C-g from Isearch (was: Isearch: retrieve last successful search string from when you quit (`C-g'))

C-g from Isearch (was: Isearch: retrieve last successful search string from when you quit (`C-g'))

[Dani Moncayo]

>> Quite. Perhaps I'm just an idiot, but I always thought the current
>> isearch behaviour of not immediately quitting when C-g was stuck on a
>> failed search was a *bug*.
>
> That is something very different from what this thread is about (and not a bug,
> IMO).  The C-g behavior in Emacs incremental search is somewhat complex (depends
> on the search state), and fine as is (IMO).

IIUC, the behavior of C-g from Isearch is not that complex.  As
documented in (info "(emacs)Error in Isearch"), `C-g' from Isearch
does this:
* If the current search is successful, `C-g' cancels the Isearch.
* Else, `C-g' restores the last succesful search string.

And I, like Nix, don't like the second case.  I would like C-g to
_always_ cancel the Isearch.

So IMO this behavior should be configurable.

-- 
Dani Moncayo

RE: C-g from Isearch (was: Isearch: retrieve last successful search string from when you quit (`C-g'))

[Drew Adams]

> > `C-g' from Isearch does this:
> * If the current search is successful, `C-g' cancels the Isearch.
> * Else, `C-g' restores the last succesful search string.
> 
> And I, like Nix, don't like the second case.  I would like C-g to
> _always_ cancel the Isearch.  So IMO this behavior should be
> configurable.

It already is.  Just do this:

(define-key isearch-mode-map 'isearch-cancel)

And you also already have a key binding for this: `ESC ESC ESC'. 

Re: C-g from Isearch (was: Isearch: retrieve last successful search string from when you quit (`C-g'))

[Dani Moncayo]

On Tue, Oct 2, 2012 at 6:22 PM, Drew Adams <drew.adams@oracle.com> wrote:
>> > `C-g' from Isearch does this:
>> * If the current search is successful, `C-g' cancels the Isearch.
>> * Else, `C-g' restores the last succesful search string.
>>
>> And I, like Nix, don't like the second case.  I would like C-g to
>> _always_ cancel the Isearch.  So IMO this behavior should be
>> configurable.
>
> It already is.  Just do this:
>
> (define-key isearch-mode-map 'isearch-cancel)

Right.  I'll add that (well, adding the missing "C-g") to my init file.  Thanks.

>
> And you also already have a key binding for this: `ESC ESC ESC'.
>

I don't want to use another binding.  I just want C-g to
unconditionally exit Isearch.

-- 
Dani Moncayo

Re: C-g from Isearch

[Stefan Monnier]

> And I, like Nix, don't like the second case.  I would like C-g to
> _always_ cancel the Isearch.
> So IMO this behavior should be configurable.

I don't think the difference is terribly important, but a configuration
option for that is acceptable.


        Stefan

Re: C-g from Isearch

[Nix]

On 2 Oct 2012, Stefan Monnier said:

>> And I, like Nix, don't like the second case.  I would like C-g to
>> _always_ cancel the Isearch.
>> So IMO this behavior should be configurable.
>
> I don't think the difference is terribly important, but a configuration
> option for that is acceptable.

Neither do I. Like I said, it is initially confusing and prone to seem
like a bug, but is both documented and a useful behaviour for which no
other keybinding is particularly obviously appropriate.

-- 
NULL && (void)

RE: C-g from Isearch

[Drew Adams]

> > And I, like Nix, don't like the second case.  I would like
> > C-g to _always_ cancel the Isearch.  So IMO this behavior
> > should be configurable.
> 
> I don't think the difference is terribly important, but a 
> configuration option for that is acceptable.

Coming from someone who often resists adding yet another user option, this is a
bit surprising to me.  IMHO, this is a case where your more usual stance wrt new
options (YAGNI) would be appropriate.

It is trivial for a user to bind C-g to `isearch-cancel'.
(And that command already has another simple binding.)

I would, however, support documenting this in the manual, where we discuss the
out-of-the-box behavior of C-g.  Just mention that ESC ESC ESC (command
`isearch-cancel') always cancels isearch (and you can bind it to another key if
you want).

