Include ivy + counsel in Emacs core?

Include ivy + counsel in Emacs core?

[Andrea Corallo via Emacs development discussions.]

Hi all,

as I'm relatively new here forgive me if this was already asked.

Given all the recent discussions about modern Emacs but also fido-mode
came to my mind the following question:

was ever considered the inclusion of ivy + counsel in the Emacs core
distribution?

ivy+counsel it's a quality package that reached already noticeable
popularity.  Does fuzzy completion with vertical output display, and its
counsel-M-X shows also the command key binding (feature I consider for
my experience a _game changer_ for softening the learning curve).
Finally it's already in ELPA.

I'm probably wrong but I get the impression it has already everything
fido-mode is trying to get.

I'd see its inclusion together with a polished which-key a perfect
match.

Thanks

  Andrea

Re: Include ivy + counsel in Emacs core?

[Ricardo Wurmus]

Andrea Corallo via Emacs development discussions. <emacs-devel@gnu.org> writes:

> was ever considered the inclusion of ivy + counsel in the Emacs core
> distribution?
>
> ivy+counsel it's a quality package that reached already noticeable
> popularity.  Does fuzzy completion with vertical output display, and its
> counsel-M-X shows also the command key binding (feature I consider for
> my experience a _game changer_ for softening the learning curve).
> Finally it's already in ELPA.

As someone who has used ivy for several months I should note that it
doesn’t work well with “M-x rgrep” (it tries to complete files when
asked for a pattern); see also
https://github.com/abo-abo/swiper/issues/286

There are also a couple of things that might trip up new users, for
example the need to hit C-j to confirm a new file name instead of
selecting a similar matching file name (which is what RET gives you).

While I think it’s pretty nice in general (and I do prefer it over Helm)
and an improvement over the default of no completion, I do feel it’s a
little … invasive or coercive in that it’s a bit awkward to bypass it.

-- 
Ricardo

RE: Include ivy + counsel in Emacs core?

[arthur miller]

[-- Attachment #1: Type: text/plain, Size: 1953 bytes --]

I have used ivy+counsel for about some year or so for quite some time ago, but I have since switched to Helm and have been using it for like maybe years now.  In my opinion it is less buggy, has better and more integration with other packages and also feels for some reason more straightforward than Ivy and counsel.

In that case I would probably  vote for Helm,  but I am completely happy with such big packages being out of Emacs core. I would even like to see Org being left out or at least demoted to Elpa.



-------- Originalmeddelande --------
Från: Ricardo Wurmus <rekado@elephly.net>
Datum: 2020-09-13 21:13 (GMT+01:00)
Till: Andrea Corallo <akrl@sdf.org>
Kopia: Oleh Krehel <ohwoeowho@gmail.com>, emacs-devel@gnu.org
Ämne: Re: Include ivy + counsel in Emacs core?


Andrea Corallo via Emacs development discussions. <emacs-devel@gnu.org> writes:

> was ever considered the inclusion of ivy + counsel in the Emacs core
> distribution?
>
> ivy+counsel it's a quality package that reached already noticeable
> popularity.  Does fuzzy completion with vertical output display, and its
> counsel-M-X shows also the command key binding (feature I consider for
> my experience a _game changer_ for softening the learning curve).
> Finally it's already in ELPA.

As someone who has used ivy for several months I should note that it
doesn’t work well with “M-x rgrep” (it tries to complete files when
asked for a pattern); see also
https://github.com/abo-abo/swiper/issues/286

There are also a couple of things that might trip up new users, for
example the need to hit C-j to confirm a new file name instead of
selecting a similar matching file name (which is what RET gives you).

While I think it’s pretty nice in general (and I do prefer it over Helm)
and an improvement over the default of no completion, I do feel it’s a
little … invasive or coercive in that it’s a bit awkward to bypass it.

--
Ricardo


[-- Attachment #2: Type: text/html, Size: 2863 bytes --]

Re: Include ivy + counsel in Emacs core?

[ej32u--- via Emacs development discussions.]

