at-point' that DTRT?

(I'm still curious how `variable-at-point' ever got to be defined that way. I
still do wonder if there wasn't perhaps a good reason at the time. I cannot
imagine one, but I've been surprised before.)

Re: why does `variable-at-point' return 0?

[Stefan Monnier]

>> All I can think of is that for some reason someone wanted to 
>> test using `symbolp' instead of testing using `not' (or `and').
>> But I cannot imagine why.

That's most likely the reason, so as not to get tripped up by a variable
whose name happens to be "nil" (unlikely in Lisp, admittedly).


        Stefan

Re: why does `variable-at-point' return 0?

[Juanma Barranquero]

On Thu, Apr 19, 2012 at 23:52, Stefan Monnier <monnier@iro.umontreal.ca> wrote:

> That's most likely the reason, so as not to get tripped up by a variable
> whose name happens to be "nil" (unlikely in Lisp, admittedly).

`variable-at-point' is used five times in the lisp/* sources, and in
every case there's a check like

  (and (symbolp val) val)

or

  (if (eq val 0) nil val)

so it seems like v-a-p needs an upgrade.

    Juanma

Re: why does `variable-at-point' return 0?

[Štěpán Němec]

On Thu, 19 Apr 2012 23:52:22 +0200
Stefan Monnier wrote:

>>> All I can think of is that for some reason someone wanted to 
>>> test using `symbolp' instead of testing using `not' (or `and').
>>> But I cannot imagine why.
>
> That's most likely the reason, so as not to get tripped up by a variable
> whose name happens to be "nil" (unlikely in Lisp, admittedly).
                                  ^^^^^^^^

AFAIK, in "nil-lisps" like Elisp or Common Lisp that is not unlikely,
but impossible. Could you explain what you had in mind?

-- 
Štěpán

Re: why does `variable-at-point' return 0?

[Juanma Barranquero]

On Fri, Apr 20, 2012 at 20:14, Štěpán Němec <stepnem@gmail.com> wrote:

> AFAIK, in "nil-lisps" like Elisp or Common Lisp that is not unlikely,
> but impossible. Could you explain what you had in mind?

C:\> emacs -Q --batch --eval "(unintern 'nil)" --eval "(message \"nil
= %%S\" (let ((nil t)) nil))"
nil = t

    Juanma

Re: why does `variable-at-point' return 0?

[Stefan Monnier]

>>>> All I can think of is that for some reason someone wanted to 
>>>> test using `symbolp' instead of testing using `not' (or `and').
>>>> But I cannot imagine why.
>> That's most likely the reason, so as not to get tripped up by a variable
>> whose name happens to be "nil" (unlikely in Lisp, admittedly).
> AFAIK, in "nil-lisps" like Elisp or Common Lisp that is not unlikely,
> but impossible.

In the context of the discussion, the difference between "unlikely" and
"impossible" is irrelevant.

> Could you explain what you had in mind?

I work in formal methods, and from that point of view, "impossible"
would basically mean you have a proof that it can't happen, yet in
languages like Lisp "can't happen" is pretty much ... "unlikely"?


        Stefan