merging results from completion-at-point-functions

merging results from completion-at-point-functions

[Eric S Fraga]

Hello all,

I currently use '(cape-dabbrev cape-ispell cape-dict) as the list for
completion-at-point-functions when writing prose.

I keep trying different orderings of these entries as the behaviour
changes significantly, not surprisingly given the /greedy/ nature of
completing-at-point, if I understand things correctly.  Ideally, I would
like a merged search for potential completions, such as I believe
company-mode does if you group backends.

Is it possible to specify a grouping or merge operation?  Or is there a
generalised capf that does this, I guess?

Thank you,
eric

-- 
Professor Eric S Fraga, UCL; GnuPG: 0xc89193d8fffcf67d

Re: merging results from completion-at-point-functions

[Eric Abrahamsen]

Eric S Fraga <e.fraga@ucl.ac.uk> writes:

> Hello all,
>
> I currently use '(cape-dabbrev cape-ispell cape-dict) as the list for
> completion-at-point-functions when writing prose.
>
> I keep trying different orderings of these entries as the behaviour
> changes significantly, not surprisingly given the /greedy/ nature of
> completing-at-point, if I understand things correctly.  Ideally, I would
> like a merged search for potential completions, such as I believe
> company-mode does if you group backends.
>
> Is it possible to specify a grouping or merge operation?  Or is there a
> generalised capf that does this, I guess?

Two possibilities: since you're using cape, use `cape-super-capf' to
merge a bunch of cape capf functions into one.

Other possibility: use `completion-table-merge' in exactly the same way.
This should be the general version of what `cape-super-capf' does.

Please try both and report back! I'm very curious, as well.

Eric

Re: merging results from completion-at-point-functions

[Eric S Fraga]

Hi Eric,

thank you for the quick response.

On Friday, 29 Apr 2022 at 08:08, Eric Abrahamsen wrote:
> Two possibilities: since you're using cape, use `cape-super-capf' to
> merge a bunch of cape capf functions into one.

First of all, apologies: this is in the cape documentation which I
forgot to look at again.  Nevertheless, this doesn't work at all.  If I
do

(setq-local completion-at-point-functions
    (list (cape-super-capf #'cape-dabbrev #'cape-dict #'cape-keyword)))

when visiting an appropriate file, as the web page
(https://github.com/minad/cape) suggests, nothing happens when invoking
completion-at-point.

> Other possibility: use `completion-table-merge' in exactly the same way.

This one I cannot figure out how to use.  My elisp-fu is obviously not
up to scratch. :-(

Thanks again,
eric

-- 
Eric S Fraga with org 9.5.3 in Emacs 29.0.50 on Debian 11.3

Re: merging results from completion-at-point-functions

[Eric Abrahamsen]

Eric S Fraga <e.fraga@ucl.ac.uk> writes:

> Hi Eric,
>
> thank you for the quick response.
>
> On Friday, 29 Apr 2022 at 08:08, Eric Abrahamsen wrote:
>> Two possibilities: since you're using cape, use `cape-super-capf' to
>> merge a bunch of cape capf functions into one.
>
> First of all, apologies: this is in the cape documentation which I
> forgot to look at again.  Nevertheless, this doesn't work at all.  If I
> do
>
> (setq-local completion-at-point-functions
>     (list (cape-super-capf #'cape-dabbrev #'cape-dict #'cape-keyword)))
>
> when visiting an appropriate file, as the web page
> (https://github.com/minad/cape) suggests, nothing happens when invoking
> completion-at-point.

Sounds like some kind of bug, then. I'd check the github issues (several
of the closed issues are about super capes) and maybe open a new one.

>> Other possibility: use `completion-table-merge' in exactly the same way.
>
> This one I cannot figure out how to use.  My elisp-fu is obviously not
> up to scratch. :-(

No worries! It's supposed to work the same way:

(setq my-merged-table (completion-table-merge #'cap-dabbrev
#'cape-dict))

(setq-local completion-at-point-functions (list #'my-merged-table))

That appears to be working for me (at least, it raises a "lookup-words"
error from `cape-dict', so at least it's *trying* to work).