On 9/13/20 4:13 PM, arthur miller wrote:
> I have used ivy+counsel for about some year or so for quite some time ago, but I 
> have since switched to Helm and have been using it for like maybe years now.  In 
> my opinion it is less buggy, has better and more integration with other packages 
> and also feels for some reason more straightforward than Ivy and counsel.
> 
> In that case I would probably  vote for Helm,  but I am completely happy with 
> such big packages being out of Emacs core. I would even like to see Org being 
> left out or at least demoted to Elpa.

I believe the Helm maintainer no longer wishes to work on Helm (see here: 
https://github.com/emacs-helm/helm/issues/2386).

Re: Include ivy + counsel in Emacs core?

[Doug Davis]

ej32u--- via "Emacs development discussions." <emacs-devel@gnu.org>
writes:

> On 9/13/20 4:13 PM, arthur miller wrote:
>> I have used ivy+counsel for about some year or so for quite some time ago, but I 
>> have since switched to Helm and have been using it for like maybe years now.  In 
>> my opinion it is less buggy, has better and more integration with other packages 
>> and also feels for some reason more straightforward than Ivy and counsel.
>> 
>> In that case I would probably  vote for Helm,  but I am completely happy with 
>> such big packages being out of Emacs core. I would even like to see Org being 
>> left out or at least demoted to Elpa.
>
> I believe the Helm maintainer no longer wishes to work on Helm (see here: 
> https://github.com/emacs-helm/helm/issues/2386).

Perhaps this is a good reason to bring it into Emacs core! The package
is quite mature. But I imagine the copyright assignment from all 155
contributors (listed on the GitHub repo) would be potential road block?

Re: Include ivy + counsel in Emacs core?

[Stefan Monnier]

> ivy+counsel it's a quality package that reached already noticeable
> popularity.  Does fuzzy completion with vertical output display, and its
> counsel-M-X shows also the command key binding (feature I consider for
> my experience a _game changer_ for softening the learning curve).
> Finally it's already in ELPA.

Last I heard Ivy still had problems with some completion tables (IIRC
with those completion tables that return non-trivial values from
`completion-boundaries`, such as the completion table used for
filenames, tho there might be other issues such as the support for
unquote/requote).  If that's still the case (it was several years ago),
then enabling Ivy by default would either introduce regressions when
completing against such completion tables, or would force us to use
2 different UIs (the Ivy one where it works well, as some other UI for
those other cases) ;-(


        Stefan

RE: Include ivy + counsel in Emacs core?

[arthur miller]

[-- Attachment #1: Type: text/plain, Size: 1172 bytes --]

Oki, that was news for me. Let us hope Thierry will change his mind and continue or someone else will help.

I remember that he announced some time ago that he will maybe stop developing Helm due to lack of time and resource.




-------- Originalmeddelande --------
Från: "ej32u--- via Emacs development discussions." <emacs-devel@gnu.org>
Datum: 2020-09-13 22:26 (GMT+01:00)
Till: emacs-devel@gnu.org
Ämne: Re: Include ivy + counsel in Emacs core?

On 9/13/20 4:13 PM, arthur miller wrote:
> I have used ivy+counsel for about some year or so for quite some time ago, but I
> have since switched to Helm and have been using it for like maybe years now.  In
> my opinion it is less buggy, has better and more integration with other packages
> and also feels for some reason more straightforward than Ivy and counsel.
>
> In that case I would probably  vote for Helm,  but I am completely happy with
> such big packages being out of Emacs core. I would even like to see Org being
> left out or at least demoted to Elpa.

I believe the Helm maintainer no longer wishes to work on Helm (see here:
https://github.com/emacs-helm/helm/issues/2386).



[-- Attachment #2: Type: text/html, Size: 2042 bytes --]

Re: Include ivy + counsel in Emacs core?

[Ergus]

On Sun, Sep 13, 2020 at 04:36:13PM -0400, Stefan Monnier wrote:
>> ivy+counsel it's a quality package that reached already noticeable
>> popularity.  Does fuzzy completion with vertical output display, and its
>> counsel-M-X shows also the command key binding (feature I consider for
>> my experience a _game changer_ for softening the learning curve).
>> Finally it's already in ELPA.
>
>Last I heard Ivy still had problems with some completion tables (IIRC
>with those completion tables that return non-trivial values from
>`completion-boundaries`, such as the completion table used for
>filenames, tho there might be other issues such as the support for
>unquote/requote).  If that's still the case (it was several years ago),
>then enabling Ivy by default would either introduce regressions when
>completing against such completion tables, or would force us to use
>2 different UIs (the Ivy one where it works well, as some other UI for
>those other cases) ;-(
>
>
>        Stefan
>
An important part of ivy was re-implemented the last year to simplify
and unify some interfaces and many errors were corrected.

