bug#21737: 24.5; Minor feature request: `comint-previous-matching-input-from-input' and point

bug#21737: 24.5; Minor feature request: `comint-previous-matching-input-from-input' and point

[Eli Barzilay]

`comint-previous-matching-input-from-input' used to insert a match and
leave the point at the end of the line.  I found that behavior much
better than zsh and bash since when recalling old commands I edit the
last part of the command much more often.  (In fact, I think that
comint used to leave the point where it was years ago, since I
remember having to tweak emacs to put it at the end.)

A few years ago, there was a bug & a patch reported by Le Wang about
keeping the point at the same place -- something that I find really
annoying...  The problem with this patch is that it has no option to
leave the point at the end, forcing me to duplicate the whole function
if I want that behavior.

So, this is a feature request to add such an option with a new custom
variable.

-- 
                    ((x=>x(x))(x=>x(x)))                   Eli Barzilay:
                    http://barzilay.org/                   Maze is Life!

bug#21737: Acknowledgement (24.5; Minor feature request: `comint-previous-matching-input-from-input' and point)

[Eli Barzilay]

On Thu, Oct 22, 2015 at 4:30 PM, GNU bug Tracking System
<help-debbugs@gnu.org> wrote:
> If you wish to submit further information on this problem, please
> send it to 21737@debbugs.gnu.org.

I forgot to also note that this bit from the mentioned bug report:

| The point should not be moved so that I can press M-p again to find
| the next match.

is wrong: `comint-previous-matching-input-from-input' already had that
functionality using `comint-matching-input-from-input-string' to keep
the same search string on consequent searches regardless of the point
position.

(I tend to think that because of this the default behavior should be
kept as it was, but in practice I'm guessing that most people expect the
zsh/bash thing.)

-- 
                    ((x=>x(x))(x=>x(x)))                   Eli Barzilay:
                    http://barzilay.org/                   Maze is Life!

bug#21737: 24.5; Minor feature request: `comint-previous-matching-input-from-input' and point

[Eli Zaretskii]

> From: Eli Barzilay <eli@barzilay.org>
> Date: Thu, 22 Oct 2015 16:29:15 -0400
> Cc: Chong Yidong <cyd@gnu.org>, Le Wang <l26wang@gmail.com>
> 
> `comint-previous-matching-input-from-input' used to insert a match and
> leave the point at the end of the line.  I found that behavior much
> better than zsh and bash since when recalling old commands I edit the
> last part of the command much more often.  (In fact, I think that
> comint used to leave the point where it was years ago, since I
> remember having to tweak emacs to put it at the end.)
> 
> A few years ago, there was a bug & a patch reported by Le Wang about
> keeping the point at the same place -- something that I find really
> annoying...  The problem with this patch is that it has no option to
> leave the point at the end, forcing me to duplicate the whole function
> if I want that behavior.
> 
> So, this is a feature request to add such an option with a new custom
> variable.

Thanks.  Would you like to prepare a patch along these lines?

bug#21737: Acknowledgement (24.5; Minor feature request: `comint-previous-matching-input-from-input' and point)

[Stephen Perry]

How about this for a patch?  It’s my first so be kind :-)

diff --git a/lisp/comint.el b/lisp/comint.el
index e70fe88..8d61226 100644
--- a/lisp/comint.el
+++ b/lisp/comint.el
@@ -283,6 +283,14 @@ comint-move-point-for-output
 		 (const others))
   :group 'comint)
 
+(defcustom comint-move-point-to-end-on-match nil
+  "Controls whether point moves to the end of the matched input.
+If nil, then point remains in the current position.
+If t, point moves to the end of matched input."
+  :type '(choice (const :tag "off" nil)
+                 (const t))
+  :group 'comint)
+
 (defvaralias 'comint-scroll-to-bottom-on-output 'comint-move-point-for-output)
 
 (defcustom comint-scroll-show-maximum-output t
@@ -1220,7 +1228,8 @@ comint-previous-matching-input-from-input
     (comint-previous-matching-input
      (concat "^" (regexp-quote comint-matching-input-from-input-string))
      n)
-    (goto-char opoint)))
+    (unless comint-move-point-to-end-on-match
+      (goto-char opoint))))
 
 (defun comint-next-matching-input-from-input (n)
   "Search forwards through input history for match for current input.

bug#21737: Acknowledgement (24.5; Minor feature request: `comint-previous-matching-input-from-input' and point)

[John Wiegley]

>>>>> Stephen Perry <stevoooo@gmail.com> writes:

> How about this for a patch?  It’s my first so be kind :-)

Hi Stephen,

I think that using ":type 'boolean" is better than specifiying a custom
version.

Otherwise, it looks good to me.

