bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Kaushal Modi]

[-- Attachment #1: Type: text/plain, Size: 931 bytes --]

Tested in the latest emacs-25 build as of today.

emacs -Q
C-u C-s

The echo area will show "I-search:".
Earlier it used to show "Regexp I-search:" and that made it clear that we
are doing regexp search.

This change happened after this revert:
http://git.savannah.gnu.org/cgit/emacs.git/commit/?h=emacs-25&id=b417c5a3b35bf31e66a170529c5aeb34cb3318d0

So it looks like that isearch-message-prefix was bound to
search-default-mode being non-nil.

I verified that "C-u C-s" shows "Regexp I-search:" fine in emacs 24.5. But
that version did not have the search-default-mode.

I believe that the clarity of regexp isearch being on/off is missing,
especially when C-u C-s DOES do regexp search and still displays "I-search:
" in the echo.

If I set search-default-mode back to "#'character-fold-to-regexp", this
issue is gone. So displaying of "Regexp I-search:" should be decoupled from
the character-fold.el library.

--
Kaushal Modi

[-- Attachment #2: Type: text/html, Size: 1436 bytes --]

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Kaushal Modi]

[-- Attachment #1: Type: text/plain, Size: 1300 bytes --]

Hi all,

The below patch fixes this bug for me.

I simply promoted the priority of isearch-regexp in that cond.

I hope I am not overlooking something else by doing so.

With that, now C-u C-s shows "Regexp I-search:" when the search of
search-default-mode is either nil or character-fold-to-regexp.


diff --git a/lisp/isearch.el b/lisp/isearch.el
index b8ada2c..47dedfa 100644
--- a/lisp/isearch.el
+++ b/lisp/isearch.el
@@ -2575,12 +2575,12 @@ isearch--describe-regexp-mode
     (setq regexp-function #'word-search-regexp))
   (let ((description
          ;; Don't use a description on the default search mode.
-         (cond ((equal regexp-function search-default-mode) "")
+         (cond (isearch-regexp "regexp ")
+               ((equal regexp-function search-default-mode) "")
                (regexp-function
                 (and (symbolp regexp-function)
                      (or (get regexp-function 'isearch-message-prefix)
                          "")))
-               (isearch-regexp "regexp ")
                ;; We're in literal mode. If the default mode is not
                ;; literal, then describe it.
                ((functionp search-default-mode) "literal "))))


Apologies for earlier misunderstanding that this bug had to do with
character-fold.el.

--
Kaushal Modi

[-- Attachment #2: Type: text/html, Size: 1942 bytes --]

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Eli Zaretskii]

> From: Kaushal Modi <kaushal.modi@gmail.com>
> Date: Fri, 11 Mar 2016 15:46:29 -0500
> 
> The below patch fixes this bug for me.
> 
> I simply promoted the priority of isearch-regexp in that cond.
> 
> I hope I am not overlooking something else by doing so.
> 
> With that, now C-u C-s shows "Regexp I-search:" when the search of
> search-default-mode is either nil or character-fold-to-regexp.
> 
> 
> diff --git a/lisp/isearch.el b/lisp/isearch.el
> index b8ada2c..47dedfa 100644
> --- a/lisp/isearch.el
> +++ b/lisp/isearch.el
> @@ -2575,12 +2575,12 @@ isearch--describe-regexp-mode
>      (setq regexp-function #'word-search-regexp))
>    (let ((description
>           ;; Don't use a description on the default search mode.
> -         (cond ((equal regexp-function search-default-mode) "")
> +         (cond (isearch-regexp "regexp ")
> +               ((equal regexp-function search-default-mode) "")
>                 (regexp-function
>                  (and (symbolp regexp-function)
>                       (or (get regexp-function 'isearch-message-prefix)
>                           "")))
> -               (isearch-regexp "regexp ")
>                 ;; We're in literal mode. If the default mode is not
>                 ;; literal, then describe it.
>                 ((functionp search-default-mode) "literal "))))

What if search-default-mode is t, i.e. the default is regexp search?

