:extend face attribute at EOB without end-of-line char

:extend face attribute at EOB without end-of-line char

[Juanma Barranquero]

[-- Attachment #1.1: Type: text/plain, Size: 168 bytes --]

Is this effect intended?

[image: image.png]

emacs -Q
M-x bs-show <RET>
a
M-x hl-line-mode <RET>

The last line in the *buffer-selection* buffer has no EOL character.

[-- Attachment #1.2: Type: text/html, Size: 369 bytes --]

[-- Attachment #2: image.png --]
[-- Type: image/png, Size: 11768 bytes --]

Re: :extend face attribute at EOB without end-of-line char

[Ergus]

Hi Juanma:

By default up to now hl-face and region are the only faces that are
extensible by default (as before), so their behavior should be the same
than before; no change should be observed.

I've tried it now in emacs 26 and the actual master branch and up to now
the behavior seems to be the same for me.

If you observe any unexpected difference, please, tell me.

On Mon, Oct 14, 2019 at 06:02:27PM +0200, Juanma Barranquero wrote:
>Is this effect intended?
>
>[image: image.png]
>
>emacs -Q
>M-x bs-show <RET>
>a
>M-x hl-line-mode <RET>
>
>The last line in the *buffer-selection* buffer has no EOL character.

Re: :extend face attribute at EOB without end-of-line char

[Juanma Barranquero]

[-- Attachment #1: Type: text/plain, Size: 329 bytes --]

On Mon, Oct 14, 2019 at 6:32 PM Ergus <spacibba@aol.com> wrote:

> If you observe any unexpected difference, please, tell me.

No, I don't see a difference. I *expected* a difference. I thought the face
would be extended in the last line of the buffer up to the right side of
the window, even if it had no end-of-line character.

[-- Attachment #2: Type: text/html, Size: 444 bytes --]

Re: :extend face attribute at EOB without end-of-line char

[Ergus]

On Mon, Oct 14, 2019 at 06:37:22PM +0200, Juanma Barranquero wrote:
>On Mon, Oct 14, 2019 at 6:32 PM Ergus <spacibba@aol.com> wrote:
>
>> If you observe any unexpected difference, please, tell me.
>
>No, I don't see a difference. I *expected* a difference. I thought the face
>would be extended in the last line of the buffer up to the right side of
>the window, even if it had no end-of-line character.

Ok, I am not sure, but I think that actually bs-show is a table, so the
hl-face does not apply. to it because there is a last column (in this
case empty) with the default face. Maybe Eli will confirm.

Re: :extend face attribute at EOB without end-of-line char

[Juanma Barranquero]

[-- Attachment #1: Type: text/plain, Size: 1189 bytes --]

On Mon, Oct 14, 2019 at 7:07 PM Ergus <spacibba@aol.com> wrote:

> Ok, I am not sure, but I think that actually bs-show is a table, so the
> hl-face does not apply. to it because there is a last column (in this
> case empty) with the default face.

No. The *buffer-selection* buffer is filled with this code in
bs-show-in-buffer:

    (erase-buffer)
    (setq bs--name-entry-length name-entry-length)
    (bs--show-header)
    (dolist (buffer list)
      (bs--insert-one-entry buffer)
      (insert "\n"))
    (delete-char -1)

which inserts lines (strings) followed by \n, and then deletes the last \n
in the buffer (so the last line does not have a \n at the end, which
theoretically simplifies a bit moving around the buffer).

The function that creates each line is bs--insert-one-entry, which
accumulates the different "fields" into one string, and then just inserts
it in the buffer.

The effect in the picture in my original post is not showing an empty
column in a table. It's an artifact of the last line not having a closing
\n. Whether that's intended with face extension or not, I don't know. IMO
it would be nicer if the face extended to the end of the line in that case
too.

[-- Attachment #2: Type: text/html, Size: 1434 bytes --]

Re: :extend face attribute at EOB without end-of-line char

[Juanma Barranquero]

[-- Attachment #1: Type: text/plain, Size: 339 bytes --]

In any case, what I'm suggesting is not just to "fix" hl-line-mode in
bs.el. IMHO, bs.el needs someone* to take a serious look at it. I think it
could benefit from tabulated-mode, and the new window-selection variables
and functions. Currently it does tricks with window configurations which
are a bit of a PITA.