John

bug#21737: Acknowledgement (24.5; Minor feature request: `comint-previous-matching-input-from-input' and point)

[Stephen Perry]

Updated patch:

diff --git a/lisp/comint.el b/lisp/comint.el
index e70fe88..b7a906e 100644
--- a/lisp/comint.el
+++ b/lisp/comint.el
@@ -283,6 +283,13 @@ comint-move-point-for-output
 		 (const others))
   :group 'comint)
 
+(defcustom comint-move-point-to-end-on-match nil
+  "Controls whether point moves to the end of the matched input.
+If nil, then point remains in the current position.
+If t, point moves to the end of matched input."
+  :type 'boolean
+  :group 'comint)
+
 (defvaralias 'comint-scroll-to-bottom-on-output 'comint-move-point-for-output)
 
 (defcustom comint-scroll-show-maximum-output t
@@ -1220,7 +1227,8 @@ comint-previous-matching-input-from-input
     (comint-previous-matching-input
      (concat "^" (regexp-quote comint-matching-input-from-input-string))
      n)
-    (goto-char opoint)))
+    (unless comint-move-point-to-end-on-match
+      (goto-char opoint))))
 
 (defun comint-next-matching-input-from-input (n)
   "Search forwards through input history for match for current input.

bug#21737: Acknowledgement (24.5; Minor feature request: `comint-previous-matching-input-from-input' and point)

[Juri Linkov]

> +(defcustom comint-move-point-to-end-on-match nil
> +  "Controls whether point moves to the end of the matched input.
> +If nil, then point remains in the current position.
> +If t, point moves to the end of matched input."
> +  :type 'boolean
> +  :group 'comint)
> +
>  (defvaralias 'comint-scroll-to-bottom-on-output 'comint-move-point-for-output)
>
>  (defcustom comint-scroll-show-maximum-output t
> @@ -1220,7 +1227,8 @@ comint-previous-matching-input-from-input
>      (comint-previous-matching-input
>       (concat "^" (regexp-quote comint-matching-input-from-input-string))
>       n)
> -    (goto-char opoint)))
> +    (unless comint-move-point-to-end-on-match
> +      (goto-char opoint))))

I wonder could we use the logic of 'track-eol' here?  This means that going
to the previous command lines starting at the end of the command line will
keep to the ends of command lines.

bug#21737: Acknowledgement (24.5; Minor feature request: `comint-previous-matching-input-from-input' and point)

[Stephen Perry]

> On 17 Nov 2015, at 00:58, Juri Linkov <juri@linkov.net> wrote:
> 
>> +(defcustom comint-move-point-to-end-on-match nil
>> +  "Controls whether point moves to the end of the matched input.
>> +If nil, then point remains in the current position.
>> +If t, point moves to the end of matched input."
>> +  :type 'boolean
>> +  :group 'comint)
>> +
>> (defvaralias 'comint-scroll-to-bottom-on-output 'comint-move-point-for-output)
>> 
>> (defcustom comint-scroll-show-maximum-output t
>> @@ -1220,7 +1227,8 @@ comint-previous-matching-input-from-input
>>     (comint-previous-matching-input
>>      (concat "^" (regexp-quote comint-matching-input-from-input-string))
>>      n)
>> -    (goto-char opoint)))
>> +    (unless comint-move-point-to-end-on-match
>> +      (goto-char opoint))))
> 
> I wonder could we use the logic of 'track-eol' here?  This means that going
> to the previous command lines starting at the end of the command line will
> keep to the ends of command lines.

It’s a trivial change to make, but I’m not sure if having it do
the same thing will be annoying for people who don’t want
track-eol turned on everywhere?

bug#21737: Acknowledgement (24.5; Minor feature request: `comint-previous-matching-input-from-input' and point)

[Juri Linkov]

>> I wonder could we use the logic of 'track-eol' here?  This means that going
>> to the previous command lines starting at the end of the command line will
>> keep to the ends of command lines.
>
> It’s a trivial change to make, but I’m not sure if having it do
> the same thing will be annoying for people who don’t want
> track-eol turned on everywhere?

We have it working this way in navigating the minibuffer's history,
so users are already accustomed to this behavior.

bug#21737: Acknowledgement (24.5; Minor feature request: `comint-previous-matching-input-from-input' and point)

[Lars Ingebrigtsen]

Stephen Perry <stevoooo@gmail.com> writes:

> +(defcustom comint-move-point-to-end-on-match nil
> +  "Controls whether point moves to the end of the matched input.
> +If nil, then point remains in the current position.
> +If t, point moves to the end of matched input."
> +  :type 'boolean
> +  :group 'comint)

It seems like this was fixed by a somewhat similar patch the year
after, so I'm closing this bug report:


commit afe2997119a5e6b108851f2e236037da1d4acef2
Author:     Philipp Stephani <phst@google.com>
AuthorDate: Thu Oct 27 00:28:34 2016 +0200

    Comint: new user option for C-c M-r and C-c M-s
    
    This option allows the user to specify where to place point after these
    commands.
    
    * comint.el (comint-move-point-for-matching-input): New user option.
    (comint-previous-matching-input-from-input): Use user option.


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