Thanks.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Kaushal Modi]

[-- Attachment #1: Type: text/plain, Size: 341 bytes --]

On Fri, Mar 11, 2016 at 3:54 PM, Eli Zaretskii <eliz@gnu.org> wrote:

> What if search-default-mode is t, i.e. the default is regexp search?


It works fine in that case too (with this patch).

When search-default-mode is set to t, doing just C-s will now begin regexp
isearch showing "Regexp I-search:" in the minibuffer.


--
Kaushal Modi

[-- Attachment #2: Type: text/html, Size: 791 bytes --]

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Eli Zaretskii]

> From: Kaushal Modi <kaushal.modi@gmail.com>
> Date: Fri, 11 Mar 2016 15:58:20 -0500
> Cc: 22991@debbugs.gnu.org, Artur Malabarba <arturmalabarba@gmail.com>
> 
>  What if search-default-mode is t, i.e. the default is regexp search?
> 
> It works fine in that case too (with this patch).
> 
> When search-default-mode is set to t, doing just C-s will now begin regexp isearch showing "Regexp
> I-search:" in the minibuffer.

Which I think is wrong, since the default mode shouldn't show anything
except "I-search: ".  Right?

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Kaushal Modi]

[-- Attachment #1: Type: text/plain, Size: 2329 bytes --]

> Which I think is wrong, since the default mode shouldn't show anything
> except "I-search: ".  Right?

Even before this patch, that behavior (Displaying "Regexp I-search:" on C-s
even when search-default-mode is t) was that way. Because, even when
search-default-mode is t, regexp-function still stays nil. So that first
cond form condition evals to nil.

But I discovered a major issue with my patch. So here's a new one which is
less intrusive. I also added explanation for the current priority order in
the cond.

diff --git a/lisp/isearch.el b/lisp/isearch.el
index b8ada2c..7c3d4e3 100644
--- a/lisp/isearch.el
+++ b/lisp/isearch.el
@@ -2574,16 +2574,23 @@ isearch--describe-regexp-mode
   (when (eq regexp-function t)
     (setq regexp-function #'word-search-regexp))
   (let ((description
-         ;; Don't use a description on the default search mode.
-         (cond ((equal regexp-function search-default-mode) "")
-               (regexp-function
-                (and (symbolp regexp-function)
-                     (or (get regexp-function 'isearch-message-prefix)
-                         "")))
-               (isearch-regexp "regexp ")
-               ;; We're in literal mode. If the default mode is not
-               ;; literal, then describe it.
-               ((functionp search-default-mode) "literal "))))
+         (cond
+          ;; 1. Don't use a description on the default search mode,
+          ;;    only if the default search mode is non-nil.
+          ((and search-default-mode
+                (equal regexp-function search-default-mode)) "")
+          ;; 2. Use the `isearch-message-prefix' set for
+          ;;    `regexp-function' if available.
+          (regexp-function
+           (and (symbolp regexp-function)
+                (or (get regexp-function 'isearch-message-prefix)
+                    "")))
+          ;; 3. Else if `isearch-regexp' is non-nil, set description
+          ;;    to "regexp ".
+          (isearch-regexp "regexp ")
+          ;; 4. And finally, if we're in literal mode (AND if default
+          ;; mode is not literal too), then describe it.
+          ((functionp search-default-mode) "literal "))))
     (if space-before
         ;; Move space from the end to the beginning.
         (replace-regexp-in-string "\\(.*\\) \\'" " \\1" description)

[-- Attachment #2: Type: text/html, Size: 3209 bytes --]

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Artur Malabarba]

[-- Attachment #1: Type: text/plain, Size: 655 bytes --]

On 11 Mar 2016 6:23 pm, "Kaushal Modi" <kaushal.modi@gmail.com> wrote:
>
> > Which I think is wrong, since the default mode shouldn't show anything
> > except "I-search: ".  Right?

> Even before this patch, that behavior (Displaying "Regexp I-search:" on
C-s even when search-default-mode is t) was that way. Because, even when
search-default-mode is t, regexp-function still stays nil. So that first
cond form condition evals to nil.
>
> But I discovered a major issue with my patch. So here's a new one which
is less intrusive. I also added explanation for the current priority order
in the cond.

