Re: Fix nreverse

Re: Fix nreverse

[Barry OReilly]

[-- Attachment #1: Type: text/plain, Size: 904 bytes --]

> I also fix nreverse doc-string not to mention its implementation
> details.

That the sequence is mutated is not an implementation detail. It is a
side effect that should be documented as part of the interface.

  (let ((lst '(1 2 3)))
    (message "Reversed: %s" (reverse lst))
    (message "Original: %s" lst)
    (message "NReversed: %s" (nreverse lst))
    (message "Original: %s" lst))
  ;; Output:
  ;; Reversed: (3 2 1)
  ;; Original: (1 2 3)
  ;; NReversed: (3 2 1)
  ;; Original: (1)

The user may well care about the difference.

> This would make nreverse and reverse symmetric.

  (let ((str "123"))
    (message "Reversed: %s" (reverse str))
    (message "Original: %s" str)
    (message "NReversed: %s" (nreverse str))
    (message "Original: %s" str))
  ;; Output:
  ;; Reversed: 321
  ;; Original: 123
  ;; NReversed: 321
  ;; Original: 123

This is not symmetrical to the lst example.

[-- Attachment #2: Type: text/html, Size: 1177 bytes --]

Re: Fix nreverse

[Stefan Monnier]

> That the sequence is mutated is not an implementation detail. It is a
> side effect that should be documented as part of the interface.



>   (let ((lst '(1 2 3)))
>     (message "Reversed: %s" (reverse lst))
>     (message "Original: %s" lst)
>     (message "NReversed: %s" (nreverse lst))
>     (message "Original: %s" lst))
>   ;; Output:
>   ;; Reversed: (3 2 1)
>   ;; Original: (1 2 3)
>   ;; NReversed: (3 2 1)
>   ;; Original: (1)

This last "Original" value should not be depended upon by the user.


        Stefan

Re: Fix nreverse

[Leo Liu]

On 2014-05-21 10:07 -0400, Barry OReilly wrote:
> That the sequence is mutated is not an implementation detail. It is a
> side effect that should be documented as part of the interface.

No disagreement and that is how nreverse is documented. What I meant is
really implementation detail i.e. how we set the list pointers to
reverse a list.

Leo

Fix nreverse

[Leo Liu]

Hi there,

I find the decision to leave out strings in nreverse inconvenient. Why
don't we dispatch to reverse when SEQ is string instead of throwing an
error? This would make nreverse and reverse symmetric. I hope "N" mean
we `may' not `must' destructively modify the argument. Objections?

Leo

Re: Fix nreverse

[Stefan Monnier]

> I find the decision to leave out strings in nreverse inconvenient.  Why
> don't we dispatch to reverse when SEQ is string instead of throwing an
> error? This would make nreverse and reverse symmetric.  I hope "N" mean
> we `may' not `must' destructively modify the argument. Objections?

I guess it's OK if we change the meaning of nreverse to "may modify
destructively", but then please change the docstring to say so, and
*generically* (rather than saying explicitly how that's interpreted for
each possible type).


        Stefan

Re: Fix nreverse

[Leo Liu]

On 2014-05-20 22:53 -0400, Stefan Monnier wrote:
> I guess it's OK if we change the meaning of nreverse to "may modify
> destructively", but then please change the docstring to say so, and
> *generically* (rather than saying explicitly how that's interpreted for
> each possible type).

Done. I also fix nreverse doc-string not to mention its implementation
details.

Leo