bug#24153: 25.1; Feature request: per-buffer cursor-color variable

bug#24153: 25.1; Feature request: per-buffer cursor-color variable

[Tej Chajed]

Currently there's a per-buffer cursor-type variable that is respected
per window and a single cursor-color variable that applies to the
entire frame.

Evil emulates a per-window cursor color in order to have the cursor
color of the active window reflect the buffer state (eg,
distinguishing insert and normal mode). It does so by advising
set-window and calling set-cursor-color if necessary; this can lead to
slowdowns when code calls set-window repeatedly (eg, through
with-selected-window).

It seems like the right way for evil to manage the cursor color is to
have an analogous per-buffer cursor-color variable natively provided
by emacs. Is this a doable feature?

Thanks!
Tej

bug#24153: 25.1; Feature request: per-buffer cursor-color variable

[Eli Zaretskii]

> From: Tej Chajed <tchajed@mit.edu>
> Date: Thu, 4 Aug 2016 11:08:12 -0400
> 
> It seems like the right way for evil to manage the cursor color is to
> have an analogous per-buffer cursor-color variable natively provided
> by emacs. Is this a doable feature?

I don't see up front why not.  Patches welcome.

Thanks.

bug#24153: 25.1; Feature request: per-buffer cursor-color variable

[npostavs]

Tej Chajed <tchajed@mit.edu> writes:
>
> Evil emulates a per-window cursor color in order to have the cursor
> color of the active window reflect the buffer state (eg,
> distinguishing insert and normal mode). It does so by advising
> set-window and calling set-cursor-color if necessary; this can lead to
> slowdowns when code calls set-window repeatedly (eg, through
> with-selected-window).

There's no `set-window' function, did you mean `select-window'?  I find
in the docstring for that function:

    Selections that "really count" are those causing a visible change in
    the next redisplay of WINDOW’s frame and should be always recorded.
    So if you think of running a function each time a window gets
    selected put it on ‘buffer-list-update-hook’.

Perhaps using ‘buffer-list-update-hook’ instead of advising
`select-window' would help?

bug#24153: 25.1; Feature request: per-buffer cursor-color variable

[Tej Chajed]

On Thu, Aug 4, 2016 at 9:21 PM,  <npostavs@users.sourceforge.net> wrote:
> Tej Chajed <tchajed@mit.edu> writes:
>>
>> Evil emulates a per-window cursor color in order to have the cursor
>> color of the active window reflect the buffer state (eg,
>> distinguishing insert and normal mode). It does so by advising
>> set-window and calling set-cursor-color if necessary; this can lead to
>> slowdowns when code calls set-window repeatedly (eg, through
>> with-selected-window).
>
> There's no `set-window' function, did you mean `select-window'?

Yes, my mistake - the advice is on select-window.

>
> Perhaps using ‘buffer-list-update-hook’ instead of advising
> `select-window' would help?

I haven't fully debugged this alternative, but it doesn't seem to
work. company-coq still uses save-window-excursion in
company-coq-ask-prover, which triggers the buffer-list-update-hook. In
addition, even with company-coq disabled this somehow doesn't call
evil-refresh-cursor enough; starting from insert mode, after the first
proof command, the color gets stuck on the normal mode color, and then
doesn't change until some other buffer list update.

bug#24153: 25.1; Feature request: per-buffer cursor-color variable

[Clément Pit--Claudel]

[-- Attachment #1.1: Type: text/plain, Size: 642 bytes --]

On 2016-08-04 21:21, npostavs@users.sourceforge.net wrote:
> Perhaps using ‘buffer-list-update-hook’ instead of advising
> `select-window' would help?

Hmmm. I don't think that would work.

Assume there are two windows, one showing a buffer in evil's normal editing mode, and one showing a buffer in evil's edit mode.  If the user has configured evil to show the normal-mode cursor in green and the insert-mode cursor in purple, then evil needs to change the cursor color every time the user switches between these two windows, and (I don't think that) buffer-list-update-hook won't be sufficient for that.

Cheers,
Clément.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

bug#24153: 25.1; Feature request: per-buffer cursor-color variable

[martin rudalics]

 >>> Evil emulates a per-window cursor color in order to have the cursor
 >>> color of the active window reflect the buffer state (eg,
 >>> distinguishing insert and normal mode). It does so by advising
 >>> set-window and calling set-cursor-color if necessary; this can lead to
 >>> slowdowns when code calls set-window repeatedly (eg, through
 >>> with-selected-window).
 >>
 >> There's no `set-window' function, did you mean `select-window'?
 >
 > Yes, my mistake - the advice is on select-window.
 >
 >>
 >> Perhaps using ‘buffer-list-update-hook’ instead of advising
 >> `select-window' would help?
 >
 > I haven't fully debugged this alternative, but it doesn't seem to
 > work. company-coq still uses save-window-excursion in
 > company-coq-ask-prover, which triggers the buffer-list-update-hook.

Please tell us more about where and why ‘buffer-list-update-hook’
doesn't work in any of your use cases.  From what you say here you
expect ‘save-window-excursion’ to _not_ run ‘buffer-list-update-hook’.
Is that correct?

 > In
 > addition, even with company-coq disabled this somehow doesn't call
 > evil-refresh-cursor enough; starting from insert mode, after the first
 > proof command, the color gets stuck on the normal mode color, and then
 > doesn't change until some other buffer list update.

But IIUC you do not set the buffer of any window in these steps, so why
would you expect ‘buffer-list-update-hook’ to run?

martin

bug#24153: 25.1; Feature request: per-buffer cursor-color variable

[martin rudalics]

 > Assume there are two windows, one showing a buffer in evil's normal
 > editing mode, and one showing a buffer in evil's edit mode.  If the
 > user has configured evil to show the normal-mode cursor in green and
 > the insert-mode cursor in purple, then evil needs to change the cursor
 > color every time the user switches between these two windows, and (I
 > don't think that) buffer-list-update-hook won't be sufficient for
 > that.

This would be a bug.  ‘select-window’ should always run
‘buffer-list-update-hook’ when NORECORD is nil.

martin

bug#24153: 25.1; Feature request: per-buffer cursor-color variable

[Clément Pit--Claudel]

[-- Attachment #1.1: Type: text/plain, Size: 649 bytes --]

On 2016-08-05 12:37, martin rudalics wrote:
>> Assume there are two windows, one showing a buffer in evil's normal
>> editing mode, and one showing a buffer in evil's edit mode.  If the
>> user has configured evil to show the normal-mode cursor in green and
>> the insert-mode cursor in purple, then evil needs to change the cursor
>> color every time the user switches between these two windows, and (I
>> don't think that) buffer-list-update-hook won't be sufficient for
>> that.
> 
> This would be a bug.  ‘select-window’ should always run
> ‘buffer-list-update-hook’ when NORECORD is nil.

I see; thanks for clarifying!



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

bug#24153: 25.1; Feature request: per-buffer cursor-color variable

[Clément Pit--Claudel]

[-- Attachment #1.1: Type: text/plain, Size: 596 bytes --]

On 2016-08-04 22:17, Tej Chajed wrote:
>>> Perhaps using ‘buffer-list-update-hook’ instead of advising
>>> `select-window' would help?
>
> I haven't fully debugged this alternative, but it doesn't seem to
> work. company-coq still uses save-window-excursion in
> company-coq-ask-prover, which triggers the buffer-list-update-hook. 

IIRC I added this save-window-excursion call to work a Proof General issue, back when I didn't know much about PG.  Now that I have commit rights there, I could probably move the fix upstream and remove the save-window-excursion call.

Clément.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]