This last patch looks fine to me. 👍

[-- Attachment #2: Type: text/html, Size: 830 bytes --]

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Kaushal Modi]

[-- Attachment #1: Type: text/plain, Size: 3009 bytes --]

On Fri, Mar 11, 2016 at 5:22 PM, Artur Malabarba <bruce.connor.am@gmail.com>
wrote:

> This last patch looks fine to me. 👍
>

Thanks.

Here's one more shot with fix for that bug that was already there that Eli
mentioned.
Now if the user has set search-default-mode to t and if they do C-s, the
description is still "" (not "regexp "). So the minibuffer would read
"I-search: ".

Below is the patch created using git format-patch with appropriate commit
log.

From df1a6e8e9d8227f5c137236a09fda9818c479d0c Mon Sep 17 00:00:00 2001
From: Kaushal Modi <kaushal.modi@gmail.com>
Date: Fri, 11 Mar 2016 17:34:50 -0500
Subject: [PATCH] Fix description for regexp searches (bug 22991)

* lisp/isearch.el (isearch--describe-regexp-mode): With
  `search-default-mode' set to nil, if user does C-u C-s, the minibuffer
  now displays "Regexp I-search: ".  But if the user has set
  `search-default-mode' to t, and then does C-s, the minibuffer now
  displays "I-search: " because the default search mode is now regexp
  mode.  Comments have been added to explain the priority of conditions
  in the `cond' form.
---
 lisp/isearch.el | 19 +++++++++++++++----
 1 file changed, 15 insertions(+), 4 deletions(-)

diff --git a/lisp/isearch.el b/lisp/isearch.el
index b8ada2c..646a906 100644
--- a/lisp/isearch.el
+++ b/lisp/isearch.el
@@ -2574,15 +2574,26 @@ isearch--describe-regexp-mode
   (when (eq regexp-function t)
     (setq regexp-function #'word-search-regexp))
   (let ((description
-         ;; Don't use a description on the default search mode.
-         (cond ((equal regexp-function search-default-mode) "")
+         (cond
+          ;; 1. Do not use a description on the default search mode,
+          ;;    but only if the default search mode is non-nil.
+          ((or (and search-default-mode
+                    (equal search-default-mode regexp-function))
+               ;; Special case where `search-default-mode' is t
+               ;; (defaults to regexp searches).
+               (and (eq search-default-mode t)
+                    (eq search-default-mode isearch-regexp))) "")
+          ;; 2. Use the `isearch-message-prefix' set for
+          ;;    `regexp-function' if available.
           (regexp-function
            (and (symbolp regexp-function)
                 (or (get regexp-function 'isearch-message-prefix)
                     "")))
+          ;; 3. Else if `isearch-regexp' is non-nil, set description
+          ;;    to "regexp ".
           (isearch-regexp "regexp ")
-               ;; We're in literal mode. If the default mode is not
-               ;; literal, then describe it.
+          ;; 4. And finally, if we're in literal mode (and if the
+          ;;    default mode is also not literal), describe it.
           ((functionp search-default-mode) "literal "))))
     (if space-before
         ;; Move space from the end to the beginning.
--
2.6.0.rc0.24.gec371ff



--
Kaushal Modi

[-- Attachment #2: Type: text/html, Size: 5652 bytes --]

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Michael Heerdegen]

Eli Zaretskii <eliz@gnu.org> writes:

> Which I think is wrong, since the default mode shouldn't show anything
> except "I-search: ".  Right?

I'm not so sure.  I use different defaults in different modes for
convenience (symbol search in programming modes, word search in Info,
literal search in the rest), and I would soon get confused if the
default mode would not be indicated, especially if I change to a
different mode and back again etc.


Michael.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Kaushal Modi]

[-- Attachment #1: Type: text/plain, Size: 757 bytes --]

In that case, we will have to figure out why this condition was put as a
high priority to begin with.

