bug#21505: 24.4; Buffer order

bug#21505: 24.4; Buffer order

[hoppe]

[-- Attachment #1: Type: text/plain, Size: 893 bytes --]

This is prompted by the Stack Overflow question

http://emacs.stackexchange.com/q/16675/9553

Let's say I have a Linux directory that looks like

    -rw-r--r--  1 hooked se     0 Sep 16 16:02 a
    -rw-r--r--  1 hooked se     0 Sep 16 16:02 b
    -rw-r--r--  1 hooked se     0 Sep 16 16:02 c

When I type emacs * it opens all three files but it puts me in the middle
of the buffer chain. For example when I run it, I start off at c and
NextBuffer takes me along the chain

    c -> b -> *Messages* -> *scratch* -> a ->

This is really, really annoying. I'd like to open up emacs and have the
order be any of the permutations

    c -> b -> a -> *Messages* -> *scratch* ->
    b -> a -> c -> *Messages* -> *scratch* ->
    a -> b -> c -> *Messages* -> *scratch* ->

This behavior is the most natural one IMHO (user @Stefan agrees). This is a
feature request to make this the default behavior.

[-- Attachment #2: Type: text/html, Size: 1334 bytes --]

bug#21505: 24.4; Buffer order

[Eli Zaretskii]

> From: hoppe <travis.hoppe@gmail.com>
> Date: Wed, 16 Sep 2015 22:57:07 -0400
> 
> Let's say I have a Linux directory that looks like
> 
> -rw-r--r-- 1 hooked se 0 Sep 16 16:02 a
> -rw-r--r-- 1 hooked se 0 Sep 16 16:02 b
> -rw-r--r-- 1 hooked se 0 Sep 16 16:02 c
> When I type emacs * it opens all three files but it puts me in the middle of
> the buffer chain. For example when I run it, I start off at c and NextBuffer
> takes me along the chain
> 
> c -> b -> *Messages* -> *scratch* -> a ->
> This is really, really annoying. I'd like to open up emacs and have the order
> be any of the permutations
> 
> c -> b -> a -> *Messages* -> *scratch* -> 
> b -> a -> c -> *Messages* -> *scratch* -> 
> a -> b -> c -> *Messages* -> *scratch* -> 
> 
> This behavior is the most natural one IMHO (user @Stefan agrees). This is a
> feature request to make this the default behavior.

Is the feature request to produce the above buffer order only once,
upon entry to Emacs in this scenario, or is it more general,
i.e. should hold at any arbitrary point in time during the Emacs
session?  If the latter, please define the desired buffer order more
generally, since I'm guessing *scratch* and *Messages* are not the
only buffers you dislike.

In any case, the latter possibility is much harder to accomplish; the
former is relatively easy, but I question its usefulness, given that
it's limited to the initial entry.

bug#21505: 24.4; Buffer order

[Tassilo Horn]

Eli Zaretskii <eliz@gnu.org> writes:

>> Let's say I have a Linux directory that looks like
>> 
>> -rw-r--r-- 1 hooked se 0 Sep 16 16:02 a
>> -rw-r--r-- 1 hooked se 0 Sep 16 16:02 b
>> -rw-r--r-- 1 hooked se 0 Sep 16 16:02 c
>> When I type emacs * it opens all three files but it puts me in the middle of
>> the buffer chain. For example when I run it, I start off at c and NextBuffer
>> takes me along the chain
>> 
>> c -> b -> *Messages* -> *scratch* -> a ->
>> This is really, really annoying. I'd like to open up emacs and have the order
>> be any of the permutations
>> 
>> c -> b -> a -> *Messages* -> *scratch* -> 
>> b -> a -> c -> *Messages* -> *scratch* -> 
>> a -> b -> c -> *Messages* -> *scratch* -> 
>> 
>> This behavior is the most natural one IMHO (user @Stefan
>> agrees). This is a feature request to make this the default behavior.
>
> Is the feature request to produce the above buffer order only once,
> upon entry to Emacs in this scenario, or is it more general,
> i.e. should hold at any arbitrary point in time during the Emacs
> session? [...]
>
> In any case, the latter possibility is much harder to accomplish; the
> former is relatively easy, but I question its usefulness, given that
> it's limited to the initial entry.