(*Note: Perhaps even me)

[-- Attachment #2: Type: text/html, Size: 381 bytes --]

Re: :extend face attribute at EOB without end-of-line char

[Eli Zaretskii]

> From: Juanma Barranquero <lekktu@gmail.com>
> Date: Mon, 14 Oct 2019 18:37:22 +0200
> Cc: Emacs developers <emacs-devel@gnu.org>
> 
> No, I don't see a difference. I *expected* a difference. I thought the face would be extended in the last line of
> the buffer up to the right side of the window, even if it had no end-of-line character.

No, that was never the intent.  The extension happens only of the face
crosses the line boundary, i.e. it "covers" the newline at the end of
the line.

If the documentation doesn't make that clear, please suggest how to
clarify it.

Re: :extend face attribute at EOB without end-of-line char

[Juanma Barranquero]

[-- Attachment #1: Type: text/plain, Size: 1106 bytes --]

On Mon, Oct 14, 2019 at 7:59 PM Eli Zaretskii <eliz@gnu.org> wrote:

> No, that was never the intent.  The extension happens only of the face
> crosses the line boundary, i.e. it "covers" the newline at the end of
> the line.

"the line boundary" is a bit ambiguous; count-lines counts the last line in
a buffer even if it has no newline char at the end. It wouldn't be absurd
to suppose there's an implicit line boundary between the last line and the
end of the buffer, whether that last line has a newline character or not.

But you say that's not the intent, so ok. I still think it would make more
sense. The bs-show example above shows why could it be useful.

> If the documentation doesn't make that clear, please suggest how to
> clarify it.

I don't know. I'm bad at thinking about documentation in english. I suppose
just adding somewhere: "The face extends from the newline at the end of the
line up to the right margin of the window; it doesn't happen if there's no
newline at the end". But the simple fact of having to clarify that fact
points to it being a bit... unexpected. IMO, of course.

[-- Attachment #2: Type: text/html, Size: 1329 bytes --]

Re: :extend face attribute at EOB without end-of-line char

[Eli Zaretskii]

> From: Juanma Barranquero <lekktu@gmail.com>
> Date: Mon, 14 Oct 2019 21:41:46 +0200
> Cc: spacibba@aol.com, Emacs developers <emacs-devel@gnu.org>
> 
> > No, that was never the intent.  The extension happens only of the face
> > crosses the line boundary, i.e. it "covers" the newline at the end of
> > the line.
> 
> "the line boundary" is a bit ambiguous

That's why I added that it should "cover" the newline.  That's the
meaning of "crosses the line boundary".

Re: :extend face attribute at EOB without end-of-line char

[Juanma Barranquero]

[-- Attachment #1: Type: text/plain, Size: 414 bytes --]

On Mon, Oct 14, 2019 at 9:47 PM Eli Zaretskii <eliz@gnu.org> wrote:

> That's why I added that it should "cover" the newline.  That's the
> meaning of "crosses the line boundary".

Ok. That's how it was defined, and that's how it works. Agreed.

My question is: would it be better, worse or indifferent that, in the
specific case of the last line of the buffer, it worked even without a
boundary newline to cross?

[-- Attachment #2: Type: text/html, Size: 581 bytes --]

Re: :extend face attribute at EOB without end-of-line char

[Eli Zaretskii]

> From: Juanma Barranquero <lekktu@gmail.com>
> Date: Mon, 14 Oct 2019 21:50:22 +0200
> Cc: spacibba@aol.com, Emacs developers <emacs-devel@gnu.org>
> 
> My question is: would it be better, worse or indifferent that, in the specific case of the last line of the buffer, it
> worked even without a boundary newline to cross?

I don't know.  This change didn't _add_ any cases that the face is
extended where it wasn't before.  This change only lets us _refrain_
from extending the face in some cases where we unconditionally
extended before.  We never extended the face in the case which you
describe, so we don't extend it now.

So you are talking about a different, almost unrelated, feature.  It
would have to have its own rationale.

Re: :extend face attribute at EOB without end-of-line char

[Juanma Barranquero]

[-- Attachment #1: Type: text/plain, Size: 223 bytes --]

On Mon, Oct 14, 2019 at 10:04 PM Eli Zaretskii <eliz@gnu.org> wrote:

> So you are talking about a different,

Agreed.

> almost unrelated, feature.

Obviously, we don't define "almost unrelated" the same way. Fair enough.