(cond ((equal regexp-function search-default-mode) "")

On Friday, March 11, 2016, Michael Heerdegen <michael_heerdegen@web.de>
wrote:

> Eli Zaretskii <eliz@gnu.org <javascript:;>> writes:
>
> > Which I think is wrong, since the default mode shouldn't show anything
> > except "I-search: ".  Right?
>
> I'm not so sure.  I use different defaults in different modes for
> convenience (symbol search in programming modes, word search in Info,
> literal search in the rest), and I would soon get confused if the
> default mode would not be indicated, especially if I change to a
> different mode and back again etc.
>
>
> Michael.
>


-- 

--
Kaushal Modi

[-- Attachment #2: Type: text/html, Size: 1247 bytes --]

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Artur Malabarba]

[-- Attachment #1: Type: text/plain, Size: 395 bytes --]

On 11 Mar 2016 8:40 pm, "Kaushal Modi" <kaushal.modi@gmail.com> wrote:
>
> In that case, we will have to figure out why this condition was put as a
high priority to begin with.
>
> (cond ((equal regexp-function search-default-mode) "")
>

Because, since char-folding was turned on by default, some people said it
was odd that the stock search said "Isearch char fold" instead of just
"Isearch".

[-- Attachment #2: Type: text/html, Size: 569 bytes --]

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Kaushal Modi]

[-- Attachment #1: Type: text/plain, Size: 561 bytes --]

Sorry, I just remembered that I was one of the people (or the only one?)
who suggested not displaying "Char-fold" prefix.

But considering the example that Michael gave, it makes sense to remove
that condition from the cond form. I would actually like to see "Regexp
I-search:" even if I have set that by default.

The result will be a less complex cond statement (will get rid of that
"oring of ands" case in my last patch) and clarity as to what's the current
isearch mode.

What do you guys think? Or should we bring this up on emacs-devel?

--
Kaushal Modi

[-- Attachment #2: Type: text/html, Size: 683 bytes --]

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Eli Zaretskii]

> From: Michael Heerdegen <michael_heerdegen@web.de>
> Cc: Kaushal Modi <kaushal.modi@gmail.com>,  arturmalabarba@gmail.com,  22991@debbugs.gnu.org
> Date: Sat, 12 Mar 2016 00:37:41 +0100
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > Which I think is wrong, since the default mode shouldn't show anything
> > except "I-search: ".  Right?
> 
> I'm not so sure.  I use different defaults in different modes for
> convenience (symbol search in programming modes, word search in Info,
> literal search in the rest), and I would soon get confused if the
> default mode would not be indicated, especially if I change to a
> different mode and back again etc.

Fair enough, but the code clearly wants to produce the behavior I
hinted at.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Eli Zaretskii]

> Date: Fri, 11 Mar 2016 18:40:23 -0500
> From: Kaushal Modi <kaushal.modi@gmail.com>
> Cc: Eli Zaretskii <eliz@gnu.org>, "arturmalabarba@gmail.com" <arturmalabarba@gmail.com>, 
> 	"22991@debbugs.gnu.org" <22991@debbugs.gnu.org>
> 
> In that case, we will have to figure out why this condition was put as a high priority to begin with. 
> 
> (cond ((equal regexp-function search-default-mode) "")

Clearly, because the intent was not to produce any prefix for the
default mode.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Eli Zaretskii]

> Date: Fri, 11 Mar 2016 23:56:56 -0500
> From: Kaushal Modi <kaushal.modi@gmail.com>
> Cc: 22991@debbugs.gnu.org, Michael Heerdegen <michael_heerdegen@web.de>, 
> 	Eli Zaretskii <eliz@gnu.org>
> 
> Sorry, I just remembered that I was one of the people (or the only one?) who suggested not displaying
> "Char-fold" prefix. 
> 
> But considering the example that Michael gave, it makes sense to remove that condition from the cond form. I
> would actually like to see "Regexp I-search:" even if I have set that by default. 
> 
> The result will be a less complex cond statement (will get rid of that "oring of ands" case in my last patch) and
> clarity as to what's the current isearch mode. 
> 
> What do you guys think? Or should we bring this up on emacs-devel?