IMO it could and should apply also to emacsclient invocations.  And to
me, the order a, b, c seems to be the only right one.

Bye,
Tassilo

bug#21505: 24.4; Buffer order

[Eli Zaretskii]

> From: Tassilo Horn <tsdh@gnu.org>
> Cc: hoppe <travis.hoppe@gmail.com>,  21505@debbugs.gnu.org
> Date: Thu, 17 Sep 2015 08:49:51 +0200
> 
> And to me, the order a, b, c seems to be the only right one.

You do realize that the command line said just "*", and the order of
the files Emacs saw was determined by the expansion of that wildcard
by the shell, yes?

Or are you saying that the order of the buffers should always be
alphabetical?

bug#21505: 24.4; Buffer order

[Tassilo Horn]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Tassilo Horn <tsdh@gnu.org>
>> Cc: hoppe <travis.hoppe@gmail.com>,  21505@debbugs.gnu.org
>> Date: Thu, 17 Sep 2015 08:49:51 +0200
>> 
>> And to me, the order a, b, c seems to be the only right one.
>
> You do realize that the command line said just "*", and the order of
> the files Emacs saw was determined by the expansion of that wildcard
> by the shell, yes?

Yes, I've seen that and I assumed that the usual shell expansion is
alphabetical.

> Or are you saying that the order of the buffers should always be
> alphabetical?

No, it should always be as given to emacs/emacsclient, i.e., when I
invoke

  $ emacsclient x a b

I want to have the buffer x selected and using `next-buffer' I'd switch
to a and then b, and then to the other buffers which have existed
before.  Right now, x will be selected, but a and b are far away.

Bye,
Tassilo

bug#21505: 24.4; Buffer order

[Stefan Monnier]

> No, it should always be as given to emacs/emacsclient, i.e., when I
> invoke
>
>   $ emacsclient x a b
>
> I want to have the buffer x selected and using `next-buffer' I'd switch
> to a and then b, and then to the other buffers which have existed
> before.  Right now, x will be selected, but a and b are far away.

That's right.  And same for "emacs x a b".

The precise behavior is a bit more complex since those commands, rather
than just showing the first file may/will also show the buffer-list,
but if exactly one of the files is shown it should be the first in the
list and next-buffer should go in the order in which the file names
were given (until reaching the end of this list at which point it'll
hit things like *Messages* and friends and that's fine).


        Stefan

bug#21505: 24.4; Buffer order

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: Eli Zaretskii <eliz@gnu.org>,  21505@debbugs.gnu.org,  travis.hoppe@gmail.com
> Date: Thu, 17 Sep 2015 08:50:33 -0400
> 
> > No, it should always be as given to emacs/emacsclient, i.e., when I
> > invoke
> >
> >   $ emacsclient x a b
> >
> > I want to have the buffer x selected and using `next-buffer' I'd switch
> > to a and then b, and then to the other buffers which have existed
> > before.  Right now, x will be selected, but a and b are far away.
> 
> That's right.  And same for "emacs x a b".
> 
> The precise behavior is a bit more complex since those commands, rather
> than just showing the first file may/will also show the buffer-list,
> but if exactly one of the files is shown it should be the first in the
> list and next-buffer should go in the order in which the file names
> were given (until reaching the end of this list at which point it'll
> hit things like *Messages* and friends and that's fine).

The questions I asked the OP still stand.  The required feature is not
defined in sufficient detail.  If only the initial order is required,
it's probably relatively easy, but then that order is very fragile and
could easily break as soon as the user issues the first command.

bug#21505: 24.4; Buffer order

[hoppe]

[-- Attachment #1: Type: text/plain, Size: 2190 bytes --]

In response to the first question, I'd like this to be the "easy" request,
that this behavior is only for when emacs opens and not continuously during
the session. I agree, it does not need to be maintained during the session.

Re: Tassilo Horn: I purposely showed my example with a wildcard since emacs
could be given any expansion from the shell. To me it, and for my usual use
case, it doesn't matter. I have a bunch of files I'd like to edit and I
often miss those that are past the messages and scratch buffer. Opening the
files in the order given from the command line seems very natural to me.

