bug#25665: 26.0.50; [PATCH] Indicate prefix arg in minibuffer prompt for shell-command

bug#25665: 26.0.50; [PATCH] Indicate prefix arg in minibuffer prompt for shell-command

[Chunyang Xu]

[-- Attachment #1: Type: text/plain, Size: 331 bytes --]

Hi,

When using 'C-u M-!' ('shell-command') I feel confident if the
minibuffer prompt becomes

  Shell command and insert output:

since it explains the purpose of the prefix argument. The same thing
goes for 'M-&' ('async-shell-command') and 'M-|'
('shell-command-on-region'). I have attached a patch which implements
this idea.


[-- Attachment #2: 0001-Indicate-prefix-arg-in-minibuffer-prompt-for-shell-c.patch --]
[-- Type: text/plain, Size: 3424 bytes --]

From e395aa9e601bb8071b53704b7aa27b8b28ac22b1 Mon Sep 17 00:00:00 2001
From: Chunyang Xu <mail@xuchunyang.me>
Date: Thu, 9 Feb 2017 22:47:40 +0800
Subject: [PATCH] Indicate prefix arg in minibuffer prompt for shell-command

* lisp/simple.el (async-shell-command, shell-command)
(shell-command-on-region): Show purpose of prefix argument in
the minibuffer prompt.

Copyright-paperwork-exempt: yes
---
 lisp/simple.el | 42 +++++++++++++++++++++++++++---------------
 1 file changed, 27 insertions(+), 15 deletions(-)

diff --git a/lisp/simple.el b/lisp/simple.el
index 441713a18b..9720a8935c 100644
--- a/lisp/simple.el
+++ b/lisp/simple.el
@@ -3311,13 +3311,17 @@ async-shell-command
 shell (with its need to quote arguments)."
   (interactive
    (list
-    (read-shell-command "Async shell command: " nil nil
-			(let ((filename
-			       (cond
-				(buffer-file-name)
-				((eq major-mode 'dired-mode)
-				 (dired-get-filename nil t)))))
-			  (and filename (file-relative-name filename))))
+    (read-shell-command (format "Async shell command%s: "
+                                (if current-prefix-arg
+                                    " and insert output"
+                                  ""))
+                        nil nil
+                        (let ((filename
+                               (cond
+                                (buffer-file-name)
+                                ((eq major-mode 'dired-mode)
+                                 (dired-get-filename nil t)))))
+                          (and filename (file-relative-name filename))))
     current-prefix-arg
     shell-command-default-error-buffer))
   (unless (string-match "&[ \t]*\\'" command)
@@ -3382,13 +3386,17 @@ shell-command
 
   (interactive
    (list
-    (read-shell-command "Shell command: " nil nil
-			(let ((filename
-			       (cond
-				(buffer-file-name)
-				((eq major-mode 'dired-mode)
-				 (dired-get-filename nil t)))))
-			  (and filename (file-relative-name filename))))
+    (read-shell-command (format "Shell command%s: "
+                                (if current-prefix-arg
+                                    " and insert output"
+                                  ""))
+                        nil nil
+                        (let ((filename
+                               (cond
+                                (buffer-file-name)
+                                ((eq major-mode 'dired-mode)
+                                 (dired-get-filename nil t)))))
+                          (and filename (file-relative-name filename))))
     current-prefix-arg
     shell-command-default-error-buffer))
   ;; Look for a handler in case default-directory is a remote file name.
@@ -3640,7 +3648,11 @@ shell-command-on-region
 		 ;; Do this before calling region-beginning
 		 ;; and region-end, in case subprocess output
 		 ;; relocates them while we are in the minibuffer.
-		 (setq string (read-shell-command "Shell command on region: "))
+		 (setq string (read-shell-command
+                               (format "Shell command on region%s: "
+                                       (if current-prefix-arg
+                                           " and replace with output"
+                                         ""))))
 		 ;; call-interactively recognizes region-beginning and
 		 ;; region-end specially, leaving them in the history.
 		 (list (region-beginning) (region-end)
-- 
2.11.0

bug#25665: 26.0.50; [PATCH] Indicate prefix arg in minibuffer prompt for shell-command

[Tino Calancha]

Chunyang Xu <mail@xuchunyang.me> writes:

Thanks for the report.
> Hi,
>
> When using 'C-u M-!' ('shell-command') I feel confident if the
> minibuffer prompt becomes
>
>   Shell command and insert output:
>
> since it explains the purpose of the prefix argument. The same thing
> goes for 'M-&' ('async-shell-command') and 'M-|'
> ('shell-command-on-region').
I am not sure about this.
The meaning of the prefix argument must be explained in the docstring.
That and the manual are the source of documentation, not the prompt.