First, such a change would be a change in behavior, so it should go to
master, not emacs-25.

Second, I don't agree with Michael, so if we want to allow such a
different behavior, I will insist on a user option.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Michael Heerdegen]

Eli Zaretskii <eliz@gnu.org> writes:

> Second, I don't agree with Michael, so if we want to allow such a
> different behavior, I will insist on a user option.

I mean especially, if I end up with the default mode and the literal
search mode both show the same prompt, that would not be good.  But yes,
what I really would like would be to always show the searching mode in
the prompt.  This is not what you want, obviously.

Would it make sense to you to decide what to show in the prompt based on
the fact whether `search-default-mode' has a buffer local binding -
instead of introducing a new option (which seems a bit of overkill to
me)?  I.e., if there is a buffer local binding that differs from the
global value, do what I want - else, do what you prefer.


Michael.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Eli Zaretskii]

> From: Michael Heerdegen <michael_heerdegen@web.de>
> Cc: Kaushal Modi <kaushal.modi@gmail.com>,  22991@debbugs.gnu.org,  bruce.connor.am@gmail.com
> Date: Sat, 12 Mar 2016 13:33:09 +0100
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > Second, I don't agree with Michael, so if we want to allow such a
> > different behavior, I will insist on a user option.
> 
> I mean especially, if I end up with the default mode and the literal
> search mode both show the same prompt, that would not be good.

This could only happen if the literal search is the default.  If the
default is different, a literal search should show some non-empty
prefix.

> Would it make sense to you to decide what to show in the prompt based on
> the fact whether `search-default-mode' has a buffer local binding -
> instead of introducing a new option (which seems a bit of overkill to
> me)?  I.e., if there is a buffer local binding that differs from the
> global value, do what I want - else, do what you prefer.

I'm not sure I understand how the fact of the binding being local has
any relevance to what we show in the prompt.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Michael Heerdegen]

Eli Zaretskii <eliz@gnu.org> writes:

> I'm not sure I understand how the fact of the binding being local has
> any relevance to what we show in the prompt.

If the default is different in different buffers, the user more likely
wants to see the mode in the prompt.  So, my suggestion is to show the
default in the prompt only when it is not the default default...


Michael.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Artur Malabarba]

[-- Attachment #1: Type: text/plain, Size: 795 bytes --]

On 12 Mar 2016 9:51 am, "Michael Heerdegen" <michael_heerdegen@web.de>
wrote:
>
> Eli Zaretskii <eliz@gnu.org> writes:
>
> > I'm not sure I understand how the fact of the binding being local has
> > any relevance to what we show in the prompt.
>
> If the default is different in different buffers, the user more likely
> wants to see the mode in the prompt.  So, my suggestion is to show the
> default in the prompt only when it is not the default default...
>
>
> Michael.

I agree with Michael on this. If the current search-mode is different from
the global default, then it should be displayed on the prompt. That could
happen if the user switched the mode during a search, or if something
changed the variable's buffer-local value.

I don't want to add a user option just for this.

Cheers

[-- Attachment #2: Type: text/html, Size: 1090 bytes --]

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Kaushal Modi]

[-- Attachment #1: Type: text/plain, Size: 490 bytes --]

I also think that adding an option just for this is an overkill.

It makes sense that removing that cond right now would be a big behavior
change and should not be done in emacs-25.

Can you guys please review and push the last patch I sent? Below are the 2
fixes:
1. Now C-u C-s shows the Regexp description if the search-default-mode is
nil.
2. If regexp search is default, C-s does not show the Regexp description.

Let's continue this discussion in emacs-dev.

Thanks.

--
Kaushal Modi

[-- Attachment #2: Type: text/html, Size: 610 bytes --]

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Drew Adams]

> > Sorry, I just remembered that I was one of the people (or the only one?)
> > who suggested not displaying "Char-fold" prefix.
> >
> > But considering the example that Michael gave, it makes sense to remove
> > that condition from the cond form. I would actually like to see
> > "Regexp I-search:" even if I have set that by default.
> >
> > The result will be a less complex cond statement (will get rid of that
> > "oring of ands" case in my last patch) and clarity as to what's the
> > current isearch mode.
> >
> > What do you guys think? Or should we bring this up on emacs-devel?
> 
> First, such a change would be a change in behavior, so it should go to
> master, not emacs-25.
> 
> Second, I don't agree with Michael, so if we want to allow such a
> different behavior, I will insist on a user option.