So to be clear, given the command run on the shell:

> emacs a c b

The buffer order should be [a,c,b,*messages*,*scratch*]. Currently the
default behavior looks something like [b,c,*messages*,*scratch*, a].

On Thu, Sep 17, 2015 at 10:55 AM, Eli Zaretskii <eliz@gnu.org> wrote:

> > From: Stefan Monnier <monnier@iro.umontreal.ca>
> > Cc: Eli Zaretskii <eliz@gnu.org>,  21505@debbugs.gnu.org,
> travis.hoppe@gmail.com
> > Date: Thu, 17 Sep 2015 08:50:33 -0400
> >
> > > No, it should always be as given to emacs/emacsclient, i.e., when I
> > > invoke
> > >
> > >   $ emacsclient x a b
> > >
> > > I want to have the buffer x selected and using `next-buffer' I'd switch
> > > to a and then b, and then to the other buffers which have existed
> > > before.  Right now, x will be selected, but a and b are far away.
> >
> > That's right.  And same for "emacs x a b".
> >
> > The precise behavior is a bit more complex since those commands, rather
> > than just showing the first file may/will also show the buffer-list,
> > but if exactly one of the files is shown it should be the first in the
> > list and next-buffer should go in the order in which the file names
> > were given (until reaching the end of this list at which point it'll
> > hit things like *Messages* and friends and that's fine).
>
> The questions I asked the OP still stand.  The required feature is not
> defined in sufficient detail.  If only the initial order is required,
> it's probably relatively easy, but then that order is very fragile and
> could easily break as soon as the user issues the first command.
>

[-- Attachment #2: Type: text/html, Size: 3423 bytes --]

bug#21505: 24.4; Buffer order

[Eli Zaretskii]

> From: hoppe <travis.hoppe@gmail.com>
> Date: Thu, 17 Sep 2015 10:59:41 -0400
> Cc: Stefan Monnier <monnier@iro.umontreal.ca>, tsdh@gnu.org, 21505@debbugs.gnu.org
> 
> In response to the first question, I'd like this to be the "easy" request, that
> this behavior is only for when emacs opens and not continuously during the
> session. I agree, it does not need to be maintained during the session. 
> 
> Re: Tassilo Horn: I purposely showed my example with a wildcard since emacs
> could be given any expansion from the shell. To me it, and for my usual use
> case, it doesn't matter. I have a bunch of files I'd like to edit and I often
> miss those that are past the messages and scratch buffer. Opening the files in
> the order given from the command line seems very natural to me.
> 
> So to be clear, given the command run on the shell:
> 
> > emacs a c b
> 
> The buffer order should be [a,c,b,*messages*,*scratch*]. Currently the default
> behavior looks something like [b,c,*messages*,*scratch*, a]. 

So it's just an issue with the order in which buffers are presented in
the prompt of "C-x b"?

bug#21505: 24.4; Buffer order

[hoppe]

[-- Attachment #1: Type: text/plain, Size: 1492 bytes --]

No, and this is why it's a little weird. On my computer when I create the
files a,b,c and run

> emacs a b c

the prompt of C-x b gives:

c
b
a
*scratch*
*Messages*

with file c opened first. BUT, the order of NextBuffer goes
[c->b->*Messages*->*scratch->a]

On Thu, Sep 17, 2015 at 11:19 AM, Eli Zaretskii <eliz@gnu.org> wrote:

> > From: hoppe <travis.hoppe@gmail.com>
> > Date: Thu, 17 Sep 2015 10:59:41 -0400
> > Cc: Stefan Monnier <monnier@iro.umontreal.ca>, tsdh@gnu.org,
> 21505@debbugs.gnu.org
> >
> > In response to the first question, I'd like this to be the "easy"
> request, that
> > this behavior is only for when emacs opens and not continuously during
> the
> > session. I agree, it does not need to be maintained during the session.
> >
> > Re: Tassilo Horn: I purposely showed my example with a wildcard since
> emacs
> > could be given any expansion from the shell. To me it, and for my usual
> use
> > case, it doesn't matter. I have a bunch of files I'd like to edit and I
> often
> > miss those that are past the messages and scratch buffer. Opening the
> files in
> > the order given from the command line seems very natural to me.
> >
> > So to be clear, given the command run on the shell:
> >
> > > emacs a c b
> >
> > The buffer order should be [a,c,b,*messages*,*scratch*]. Currently the
> default
> > behavior looks something like [b,c,*messages*,*scratch*, a].
>
> So it's just an issue with the order in which buffers are presented in
> the prompt of "C-x b"?
>