I use ivy since 4 years ago with not big issues for the moment (When I
have them I have reported and they have been solved almost immediately).
But it needs some improve before becoming part of vanilla.

It would be very nice if you comment/report/ask that to Abo-Abo. To get
an update of the current status.

FTM I will refer him to this email.

Re: Include ivy + counsel in Emacs core?

[Dmitry Gutov]

On 13.09.2020 21:53, Andrea Corallo via Emacs development discussions. 
wrote:
> I'm probably wrong but I get the impression it has already everything
> fido-mode is trying to get.

Ivy is good, but is misses one of the major goals of fido-mode: file 
name completion behavior like ido-find-file. In particular, entering a 
directory on RET (without exiting completion). The part where it goes up 
a directory on backspace is strangely there.

Re: Include ivy + counsel in Emacs core?

[Jai Flack]

Hi emacs-devel,

Would it be out-of-place for me to recommend selectrum as a
replacement?  On the GitHub (https://github.com/raxod502/selectrum) it
advertises to not break any of the existing abstractions, something
which they claim both Ivy and Helm are terrible at.

The biggest road-block I can see is that its MIT licensed but if they
were willing to re-license it, the project only has 9 contributors,
all of which seem quite active.

--
Thanks for Emacs!

Re: Include ivy + counsel in Emacs core?

[Andrea Corallo via Emacs development discussions.]

Ergus <spacibba@aol.com> writes:

> It would be very nice if you comment/report/ask that to Abo-Abo. To get
> an update of the current status.

He's in Cc on all this thread.

  Andrea

Re: Include ivy + counsel in Emacs core?

[Oleh Krehel]

> Ivy is good, but is misses one of the major goals of fido-mode: file 
> name completion behavior like ido-find-file. In particular, entering a 
> directory on RET (without exiting completion). The part where it goes 
> up a directory on backspace is strangely there.

Actually, I thought this was the default. But in my config I have:

     (define-key ivy-minibuffer-map (kbd "<return>") 'ivy-alt-done)

which makes RET enter a directory.


     Oleh

Re: Include ivy + counsel in Emacs core?

[Oleh Krehel]

On 9/13/20 11:26 PM, Ergus wrote:
> On Sun, Sep 13, 2020 at 04:36:13PM -0400, Stefan Monnier wrote:
>>> ivy+counsel it's a quality package that reached already noticeable
>>> popularity.  Does fuzzy completion with vertical output display, and 
>>> its
>>> counsel-M-X shows also the command key binding (feature I consider for
>>> my experience a _game changer_ for softening the learning curve).
>>> Finally it's already in ELPA.
>>
>> Last I heard Ivy still had problems with some completion tables (IIRC
>> with those completion tables that return non-trivial values from
>> `completion-boundaries`, such as the completion table used for
>> filenames, tho there might be other issues such as the support for
>> unquote/requote).  If that's still the case (it was several years ago),
>> then enabling Ivy by default would either introduce regressions when
>> completing against such completion tables, or would force us to use
>> 2 different UIs (the Ivy one where it works well, as some other UI for
>> those other cases) ;-(
>>
>>
>>        Stefan
>>
> An important part of ivy was re-implemented the last year to simplify
> and unify some interfaces and many errors were corrected.
>
> I use ivy since 4 years ago with not big issues for the moment (When I
> have them I have reported and they have been solved almost immediately).
> But it needs some improve before becoming part of vanilla.
>
> It would be very nice if you comment/report/ask that to Abo-Abo. To get
> an update of the current status.

I'm eager to solve the problems with completion tables. Some of them 
should already be solved.

If there was a bug report with a reproducible scenario, that would make 
things move faster.


     Oleh

Re: Include ivy + counsel in Emacs core?

[Oleh Krehel]

Hi Andrea,

Thanks for bringing this up. I'm also interested in including Ivy as a 
core package, pre-configured

in a no-surprise way. If the maintainers want to proceed, I'd be happy 
to iron out the small annoyances that prevent it

from being a drop-in replacement of e.g. ido-mode.


