bug#16545: 24.3.50; Patch to coalesce calls to linum-update-current

bug#16545: 24.3.50; Patch to coalesce calls to linum-update-current

[Chow Loong Jin]

[-- Attachment #1.1: Type: text/plain, Size: 479 bytes --]

When navigating using C-n/C-p with a high key repeat rate, causing
linum-schedule to be called many times in quick succession,
linum-update-current gets called as many times in linum-schedule,
eventhough one call per round should be enough to renumber the lines.

This patch adds a linum-delay-seconds option defaulting to 10ms to delay
the call to linum-update-current, thus coalescing all calls to
linum-update-current until Emacs settles for 10ms.

--
Kind regards,
Loong Jin

[-- Attachment #1.2: 0001-Add-customization-option-to-delay-linum-updates.patch --]
[-- Type: text/x-diff, Size: 1999 bytes --]

From 13a9fde7a9b0a4f7a40e6d5e9ce3447bbfed777b Mon Sep 17 00:00:00 2001
From: Chow Loong Jin <hyperair@debian.org>
Date: Fri, 24 Jan 2014 09:47:19 +0800
Subject: [PATCH] Add customization option to delay linum updates

When linum-schedule is called many times in quick succession, delay
the update by linum-delay-seconds, and coalesce all calls to
linum-update-current to avoid useless repeated calls.

Signed-off-by: Chow Loong Jin <hyperair@debian.org>
---
 lisp/linum.el | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/lisp/linum.el b/lisp/linum.el
index d91ce11..50759b5 100644
--- a/lisp/linum.el
+++ b/lisp/linum.el
@@ -37,6 +37,7 @@
 (defvar linum-available nil "Overlays available for reuse.")
 (defvar linum-before-numbering-hook nil
   "Functions run in each buffer before line numbering starts.")
+(defvar linum-timer-object nil "Timer object for use with ilnum-delay")
 
 (mapc #'make-variable-buffer-local '(linum-overlays linum-available))
 
@@ -71,6 +72,10 @@ and you have to scroll or press \\[recenter-top-bottom] to update the numbers."
   "Delay updates to give Emacs a chance for other changes."
   :group 'linum
   :type 'boolean)
+(defcustom linum-delay-seconds 0.01
+  "Number of seconds to delay before running linum-update-window"
+  :group 'linum
+  :type 'floatp)
 
 ;;;###autoload
 (define-minor-mode linum-mode
@@ -195,7 +200,12 @@ Linum mode is a buffer-local minor mode."
 
 (defun linum-schedule ()
   ;; schedule an update; the delay gives Emacs a chance for display changes
-  (run-with-idle-timer 0 nil #'linum-update-current))
+  (if linum-timer-object
+      (cancel-timer linum-timer-object))
+  (setq linum-timer-object
+        (run-with-timer linum-delay-seconds
+                        nil
+                        #'linum-update-current)))
 
 ;; (defun linum-after-config ()
 ;;   (walk-windows (lambda (w) (linum-update (window-buffer w))) nil 'visible))
-- 
1.8.3.2


[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

bug#16545: 24.3.50; Patch to coalesce calls to linum-update-current

[Glenn Morris]

Thanks for the patch.
The usual answer to anything involving linum.el is to try nlinum.el from
GNU ELPA instead (it's hoped this will replace linum.el one day).
Does nlinum.el have the same issue?

bug#16545: 24.3.50; Patch to coalesce calls to linum-update-current

[Chow Loong Jin]

[-- Attachment #1: Type: text/plain, Size: 830 bytes --]

On Sat, Jan 25, 2014 at 08:18:57PM -0500, Glenn Morris wrote:
> 
> Thanks for the patch.
> The usual answer to anything involving linum.el is to try nlinum.el from
> GNU ELPA instead (it's hoped this will replace linum.el one day).
> Does nlinum.el have the same issue?

Not sure, I was only able to reproduce the issue when using linum-relative.el,
as that made the linum updates slower, and I've since upgraded my machine.

Looking at the source of nlinum.el, though, it doesn't delay the updates like
linum.el does, so updates to the line numbering is unlikely to lag. Instead, it
just makes the entire Emacs instance lag.


P.S. I hope nlinum.el in its current state does not replace linum.el -- neither
global-nlinum-mode nor nlinum-relative seem to exist for this just yet.

-- 
Kind regards,
Loong Jin

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

bug#16545: 24.3.50; Patch to coalesce calls to linum-update-current

[Stefan Monnier]

> P.S. I hope nlinum.el in its current state does not replace
> linum.el -- neither global-nlinum-mode nor nlinum-relative seem to
> exist for this just yet.

nlinum.el in the `elpa' branch does include global-nlinum-mode (it was
trivial to define).

But indeed there's no nlinum-relative.  Not sure how best to implement it.


        Stefan

bug#16545: 24.3.50; Patch to coalesce calls to linum-update-current

[Chow Loong Jin]

[-- Attachment #1: Type: text/plain, Size: 474 bytes --]

On Mon, Jan 27, 2014 at 09:29:57AM -0500, Stefan Monnier wrote:
> > P.S. I hope nlinum.el in its current state does not replace
> > linum.el -- neither global-nlinum-mode nor nlinum-relative seem to
> > exist for this just yet.
> 
> nlinum.el in the `elpa' branch does include global-nlinum-mode (it was
> trivial to define).

Odd, I didn't see it yesterday when I installed it. Looks like it's there after
I reinstalled it today.

-- 
Kind regards,
Loong Jin

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

bug#16545: 24.3.50; Patch to coalesce calls to linum-update-current

[Lars Ingebrigtsen]

Chow Loong Jin <hyperair@debian.org> writes:

> When navigating using C-n/C-p with a high key repeat rate, causing
> linum-schedule to be called many times in quick succession,
> linum-update-current gets called as many times in linum-schedule,
> eventhough one call per round should be enough to renumber the lines.
>
> This patch adds a linum-delay-seconds option defaulting to 10ms to delay
> the call to linum-update-current, thus coalescing all calls to
> linum-update-current until Emacs settles for 10ms.

This was five years ago, and in the meantime, Emacs has gotten native
line number support, so I don't think this is applicable any more, and
I'm closing this bug report.  If I've misunderstood, please reopen.

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