[-- Attachment #2: Type: text/html, Size: 2292 bytes --]

bug#21505: 24.4; Buffer order

[Eli Zaretskii]

> From: hoppe <travis.hoppe@gmail.com>
> Date: Thu, 17 Sep 2015 11:23:58 -0400
> Cc: Stefan Monnier <monnier@iro.umontreal.ca>, tsdh@gnu.org, 21505@debbugs.gnu.org
> 
> No, and this is why it's a little weird. On my computer when I create the files
> a,b,c and run 
> 
> > emacs a b c 
> 
> the prompt of C-x b gives:
> 
> c 
> b
> a
> *scratch*
> *Messages*
> 
> with file c opened first. BUT, the order of NextBuffer goes
> [c->b->*Messages*->*scratch->a]

What or who is NextBuffer??  I cannot find any such string in the
entire Emacs source tree.  What am I missing?

bug#21505: 24.4; Buffer order

[hoppe]

[-- Attachment #1: Type: text/plain, Size: 997 bytes --]

I'm sorry, this is my first bug report and I'm not a power user of emacs.

When I say "Next Buffer" I mean, when I click on Buffers in the File Menu
one of the options is

Next Buffer <XF86FORWARD>

with the tooltip: Switch to the "next" buffer in a cyclic order. This, I
think, is identical to C-x right arrow.


On Thu, Sep 17, 2015 at 11:44 AM, Eli Zaretskii <eliz@gnu.org> wrote:

> > From: hoppe <travis.hoppe@gmail.com>
> > Date: Thu, 17 Sep 2015 11:23:58 -0400
> > Cc: Stefan Monnier <monnier@iro.umontreal.ca>, tsdh@gnu.org,
> 21505@debbugs.gnu.org
> >
> > No, and this is why it's a little weird. On my computer when I create
> the files
> > a,b,c and run
> >
> > > emacs a b c
> >
> > the prompt of C-x b gives:
> >
> > c
> > b
> > a
> > *scratch*
> > *Messages*
> >
> > with file c opened first. BUT, the order of NextBuffer goes
> > [c->b->*Messages*->*scratch->a]
>
> What or who is NextBuffer??  I cannot find any such string in the
> entire Emacs source tree.  What am I missing?
>

[-- Attachment #2: Type: text/html, Size: 1772 bytes --]

bug#21505: 24.4; Buffer order

[Eli Zaretskii]

> From: hoppe <travis.hoppe@gmail.com>
> Date: Thu, 17 Sep 2015 11:48:37 -0400
> Cc: Stefan Monnier <monnier@iro.umontreal.ca>, Tassilo Horn <tsdh@gnu.org>, 21505@debbugs.gnu.org
> 
> When I say "Next Buffer" I mean, when I click on Buffers in the File Menu one
> of the options is 
> 
> Next Buffer <XF86FORWARD> 
> 
> with the tooltip: Switch to the "next" buffer in a cyclic order. This, I think,
> is identical to C-x right arrow.

Ah, okay.  Now everything is clear.  Thanks.

bug#21505: 24.4; Buffer order

[Lars Ingebrigtsen]

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

> The precise behavior is a bit more complex since those commands, rather
> than just showing the first file may/will also show the buffer-list,
> but if exactly one of the files is shown it should be the first in the
> list and next-buffer should go in the order in which the file names
> were given (until reaching the end of this list at which point it'll
> hit things like *Messages* and friends and that's fine).

This behaviour is still present in Emacs 29.

But I'm wondering -- why is Emacs popping up a *Buffer List* window when
you ask it to open two files?  I'd have thought the obvious thing to do
in this situation is to open as many windows as there are files (if
possible)?  Or open a single window displaying the first file specified
(and then have the rest be in the next-buffer order you describe).