Re: C-g from Isearch

[Stefan Monnier]

> It is trivial for a user to bind C-g to `isearch-cancel'.

Indeed, I had not noticed it, thank you.  Then a config var is
indeed superfluous.


        Stefan

C-g from Isearch (was: Isearch: retrieve last successful search string from when you quit (`C-g'))

[Stephen J. Turnbull]

Dani Moncayo writes:

 > documented in (info "(emacs)Error in Isearch"), `C-g' from Isearch
 > does this:
 > * If the current search is successful, `C-g' cancels the Isearch.
 > * Else, `C-g' restores the last succesful search string.

In XEmacs, this is useless duplication.  Deleting the character has
the same effect some rather large fraction of the time, as it's
unusual that multiple not found characters are typed.

 > And I, like Nix, don't like the second case.  I would like C-g to
 > _always_ cancel the Isearch.
 > 
 > So IMO this behavior should be configurable.

I'm tempted to just change XEmacs's behavior to "always cancel" and
see how many cries of outrage can be induced.

Re: C-g from Isearch (was: Isearch: retrieve last successful search string from when you quit (`C-g'))

[Dani Moncayo]

>  > documented in (info "(emacs)Error in Isearch"), `C-g' from Isearch
>  > does this:
>  > * If the current search is successful, `C-g' cancels the Isearch.
>  > * Else, `C-g' restores the last succesful search string.
>
> In XEmacs, this is useless duplication.  Deleting the character has
> the same effect some rather large fraction of the time, as it's
> unusual that multiple not found characters are typed.

AFAIK, this is true also for GNU Emacs.  Typing <backspace> in Isearch
restores the previous Isearch state.  So I agree with you: `C-g'
duplicates this feature (almost - `C-g' restores the last successful
state) at the expense of annoying those user who expect this command
to its usual job (cancel the current command).

>  > And I, like Nix, don't like the second case.  I would like C-g to
>  > _always_ cancel the Isearch.
>  >
>  > So IMO this behavior should be configurable.
>
> I'm tempted to just change XEmacs's behavior to "always cancel" and
> see how many cries of outrage can be induced.

I've done that in my init file:
  (define-key isearch-mode-map [?\C-g] 'isearch-cancel)


--
Dani Moncayo

Re: C-g from Isearch

[Jason Rumney]

"Stephen J. Turnbull" <stephen@xemacs.org> writes:

> In XEmacs, this is useless duplication.  Deleting the character has
> the same effect some rather large fraction of the time, as it's
> unusual that multiple not found characters are typed.

Fast typing, and yank are two ways this can happen quite easily.  That
said, I never knew about this behaviour of C-g in isearch before, and
never would have expected it.  I usually end my isearch with RET, but if
I did try C-g I would be surprised to see it working like undo rather
than quit. 

Re: C-g from Isearch

[Dani Moncayo]

>> In XEmacs, this is useless duplication.  Deleting the character has
>> the same effect some rather large fraction of the time, as it's
>> unusual that multiple not found characters are typed.
>
> Fast typing, and yank are two ways this can happen quite easily.

Yank doesn't count here, since the whole yanked string will be removed
from the search string with a single <backspace> keystroke.

> That
> said, I never knew about this behaviour of C-g in isearch before, and
> never would have expected it.  I usually end my isearch with RET, but if
> I did try C-g I would be surprised to see it working like undo rather
> than quit.

I agree, but that is the current behavior in Emacs.  At least, you can
fix this with
  (define-key isearch-mode-map [?\C-g] 'isearch-cancel)

-- 
Dani Moncayo

Re: C-g from Isearch

[Jason Rumney]

Dani Moncayo <dmoncayo@gmail.com> writes:

> Yank doesn't count here, since the whole yanked string will be removed
> from the search string with a single <backspace> keystroke.

More surprising behaviour I never knew about.  Inconsistencies like
this, and the old C-y behaviour, which thankfully got fixed in 24.1,
are probably why I never use the more advanced isearch features.