1. I agree with Michael in this case.  The prompt should
faithfully reflect the current search state/mode - until we
find a better way to convey that information.  In any case,
the prompt should not mislead by reflecting a different
search behavior from what is current.

2. Whether this or that behavior is turned on by default
should not override that rule that the prompt tells you what
the state is.

3. To be more clear, perhaps the non-regexp, non-fold search
(including not folding whitespace) should also have a prompt
that unambiguously indicates this.  It is essentially
"literal" search.

4. All of that said, I think that we really need to find
some _other way than the prompt_ to indicate, completely,
the search state.  And then the prompt can perhaps just be
something like "Isearch" in all cases.  The prompt is not
adequate for conveying just what the current search state is.

5. This should be discussed on emacs-devel (again).

6. While waiting for a redesign in this regard and a
general decision, the prompt should reflect the default
state.  If it is super important that beginning users not
be frightened by a weird prompt by default, at the outset,
then we should turn off char-fold search by default - for now.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Eli Zaretskii]

> From: Michael Heerdegen <michael_heerdegen@web.de>
> Cc: kaushal.modi@gmail.com,  22991@debbugs.gnu.org,  bruce.connor.am@gmail.com
> Date: Sat, 12 Mar 2016 13:51:42 +0100
> 
> If the default is different in different buffers, the user more likely
> wants to see the mode in the prompt.  So, my suggestion is to show the
> default in the prompt only when it is not the default default...

Do many users define buffer-local defaults for isearch?  IOW, will
such a logic be useful enough to justify its existence?  If yes, I
don't mind.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Drew Adams]

> 4. All of that said, I think that we really need to find
> some _other way than the prompt_ to indicate, completely,
> the search state.  And then the prompt can perhaps just be
> something like "Isearch" in all cases.  The prompt is not
> adequate for conveying just what the current search state is.

Something that I think should be available, starting now, is
an Isearch key binding that describes the current search state
_completely_, much as `C-u C-x =' describes the char at point.

This state information should be on prefix key `C-h'.  It
could be on a new key on that prefix or it could simply be
added to `isearch-describe-mode' - the key could be `C-h m'.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Juri Linkov]

> 1. I agree with Michael in this case.  The prompt should
> faithfully reflect the current search state/mode - until we
> find a better way to convey that information.  In any case,
> the prompt should not mislead by reflecting a different
> search behavior from what is current.

Reflecting all current search states might produce overly long
default prompts as

  Character-folding case-folding whitespace-folding I-search:

>> 4. All of that said, I think that we really need to find
>> some _other way than the prompt_ to indicate, completely,
>> the search state.  And then the prompt can perhaps just be
>> something like "Isearch" in all cases.  The prompt is not
>> adequate for conveying just what the current search state is.
>
> Something that I think should be available, starting now, is
> an Isearch key binding that describes the current search state
> _completely_, much as `C-u C-x =' describes the char at point.
>
> This state information should be on prefix key `C-h'.  It
> could be on a new key on that prefix or it could simply be
> added to `isearch-describe-mode' - the key could be `C-h m'.

Or an option/state (say, ‘isearch-message-verbose’) toggleable
with ‘isearch-toggle-message-verbose’ like other states
(e.g. ‘isearch-toggle-case-fold’) that will toggle between
displaying full/brief state description in the prompt.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Drew Adams]

> > 1. I agree with Michael in this case.  The prompt should
> > faithfully reflect the current search state/mode - until we
> > find a better way to convey that information.  In any case,
> > the prompt should not mislead by reflecting a different
> > search behavior from what is current.
> 
> Reflecting all current search states might produce overly long
> default prompts as
> 
>   Character-folding case-folding whitespace-folding I-search:

Precisely.  The prompt is not the place for such information.
It might have sufficed when we had only regexp and literal.
But even then it did not express that state of case-folding.
And when "lax" whitespace matching was introduced it already
led to less clarity.