[-- Attachment #2: Type: text/html, Size: 409 bytes --]

Re: :extend face attribute at EOB without end-of-line char

[Eli Zaretskii]

> From: Juanma Barranquero <lekktu@gmail.com>
> Date: Mon, 14 Oct 2019 23:37:14 +0200
> Cc: spacibba@aol.com, Emacs developers <emacs-devel@gnu.org>
> 
> > So you are talking about a different,
> 
> Agreed.
> 
> > almost unrelated, feature.
> 
> Obviously, we don't define "almost unrelated" the same way. Fair enough.

OK, but still, we'd need a rationale.  The use case is quite obscure,
IMO, and the implementation won't be trivial (I had my share of bugs
from fiddling with display code past ZV, and you just had a taste of
that with the ticks feature).

Re: :extend face attribute at EOB without end-of-line char

[Juanma Barranquero]

[-- Attachment #1: Type: text/plain, Size: 1183 bytes --]

On Tue, Oct 15, 2019 at 8:15 AM Eli Zaretskii <eliz@gnu.org> wrote:

> OK, but still, we'd need a rationale.

I'm not sure what rationale would satisfy you. In my mind, the idea of
extending a face up to the margin of the window seems quite decoupled from
whether that line has a newline at the end or not. It's purely visual. If
I'm using a face that extends in all lines, for example, I find weird that
it does or does not extend in the last one depending of a newline. (Again:
I agree that's how it is defined now, I'm not questioning that.)

> The use case is quite obscure, IMO,

Yes. I happen to use bs-show, and hl-line-mode in bs-show, so it's not
obscure for me, obviously. (Years ago I defined the format of my bs lines
to fill entirely the window with spaces to avoid this ugly "ragged" effect
on the last one.) But whether it will be useful in other cases will depend
on how frequent is to have buffers with a non-newline last line, I suppose.
No idea.

> and the implementation won't be trivial (I had my share of bugs
> from fiddling with display code past ZV, and you just had a taste of
> that with the ticks feature).

Yes. It was quite "entertaining", so to speak.

[-- Attachment #2: Type: text/html, Size: 1503 bytes --]

Re: :extend face attribute at EOB without end-of-line char

[Eli Zaretskii]

> From: Juanma Barranquero <lekktu@gmail.com>
> Date: Tue, 15 Oct 2019 10:10:21 +0200
> Cc: spacibba@aol.com, Emacs developers <emacs-devel@gnu.org>
> 
> > OK, but still, we'd need a rationale. 
> 
> I'm not sure what rationale would satisfy you. In my mind, the idea of extending a face up to the margin of the
> window seems quite decoupled from whether that line has a newline at the end or not. It's purely visual. If I'm
> using a face that extends in all lines, for example, I find weird that it does or does not extend in the last one
> depending of a newline. (Again: I agree that's how it is defined now, I'm not questioning that.)

Face extension is only a factor when it ends on the next line,
i.e. "covers" the newline that ends the line.  When there's no newline
at EOB, the face ends with the last character on the line, so IMO it
makes no sense to extend it, because what would such an extension
indicate?  When there's a newline, extending the face indicates that
the newline (which is otherwise invisible) is "covered" by the face.

Re: :extend face attribute at EOB without end-of-line char

[Juanma Barranquero]

[-- Attachment #1: Type: text/plain, Size: 1446 bytes --]

On Tue, Oct 15, 2019 at 10:27 AM Eli Zaretskii <eliz@gnu.org> wrote:

> Face extension is only a factor when it ends on the next line,
> i.e. "covers" the newline that ends the line.  When there's no newline
> at EOB, the face ends with the last character on the line, so IMO it
> makes no sense to extend it, because what would such an extension
> indicate?  When there's a newline, extending the face indicates that
> the newline (which is otherwise invisible) is "covered" by the face.

I understand, but again, that's an explanation of how it works. For me,
face extension is useful as a way to fill the "empty background" of a line.
It is only vaguely related to the idea of text ending in a newline.

For example, if you open the speedbar and then activate hl-line-mode, the
"lines" are highlighted up to the window margin, though speedbar "lines"
are not conceptually lines of text, but an outline of some structure (of
course they *are* lines of text, implementation-wise). In that case, it
happens that the last item in the speedbar ends with a newline, so
hl-line-mode works as expected. But if the speedbar were implemented like
bs-show is, and the last item had no newline, it would be equally
surprising to see the whole "window line" highlighted in all items, and
just partially in the last one.

