Tweaking save-some-buffers-action-alist

Tweaking save-some-buffers-action-alist

[Sergey Organov]

Hello,

In this discussion:

https://lists.gnu.org/archive/html/help-gnu-emacs/2021-06/msg00165.html

I was pointed by Eli to `some-some-buffers-action-alist` as possible way
to get the feature I've asked for (revert buffer from `same-some-buffers`
prompt). I believe that this capability is really missing when
`enable-recursive-minibuffers` is nil, that is now the default.

This doesn't seem to have direct support for user-level customization,
so I came-up with the following tweak to my init.el that "works" for me,
but I'm not sure I actually do it right, so please help me with this:

#+BEGIN_SRC emacs-lisp
(nconc
 save-some-buffers-action-alist
 '((?\C-\M-g
    (lambda (buf)
      (with-current-buffer buf
        (revert-buffer)))
    "revert this buffer")))
#+END_SRC

I'm concerned as I use `nconc` that sounds unsafe, and then original
definition has some ",(" and ",(purecopy" tricks that I dunno if I need
to follow, and if so, how?

And to finally justify posting to 'emacs-devel', does it make sense to
add this capability to the default value of
`save-some-buffers-action-alist`?

To give you some more context for thoughts, not only I'm often get
annoyed myself by the absence of the feature, but I recently got this
question: "how do I discard changes to the buffer when I run `compile`
and Emacs asks me if I want to save some?" from a relatively novice
Emacs user I'm trying to help, and I really hate to answer: "no way."

Thanks,
-- Sergey Organov

Re: Tweaking save-some-buffers-action-alist

[Daniel Martín]

Sergey Organov <sorganov@gmail.com> writes:

>
> #+BEGIN_SRC emacs-lisp
> (nconc
>  save-some-buffers-action-alist
>  '((?\C-\M-g
>     (lambda (buf)
>       (with-current-buffer buf
>         (revert-buffer)))
>     "revert this buffer")))
> #+END_SRC
>
> I'm concerned as I use `nconc` that sounds unsafe, and then original
> definition has some ",(" and ",(purecopy" tricks that I dunno if I need
> to follow, and if so, how?

Purecopy is a no-op except while Emacs is being built and dumped, so
user customizations can omit it.  Evaluate (info "(elisp) Pure Storage")
in Emacs to read more about this.

Backquoting the alist will instruct the Elisp interpreter to not
evaluate KEY, but evaluate the lambda FUNC.  Evaluating lambda
expressions is generally a good idea because that enables
byte-compilation and all its benefits: Faster code, less memory usage,
some static checks, specially under lexical scope, etc.

>
> And to finally justify posting to 'emacs-devel', does it make sense to
> add this capability to the default value of
> `save-some-buffers-action-alist`?
>

I think reverting a buffer when compiling is not a very common need.
More importantly, the prompt already has a lot of options and adding one
more will make it more complex and difficult to understand.  At some
point you have to draw the line between what should be a customization
and what should be included in Emacs and I think your suggestion can
perfectly be a personal customization.

Re: Tweaking save-some-buffers-action-alist

[Sergey Organov]

Daniel Martín <mardani29@yahoo.es> writes:

> Sergey Organov <sorganov@gmail.com> writes:
>
>>
>> #+BEGIN_SRC emacs-lisp
>> (nconc
>>  save-some-buffers-action-alist
>>  '((?\C-\M-g
>>     (lambda (buf)
>>       (with-current-buffer buf
>>         (revert-buffer)))
>>     "revert this buffer")))
>> #+END_SRC
>>
>> I'm concerned as I use `nconc` that sounds unsafe, and then original
>> definition has some ",(" and ",(purecopy" tricks that I dunno if I need
>> to follow, and if so, how?
>
> Purecopy is a no-op except while Emacs is being built and dumped, so
> user customizations can omit it.  Evaluate (info "(elisp) Pure Storage")
> in Emacs to read more about this.
>
> Backquoting the alist will instruct the Elisp interpreter to not
> evaluate KEY, but evaluate the lambda FUNC.  Evaluating lambda
> expressions is generally a good idea because that enables
> byte-compilation and all its benefits: Faster code, less memory usage,
> some static checks, specially under lexical scope, etc.

Thanks for clarification and pointers!

Is `nconc` safe here, or should I `copy-seq` the original and modify
that instead?

>
>>
>> And to finally justify posting to 'emacs-devel', does it make sense to
>> add this capability to the default value of
>> `save-some-buffers-action-alist`?
>>
>
> I think reverting a buffer when compiling is not a very common need.

Surprisingly, it /is/ rather common when one has a lot of (often
unrelated to compilation) buffers, and finding the buffer in question
after compilation just to revert it is boring and is often forgotten
until the next compilation. Annoying.

> More importantly, the prompt already has a lot of options and adding one
> more will make it more complex and difficult to understand.

Well, yes, but I believe that the "Discard changes" option is the second
or third by usefulness among the rest supported: "Save Changes, Discard
Changes, Save All". To me it's "Ignore the change and continue" that
seems to be the least useful one.

I just want to have a way to finish entire prompt series and have no
modified buffers at the end. Doesn't seem to be too much to ask for.

As a side note, if there are too many options already, 'C-r' could be
removed and the buffer in question be shown automatically instead.

> At some point you have to draw the line between what should be a
> customization and what should be included in Emacs

Yep. I just think that the line in this particular case has been chosen
long ago, when `enable-recursive-minibuffers` was enabled by default, or
was unheard of, and now, when it is disabled by default, the line should
be drawn at a different level, making it possible again to conveniently
revert the buffer.

> and I think your suggestion can perfectly be a personal customization.

Probably it could, but right now it isn't. I don't think we can expect a
novice user to be able to tweak it like I did above, no way. Do you
in fact suggest to support this through `customize` or what?

The current state is that to the actual question I got from a novice
user: "How do I discard changes to the buffer when I run `compile` and
Emacs asks me if I want to save some?" I have to either give the "No
way" answer, or tweak Emacs for them. Sad.

BTW, if there were a way to cancel the `save-some-buffers` after 'C-r'
yet leave the modified buffer current, it'd solve the issue as well,
though in a different, less convenient way.

Thanks,
-- Sergey Organov

Re: Tweaking save-some-buffers-action-alist

[Stefan Monnier]

> Is `nconc` safe here, or should I `copy-seq` the original and modify
> that instead?

The general rule is: If you don't know, then it's not safe.
[ and indeed, it's not safe in this particular case.  ]


        Stefan

Re: Tweaking save-some-buffers-action-alist

[Sergey Organov]

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

>> Is `nconc` safe here, or should I `copy-seq` the original and modify
>> that instead?
>
> The general rule is: If you don't know, then it's not safe.
> [ and indeed, it's not safe in this particular case.  ]

Nice, thanks for confirming!

So I ended-up adding new item at the beginning of the list rather than
at the end:

#+BEGIN_SRC emacs-lisp
(push
  '(?R
    (lambda (buf)
      (with-current-buffer buf
        (revert-buffer)))
    "revert this buffer")
  save-some-buffers-action-alist)
#+END_SRC

And then I had to change the KEY from C-M-g to just R, as on terminal it
seems to be impossible to generate the former key-sequence from
`map-y-or-n-p`, as "ESC C-g" doesn't work as expected (ESC is simply
ignored), – yet another surprising limitation of the mode.

Thanks,
-- 
Sergey Organov