     Oleh


On 9/13/20 8:53 PM, Andrea Corallo wrote:
> Hi all,
>
> as I'm relatively new here forgive me if this was already asked.
>
> Given all the recent discussions about modern Emacs but also fido-mode
> came to my mind the following question:
>
> was ever considered the inclusion of ivy + counsel in the Emacs core
> distribution?
>
> ivy+counsel it's a quality package that reached already noticeable
> popularity.  Does fuzzy completion with vertical output display, and its
> counsel-M-X shows also the command key binding (feature I consider for
> my experience a _game changer_ for softening the learning curve).
> Finally it's already in ELPA.
>
> I'm probably wrong but I get the impression it has already everything
> fido-mode is trying to get.
>
> I'd see its inclusion together with a polished which-key a perfect
> match.
>
> Thanks
>
>    Andrea

Re: Include ivy + counsel in Emacs core?

[Dmitry Gutov]

On 14.09.2020 14:09, Oleh Krehel wrote:
>> Ivy is good, but is misses one of the major goals of fido-mode: file 
>> name completion behavior like ido-find-file. In particular, entering a 
>> directory on RET (without exiting completion). The part where it goes 
>> up a directory on backspace is strangely there.
> 
> Actually, I thought this was the default. But in my config I have:
> 
>      (define-key ivy-minibuffer-map (kbd "<return>") 'ivy-alt-done)
> 
> which makes RET enter a directory.

Good suggestion, thanks!

Re: Include ivy + counsel in Emacs core?

[Basil L. Contovounesios]

Oleh Krehel <ohwoeowho@gmail.com> writes:

> On 9/13/20 11:26 PM, Ergus wrote:
>> On Sun, Sep 13, 2020 at 04:36:13PM -0400, Stefan Monnier wrote:
>>> Last I heard Ivy still had problems with some completion tables (IIRC
>>> with those completion tables that return non-trivial values from
>>> `completion-boundaries`, such as the completion table used for
>>> filenames, tho there might be other issues such as the support for
>>> unquote/requote).  If that's still the case (it was several years ago),
>>> then enabling Ivy by default would either introduce regressions when
>>> completing against such completion tables, or would force us to use
>>> 2 different UIs (the Ivy one where it works well, as some other UI for
>>> those other cases) ;-(
>>>
>>>
>>>        Stefan
>>>
>> An important part of ivy was re-implemented the last year to simplify
>> and unify some interfaces and many errors were corrected.
>>
>> I use ivy since 4 years ago with not big issues for the moment (When I
>> have them I have reported and they have been solved almost immediately).
>> But it needs some improve before becoming part of vanilla.
>>
>> It would be very nice if you comment/report/ask that to Abo-Abo. To get
>> an update of the current status.
>
> I'm eager to solve the problems with completion tables. Some of them should
> already be solved.
>
> If there was a bug report with a reproducible scenario, that would make things
> move faster.

See e.g. https://github.com/abo-abo/swiper/issues/1597 for an issue with
programmed completion.

-- 
Basil

Re: Include ivy + counsel in Emacs core?

[Basil L. Contovounesios]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> ivy+counsel it's a quality package that reached already noticeable
>> popularity.  Does fuzzy completion with vertical output display, and its
>> counsel-M-X shows also the command key binding (feature I consider for
>> my experience a _game changer_ for softening the learning curve).
>> Finally it's already in ELPA.
>
> Last I heard Ivy still had problems with some completion tables (IIRC
> with those completion tables that return non-trivial values from
> `completion-boundaries`, such as the completion table used for
> filenames, tho there might be other issues such as the support for
> unquote/requote).  If that's still the case (it was several years ago),
> then enabling Ivy by default would either introduce regressions when
> completing against such completion tables, or would force us to use
> 2 different UIs (the Ivy one where it works well, as some other UI for
> those other cases) ;-(

I don't think anyone's suggesting enabling it by default.  But then,
given that it's already in GNU ELPA, I'm not sure what would be gained
from putting it in core, other than greater exposure to the experts on
emacs-devel and bug-gnu-emacs.

-- 
Basil