(Note: I'm not arguing for this to be implemented; just trying to explain
why I would expect a behavior different of the current one.)

[-- Attachment #2: Type: text/html, Size: 1656 bytes --]

Re: :extend face attribute at EOB without end-of-line char

[Stefan Monnier]

> hl-line-mode works as expected. But if the speedbar were implemented like
> bs-show is, and the last item had no newline, it would be equally
> surprising to see the whole "window line" highlighted in all items, and
> just partially in the last one.

I think it's easier to change bs.el ;-)


        Stefan

Re: :extend face attribute at EOB without end-of-line char

[Juanma Barranquero]

[-- Attachment #1: Type: text/plain, Size: 179 bytes --]

On Tue, Oct 15, 2019 at 3:23 PM Stefan Monnier <monnier@iro.umontreal.ca>
wrote:

> I think it's easier to change bs.el ;-)

Yes, as I said earlier, bs is in need of an overhaul.

[-- Attachment #2: Type: text/html, Size: 306 bytes --]

Re: :extend face attribute at EOB without end-of-line char

[Ergus]

Hi Eli:

Somehow this is related with something we discussed some time ago, about
the fact that we should call extend_face_to_end_of_line in the last line
of the buffer if not empty in some conditions (dfci is active for
example.)

Maybe you remember that we don't have the indicator for the last line,
which somehow we agreed must be corrected. In this case the problem is
the same: the extend_face... function is not called for the latest line
in the buffer but I didn't find a better condition to fix this (I didn't
try very hard either) But probably it just requires to extend a
condition in an if and part of this problem will be fixed (the case for
the last line at least)

There are some conditions in the display_line function to not call
extend_face_to... when the line ends at ZV, fixing this condition we
should be done right?

On Tue, Oct 15, 2019 at 11:27:10AM +0300, Eli Zaretskii wrote:
>> From: Juanma Barranquero <lekktu@gmail.com>
>> Date: Tue, 15 Oct 2019 10:10:21 +0200
>> Cc: spacibba@aol.com, Emacs developers <emacs-devel@gnu.org>
>>
>> > OK, but still, we'd need a rationale.
>>
>> I'm not sure what rationale would satisfy you. In my mind, the idea of extending a face up to the margin of the
>> window seems quite decoupled from whether that line has a newline at the end or not. It's purely visual. If I'm
>> using a face that extends in all lines, for example, I find weird that it does or does not extend in the last one
>> depending of a newline. (Again: I agree that's how it is defined now, I'm not questioning that.)
>
>Face extension is only a factor when it ends on the next line,
>i.e. "covers" the newline that ends the line.  When there's no newline
>at EOB, the face ends with the last character on the line, so IMO it
>makes no sense to extend it, because what would such an extension
>indicate?  When there's a newline, extending the face indicates that
>the newline (which is otherwise invisible) is "covered" by the face.

Re: :extend face attribute at EOB without end-of-line char

[Eli Zaretskii]

> Date: Tue, 15 Oct 2019 18:38:18 +0200
> From: Ergus <spacibba@aol.com>
> Cc: Juanma Barranquero <lekktu@gmail.com>, emacs-devel@gnu.org
> 
> There are some conditions in the display_line function to not call
> extend_face_to... when the line ends at ZV, fixing this condition we
> should be done right?

Yes, but be aware that all screen lines after that last line also "end
at ZV".  So the code should be somewhat smart to only do what you want
on the last line with text.

Re: :extend face attribute at EOB without end-of-line char

[Ergus]

On Tue, Oct 15, 2019 at 07:49:14PM +0300, Eli Zaretskii wrote:
>> Date: Tue, 15 Oct 2019 18:38:18 +0200
>> From: Ergus <spacibba@aol.com>
>> Cc: Juanma Barranquero <lekktu@gmail.com>, emacs-devel@gnu.org
>>
>> There are some conditions in the display_line function to not call
>> extend_face_to... when the line ends at ZV, fixing this condition we
>> should be done right?
>
>Yes, but be aware that all screen lines after that last line also "end
>at ZV".  So the code should be somewhat smart to only do what you want
>on the last line with text.
>
Yes, I know; that's why it didn't work for me before. Also the behavior
was different in tui and gui.

But I haven't time now. Next week maybe I'll give a look.