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#21505: 24.4; Buffer order

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Cc: Tassilo Horn <tsdh@gnu.org>,  Eli Zaretskii <eliz@gnu.org>,
>   travis.hoppe@gmail.com,  21505@debbugs.gnu.org
> Date: Sun, 20 Feb 2022 13:57:03 +0100
> 
> But I'm wondering -- why is Emacs popping up a *Buffer List* window when
> you ask it to open two files?

It doesn't.  It only shows *Buffer List* when you as it to visit 3 or
more files.  This is explicitly coded in command-line-1:

    ;; Display the first two buffers in `displayable-buffers'.  If
    ;; `initial-buffer-choice' is non-nil, its buffer will be the
    ;; first buffer in `displayable-buffers'.  The first buffer will
    ;; be focused.
    (let ((displayable-buffers-len (length displayable-buffers))
          ;; `nondisplayed-buffers-p' is true if there exist buffers
          ;; in `displayable-buffers' that were not displayed to the
          ;; user.
          (nondisplayed-buffers-p nil))
      (when (> displayable-buffers-len 0)
        (switch-to-buffer (car displayable-buffers)))
      (when (> displayable-buffers-len 1)
        (switch-to-buffer-other-window (car (cdr displayable-buffers)))
        ;; Focus on the first buffer.
        (other-window -1))
      (when (> displayable-buffers-len 2)  <<<<<<<<<<<<<<<<<<<<<<<<<<<<
        (setq nondisplayed-buffers-p t))

> I'd have thought the obvious thing to do in this situation is to
> open as many windows as there are files (if possible)?

I guess the rationale is that with too many files visited we cannot be
sure which of them the user would like to see first, and we don't want
to show them all, lest the windows become too small.

> Or open a single window displaying the first file specified
> (and then have the rest be in the next-buffer order you describe).

That'd be less useful than what we do now with 2 files, IMO.

bug#21505: 24.4; Buffer order

[Lars Ingebrigtsen]

[-- Attachment #1: Type: text/plain, Size: 1069 bytes --]

Eli Zaretskii <eliz@gnu.org> writes:

> It doesn't.  It only shows *Buffer List* when you as it to visit 3 or
> more files.  This is explicitly coded in command-line-1:

Ah, right.

>> I'd have thought the obvious thing to do in this situation is to
>> open as many windows as there are files (if possible)?
>
> I guess the rationale is that with too many files visited we cannot be
> sure which of them the user would like to see first, and we don't want
> to show them all, lest the windows become too small.

Yes, I guess it does make sense to display the *Buffer list* when
there's a lot of files.

>> Or open a single window displaying the first file specified
>> (and then have the rest be in the next-buffer order you describe).
>
> That'd be less useful than what we do now with 2 files, IMO.

Yeah.

So I guess the only this to fix here is the buffer ordering, as the
original bug reporter was talking about -- i.e., ensure that *scratch*
and *Messages* are at the bottom of the list.  Currently, this is the
buffer order if given "a b c d e f" as the files:


[-- Attachment #2: Type: image/png, Size: 71645 bytes --]

[-- Attachment #3: Type: text/plain, Size: 336 bytes --]


Which seems pretty chaotic.  I guess what we want to see here is

f
e
d
c
b
a
*scratch*
*Messages*

?  I mean, sorting the other way around (and selecting a) would also be
a possibility, but would be a greater behavioural change. 

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#21505: 24.4; Buffer order

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Cc: monnier@iro.umontreal.ca,  tsdh@gnu.org,  travis.hoppe@gmail.com,
>   21505@debbugs.gnu.org
> Date: Mon, 21 Feb 2022 14:38:19 +0100
> 
> I guess what we want to see here is
> 
> f
> e
> d
> c
> b
> a
> *scratch*
> *Messages*
> 
> ?  I mean, sorting the other way around (and selecting a) would also be
> a possibility, but would be a greater behavioural change. 

It'd be nice to have the order that is somehow related to the order of
visiting the files, yes.

bug#21505: 24.4; Buffer order

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

> It'd be nice to have the order that is somehow related to the order of
> visiting the files, yes.

I've now done this in Emacs 29.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no