Having a short prompt is also good because there are more space
in the line for the command; a short prompt is also better while tipying
a long command: sometimes you want to change something at the beginning:
< ; `beginning-of-buffer'
Now you need to `C-f' several times to jump out of the prompt.  The more
verbose prompt, the more `C-f' you need.

Just my opinion.  Let's see what other people think.
Thanks.

bug#25665: 26.0.50; [PATCH] Indicate prefix arg in minibuffer prompt for shell-command

[Drew Adams]

> I am not sure about this.
> The meaning of the prefix argument must be explained in the docstring.
> That and the manual are the source of documentation, not the prompt.
> 
> Having a short prompt is also good because there are more space
> in the line for the command; a short prompt is also better while tipying
> a long command: sometimes you want to change something at the beginning:
> < ; `beginning-of-buffer'
> Now you need to `C-f' several times to jump out of the prompt.  The more
> verbose prompt, the more `C-f' you need.
> 
> Just my opinion.  Let's see what other people think.

+1

The minibuffer should simply echo that the prefix arg has been
used, and just what its numeric value is.  It should not try
to explain what the possible values are and what they mean.

bug#25665: 26.0.50; [PATCH] Indicate prefix arg in minibuffer prompt for shell-command

[Eli Zaretskii]

> From: Tino Calancha <tino.calancha@gmail.com>
> Date: Fri, 10 Feb 2017 01:53:56 +0900
> Cc: 25665@debbugs.gnu.org, tino.calancha@gmail.com
> 
> I am not sure about this.
> The meaning of the prefix argument must be explained in the docstring.
> That and the manual are the source of documentation, not the prompt.
> 
> Having a short prompt is also good because there are more space
> in the line for the command; a short prompt is also better while tipying
> a long command: sometimes you want to change something at the beginning:
> < ; `beginning-of-buffer'
> Now you need to `C-f' several times to jump out of the prompt.  The more
> verbose prompt, the more `C-f' you need.
> 
> Just my opinion.  Let's see what other people think.

I think it goes against Emacs conventions.  We have quite a few
commands that insert output into the current buffer when invoked with
an argument.

bug#25665: 26.0.50; [PATCH] Indicate prefix arg in minibuffer prompt for shell-command

[Chunyang Xu]

Eli Zaretskii writes:

>> From: Tino Calancha <tino.calancha@gmail.com>
>> Date: Fri, 10 Feb 2017 01:53:56 +0900
>> Cc: 25665@debbugs.gnu.org, tino.calancha@gmail.com
>>
>> I am not sure about this.
>> The meaning of the prefix argument must be explained in the docstring.
>> That and the manual are the source of documentation, not the prompt.
>>
>> Having a short prompt is also good because there are more space
>> in the line for the command; a short prompt is also better while tipying
>> a long command: sometimes you want to change something at the beginning:
>> < ; `beginning-of-buffer'
>> Now you need to `C-f' several times to jump out of the prompt.  The more
>> verbose prompt, the more `C-f' you need.
>>
>> Just my opinion.  Let's see what other people think.
>
> I think it goes against Emacs conventions.  We have quite a few
> commands that insert output into the current buffer when invoked with
> an argument.

but many of these commands don't have a minibuffer prompt, such as
'C-x C-e', so we don't need to worry about them.

I just notice that 'C-u C-s' ('isearch-forward') is also showing the
meaning of prefix argument in the minibuffer prompt:

C-s       I-search:
C-u C-s   Regexp I-search:

(of course you can think it is trying to use the same prompt as 'C-M-s')

bug#25665: 26.0.50; [PATCH] Indicate prefix arg in minibuffer prompt for shell-command

[npostavs]

On Thu, Feb 9, 2017 at 11:53 AM, Tino Calancha <tino.calancha@gmail.com> wrote:

> The meaning of the prefix argument must be explained in the docstring.
> That and the manual are the source of documentation, not the prompt.