> >> 4. All of that said, I think that we really need to find
> >> some _other way than the prompt_ to indicate, completely,
> >> the search state.  And then the prompt can perhaps just be
> >> something like "Isearch" in all cases.  The prompt is not
> >> adequate for conveying just what the current search state is.
> >
> > Something that I think should be available, starting now, is
> > an Isearch key binding that describes the current search state
> > _completely_, much as `C-u C-x =' describes the char at point.
> >
> > This state information should be on prefix key `C-h'.  It
> > could be on a new key on that prefix or it could simply be
> > added to `isearch-describe-mode' - the key could be `C-h m'.
> 
> Or an option/state (say, ‘isearch-message-verbose’) toggleable
> with ‘isearch-toggle-message-verbose’ like other states
> (e.g. ‘isearch-toggle-case-fold’) that will toggle between
> displaying full/brief state description in the prompt.

No.  The point is that the prompt is inadequate for showing
search state/mode.  A more general discussion, and perhaps
more imagination, is needed to find one or more good solutions.

There was some discussion on emacs-devel already, including
suggestions about the mode-line and others, but nothing was
decided, and probably there are better suggestions to be
invented yet.

In the meantime, I suggest we anyway provide complete such
information on a help key.  This "solution" can remain,
regardless of whatother designs might be adopted to provide
some such info all of the time (instead of just on demand).

IOW, add the complete info to a help command now, in any
case.  For the rest, open another discussion on emacs-devel.

Wrt this bug and the prompt, for now: Either drop any
indication of state in the prompt (which might confuse
people) or show only what we showed before (regexp vs
literal), ignoring indicating char folding, case folding,
and whitespace folding.

Char folding, in particular, has introduced complication
in the traditional Isearch UI.  Indicating the current
state is just one such complication.  Char folding is an
area still in the process of definition, which means that
it is anyway likely that its effects on the UI are not
ready to be understood completely or decided now.

So let's be conservative wrt prompt changes etc., for now.

Just one opinion.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Juri Linkov]

> In the meantime, I suggest we anyway provide complete such
> information on a help key.  This "solution" can remain,
> regardless of whatother designs might be adopted to provide
> some such info all of the time (instead of just on demand).

You mean displaying this information in the *Help* buffer
like ‘?’ in query-replace does?

> Wrt this bug and the prompt, for now: Either drop any
> indication of state in the prompt (which might confuse
> people) or show only what we showed before (regexp vs
> literal), ignoring indicating char folding, case folding,
> and whitespace folding.

Right, let's keep the status quo for now, and display
only regexp-function statuses in the prompt.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Drew Adams]

> > In the meantime, I suggest we anyway provide complete such
> > information on a help key.  This "solution" can remain,
> > regardless of whatother designs might be adopted to provide
> > some such info all of the time (instead of just on demand).
> 
> You mean displaying this information in the *Help* buffer
> like ‘?’ in query-replace does?

I mean provide it on some help key during Isearch.

It could be a new key, just for such state/mode status.
Or it could added to the mode help, `C-h m'.

Unless it were added to the top of the `C-h m' output,
I think probably a separate help key (and separate *Help*
content) would be better, so users can easily see all of
the state info without scrolling the help window etc.

It is important to provide a succinct summary in a few
lines, and then follow that by any additional detailed
info that might be helpful.

Such additional info could give details about the
behavior of the different modes, as well as info about
how to change among them.  Yes, some of that info might
repeat some of what is in the `C-h m' output, but that's
OK.

Another possibility might be to add this info to the
regular `C-h m' output, but have a new help key that
takes you directly to the part of that output that
shows the current modes and explains the possibile modes,
i.e., scrolls to that part.  That would have the advantage
of providing this info in the context of isearch help in
general.

You might have other ideas about this.  My point is
mainly that we should provide the info on demand, via
a help key, and we should do that now (for Emacs 25.1).
The form and details are less important.

> > Wrt this bug and the prompt, for now: Either drop any
> > indication of state in the prompt (which might confuse
> > people) or show only what we showed before (regexp vs
> > literal), ignoring indicating char folding, case folding,
> > and whitespace folding.
> 
> Right, let's keep the status quo for now, and display
> only regexp-function statuses in the prompt.

Thank you.

But I do think that we should also add the state/mode
info to a help key _now_.  That info is missing for users,
and we are now complicating the possible states by adding
char folding.

Adding such *Help* info should not be difficult.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Eli Zaretskii]

> From: Juri Linkov <juri@linkov.net>
> Cc: Eli Zaretskii <eliz@gnu.org>,  Kaushal Modi <kaushal.modi@gmail.com>,  michael_heerdegen@web.de,  22991@debbugs.gnu.org,  bruce.connor.am@gmail.com
> Date: Mon, 14 Mar 2016 01:40:55 +0200
> 
> > Wrt this bug and the prompt, for now: Either drop any
> > indication of state in the prompt (which might confuse
> > people) or show only what we showed before (regexp vs
> > literal), ignoring indicating char folding, case folding,
> > and whitespace folding.
> 
> Right, let's keep the status quo for now, and display
> only regexp-function statuses in the prompt.

But what is the status quo?  This is a discussion of a bug report,
remember?  Are you saying we should leave the bug in place?

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Kaushal Modi]

[-- Attachment #1: Type: text/plain, Size: 836 bytes --]

Looks like this discussion is going in loops.

Summary:

(1) I believe that the patch I have fixes the bug I reported and the one
Eli pointed out. And it does so without bringing a drastic change to the
way the description string is generated.
(2) As for an extra binding that shows the complete list of all the active
isearch modes, I agree with Drew that it can show a complete picture of
what isearch is doing behind the scenes. @Drew: Can you please open it as a
new bug report?
(3) Put up a proposal with a patch ( probably at
https://www.emacswiki.org/emacs/Proposals ) on how we should change the way
the descriptions are shown right now.
(4) Email emacs-devel that proposal and ask for
opinions/corrections/improvements.
(5) Based on the consensus, push the resultant code to master branch.

So let's focus on step 1 for now :)

[-- Attachment #2: Type: text/html, Size: 1262 bytes --]

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Drew Adams]

> (2) As for an extra binding that shows the complete list of all the active
> isearch modes, I agree with Drew that it can show a complete picture of
> what isearch is doing behind the scenes. @Drew: Can you please open it as
> a new bug report?

Done: #23010.

(Please use plain text, not HTML text, for this list. Thx.)

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Artur Malabarba]

[-- Attachment #1: Type: text/plain, Size: 443 bytes --]

On 14 Mar 2016 10:50 am, "Kaushal Modi" <kaushal.modi@gmail.com> wrote:
>
> Looks like this discussion is going in loops.
>
> Summary:
>
> (1) I believe that the patch I have fixes the bug I reported and the one
Eli pointed out. And it does so without bringing a drastic change to the
way the description string is generated.

I can't push for the time being, but I agree with the patch.

Eli, if you agree with it too, could you push it?
TIA

[-- Attachment #2: Type: text/html, Size: 610 bytes --]

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Eli Zaretskii]

> Date: Mon, 14 Mar 2016 13:29:24 -0300
> From: Artur Malabarba <bruce.connor.am@gmail.com>
> Cc: 22991@debbugs.gnu.org, Drew Adams <drew.adams@oracle.com>, 
> 	Michael Heerdegen <michael_heerdegen@web.de>, Juri Linkov <juri@linkov.net>, Eli Zaretskii <eliz@gnu.org>
> 
> I can't push for the time being, but I agree with the patch. 
> 
> Eli, if you agree with it too, could you push it? 

Pushed.

bug#22991: 25.0.92: C-u C-s does not display "Regexp I-search:" in the echo area

[Kaushal Modi]

[-- Attachment #1: Type: text/plain, Size: 16 bytes --]

Thank you guys!

[-- Attachment #2: Type: text/html, Size: 68 bytes --]