But it's nice if Emacs gives you a different prompt, so that you can be
sure you typed in the command correctly.  Just like M-& gives a
different prompt from M-!, even though you could figure out the
different meaning just by reading the docstring of each command.

>
> Having a short prompt is also good because there are more space
> in the line for the command;

Yes, it's better to keep the prompt short if possible.  Still, having
the extra feedback might be worth it.

> a short prompt is also better while tipying
> a long command: sometimes you want to change something at the beginning:
> < ; `beginning-of-buffer'

If the command is a single line, you can use C-a instead to avoid the prompt.

> Now you need to `C-f' several times to jump out of the prompt.  The more
> verbose prompt, the more `C-f' you need.

M-e leaves the prompt immediately.

bug#25665: 26.0.50; [PATCH] Indicate prefix arg in minibuffer prompt for shell-command

[Tino Calancha]

On Thu, 9 Feb 2017, npostavs@users.sourceforge.net wrote:

> On Thu, Feb 9, 2017 at 11:53 AM, Tino Calancha <tino.calancha@gmail.com> wrote:
>
>> The meaning of the prefix argument must be explained in the docstring.
>> That and the manual are the source of documentation, not the prompt.
>
> But it's nice if Emacs gives you a different prompt, so that you can be
> sure you typed in the command correctly.
Confirmation for having typed C-u?
Well, sometimes some keys in old keyboards die... ;-)

> Just like M-& gives a
> different prompt from M-!, even though you could figure out the
> different meaning just by reading the docstring of each command.
These are different commands so its Emacs (and whatever editor)
convention to how a different prompt.  The OP suggests to use
a different prompt for the same command w/ or w/o prefix.

There are lots of commands in Emacs accepting a raw,
numeric prefix etc.  If we change M-! prompt, we might
also want to tune the prompt for all of them.  Do we want that?

bug#25665: 26.0.50; [PATCH] Indicate prefix arg in minibuffer prompt for shell-command

[npostavs]

Tino Calancha <tino.calancha@gmail.com> writes:

> On Thu, 9 Feb 2017, npostavs@users.sourceforge.net wrote:
>
>> On Thu, Feb 9, 2017 at 11:53 AM, Tino Calancha <tino.calancha@gmail.com> wrote:
>>
>>> The meaning of the prefix argument must be explained in the docstring.
>>> That and the manual are the source of documentation, not the prompt.
>>
>> But it's nice if Emacs gives you a different prompt, so that you can be
>> sure you typed in the command correctly.
> Confirmation for having typed C-u?
> Well, sometimes some keys in old keyboards die... ;-)

We already have that in `echo-keystrokes'.

But I'm thinking also of times when you do C-u M-! and then switch to
another window to copy the command, then maybe when you come back, you
don't remember 100% what you were doing.

>> Just like M-& gives a
>> different prompt from M-!, even though you could figure out the
>> different meaning just by reading the docstring of each command.
> These are different commands so its Emacs (and whatever editor)
> convention to how a different prompt.  The OP suggests to use
> a different prompt for the same command w/ or w/o prefix.

M-& and M-! are practically the same command anyway, async-shell-command
is basically just

    (defun async-shell-command (command)
      (shell-command (concat command "&")))

>
> There are lots of commands in Emacs accepting a raw,
> numeric prefix etc.  If we change M-! prompt, we might
> also want to tune the prompt for all of them.  Do we want that?

IMO yes, if the difference can be summarized in the prompt concisely
enough.

bug#25665: 26.0.50; [PATCH] Indicate prefix arg in minibuffer prompt for shell-command

[Tino Calancha]

On Thu, 9 Feb 2017, npostavs@users.sourceforge.net wrote:

> Tino Calancha <tino.calancha@gmail.com> writes:
>
>> On Thu, 9 Feb 2017, npostavs@users.sourceforge.net wrote:
>>
>>> On Thu, Feb 9, 2017 at 11:53 AM, Tino Calancha <tino.calancha@gmail.com> wrote:
>>>
>>>> The meaning of the prefix argument must be explained in the docstring.
>>>> That and the manual are the source of documentation, not the prompt.
>>>
>>> But it's nice if Emacs gives you a different prompt, so that you can be
>>> sure you typed in the command correctly.
>> Confirmation for having typed C-u?
>> Well, sometimes some keys in old keyboards die... ;-)
>
> We already have that in `echo-keystrokes'.
>
> But I'm thinking also of times when you do C-u M-! and then switch to
> another window to copy the command, then maybe when you come back, you
> don't remember 100% what you were doing.
I see.  That's a point.  I agree in that edge case would be useful.

>> There are lots of commands in Emacs accepting a raw,
>> numeric prefix etc.  If we change M-! prompt, we might
>> also want to tune the prompt for all of them.  Do we want that?
>
> IMO yes, if the difference can be summarized in the prompt concisely
> enough.
It might be a pain in the ass to follow such new policy:
emacs -Q:
   C-x b foo RET
   aeiou
I)
     C-x C-s
II)
     C-u C-x C-s
III)
     C-u C-u C-x C-s
IV)
     C-u C-u C-u C-x C-s

All have the same prompt:
File to save in: ~/

bug#25665: 26.0.50; [PATCH] Indicate prefix arg in minibuffer prompt for shell-command

[Noam Postavsky]

On Fri, Feb 10, 2017 at 2:13 AM, Tino Calancha <tino.calancha@gmail.com> wrote:
>>> There are lots of commands in Emacs accepting a raw,
>>> numeric prefix etc.  If we change M-! prompt, we might
>>> also want to tune the prompt for all of them.  Do we want that?
>>
>>
>> IMO yes, if the difference can be summarized in the prompt concisely
>> enough.
>
> It might be a pain in the ass to follow such new policy:
> emacs -Q:
>   C-x b foo RET
>   aeiou
> I)
>     C-x C-s
> II)
>     C-u C-x C-s
> III)
>     C-u C-u C-x C-s
> IV)
>     C-u C-u C-u C-x C-s
>
> All have the same prompt:
> File to save in: ~/

I think this would fall under the case of too difficult to summarize
into a prompt concisely. Furthermore, C-x C-s does not usually prompt
at all so there is not much benefit anyway. But I would not object if
someone could come up with reasonably short and informative
modifications to the prompts.

bug#25665: 26.0.50; [PATCH] Indicate prefix arg in minibuffer prompt for shell-command

[Lars Ingebrigtsen]

Tino Calancha <tino.calancha@gmail.com> writes:

>> When using 'C-u M-!' ('shell-command') I feel confident if the
>> minibuffer prompt becomes
>>
>>   Shell command and insert output:
>>
>> since it explains the purpose of the prefix argument. The same thing
>> goes for 'M-&' ('async-shell-command') and 'M-|'
>> ('shell-command-on-region').
> I am not sure about this.
> The meaning of the prefix argument must be explained in the docstring.
> That and the manual are the source of documentation, not the prompt.

There didn't seem to be much enthusiasm for the proposed change, so I'm
closing this bug report.

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

bug#25665: 26.0.50; [PATCH] Indicate prefix arg in minibuffer prompt for shell-command

[Drew Adams]

> > I am not sure about this.
> > The meaning of the prefix argument must be explained in the docstring.
> > That and the manual are the source of documentation, not the prompt.
> >
> > Having a short prompt is also good because there are more space
> > in the line for the command; a short prompt is also better while tipying
> > a long command: sometimes you want to change something at the beginning:
> > < ; `beginning-of-buffer'
> > Now you need to `C-f' several times to jump out of the prompt.  The more
> > verbose prompt, the more `C-f' you need.
> >
> > Just my opinion.  Let's see what other people think.
> 
> I think it goes against Emacs conventions.  We have quite a few
> commands that insert output into the current buffer when invoked with
> an argument.

(Sorry, I misunderstood the suggestion.  I didn't look at the
patch, and didn't get that this was about changing a _prompt_.)

I don't think the suggestion "goes against Emacs conventions".

Certainly it is the case that we do not do this for most
commands that do something different when invoked with a
prefix arg.

But there is a difference between a command that prompts
and one, such as `eval-last-sexp', that does not.  Most
"commands that insert output" when you use a prefix arg
do not prompt, I think. 

And I know of no convention that says that a command should
not use a different prompt when a prefix arg is used.

It's a judgment call.  For this one I have no opinion.

I don't think, however, that the length of the prompt is
very important here.  That's a fairly weak criterion for
deciding whether to change the prompt.

Doing something like what was suggested could be helpful
mainly in a situation where (1) the behavior is very
different if a prefix arg is used and (2) it's not easy
to undo a mistaken use of the command.

If that's not the case here then I don't see a need for it.