Eric

Re: merging results from completion-at-point-functions

[Stefan Monnier via Users list for the GNU Emacs text editor]

>> (setq-local completion-at-point-functions
>>     (list (cape-super-capf #'cape-dabbrev #'cape-dict #'cape-keyword)))

Note that this sets up these CAPF functions only for the buffer that's
current when the code is executed, so if you put it into your .emacs
it's probably going to affect completions in *scratch* only.

> (setq my-merged-table (completion-table-merge #'cap-dabbrev
> #'cape-dict))

I don't think so: `cape-dabbrev` is a CAPF function, not a completion table.


        Stefan

Re: merging results from completion-at-point-functions

[Eric Abrahamsen]

Stefan Monnier via Users list for the GNU Emacs text editor
<help-gnu-emacs@gnu.org> writes:

>>> (setq-local completion-at-point-functions
>>>     (list (cape-super-capf #'cape-dabbrev #'cape-dict #'cape-keyword)))
>
> Note that this sets up these CAPF functions only for the buffer that's
> current when the code is executed, so if you put it into your .emacs
> it's probably going to affect completions in *scratch* only.
>
>> (setq my-merged-table (completion-table-merge #'cap-dabbrev
>> #'cape-dict))
>
> I don't think so: `cape-dabbrev` is a CAPF function, not a completion table.

Ooops. Misinformation! I do often forget the distinction between capf
functions and completion tables.

Re: merging results from completion-at-point-functions

[Ergus]

On Fri, Apr 29, 2022 at 04:52:21PM -0400, Stefan Monnier via Users list for the GNU Emacs text editor wrote:
>>> (setq-local completion-at-point-functions
>>>     (list (cape-super-capf #'cape-dabbrev #'cape-dict #'cape-keyword)))
>
>Note that this sets up these CAPF functions only for the buffer that's
>current when the code is executed, so if you put it into your .emacs
>it's probably going to affect completions in *scratch* only.
>
>> (setq my-merged-table (completion-table-merge #'cap-dabbrev
>> #'cape-dict))
>
>I don't think so: `cape-dabbrev` is a CAPF function, not a completion table.
>
>
>        Stefan
>
>
Hi Stefan...

Maybe a bit off-topic... but somehow related:

Is it there a way to show the current capf backend for example in the
mode-line?? Company has such thing, and it may be useful for example, to
know which completion-at-point-functions is working better than
others... Because sometimes different functions generate similar
outputs and it may be useful to know which are the one in use, specially
when exclusive is false in some of them.

Any Idea?
Best,
Ergus.

Re: merging results from completion-at-point-functions

[Stefan Monnier]

> Is it there a way to show the current capf backend for example in the
> mode-line??

The code doesn't offer this feature yet, no.  Patch welcome,


        Stefan

Re: merging results from completion-at-point-functions

[Emanuel Berg]

Stefan Monnier wrote:

>> Is it there a way to show the current capf backend for
>> example in the mode-line??
>
> The code doesn't offer this feature yet, no. Patch welcome,

And when will the program come that will write this program
for us?

-- 
underground experts united
https://dataswamp.org/~incal

Re: merging results from completion-at-point-functions

[Stefan Monnier via Users list for the GNU Emacs text editor]

Emanuel Berg [2022-05-01 04:58:41] wrote:
> Stefan Monnier wrote:
>>> Is it there a way to show the current capf backend for
>>> example in the mode-line??
>> The code doesn't offer this feature yet, no. Patch welcome,
> And when will the program come that will write this program
> for us?

The future is long past.
That program you're talking about is called Emacs.


        Stefan

Re: merging results from completion-at-point-functions

[tomas]

[-- Attachment #1: Type: text/plain, Size: 388 bytes --]

On Sun, May 01, 2022 at 01:20:34PM -0400, Stefan Monnier via Users list for the GNU Emacs text editor wrote:
> Emanuel Berg [2022-05-01 04:58:41] wrote:

[...]

> > And when will the program come that will write this program
> > for us?
> 
> The future is long past.
> That program you're talking about is called Emacs.

You just have to tell it. Nicely :-)

Cheers
-- 
t

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

Re: merging results from completion-at-point-functions

[Emanuel Berg]

tomas wrote:

>>> And when will the program come that will write this
>>> program for us?
>> 
>> The future is long past.
>> That program you're talking about is called Emacs.
>
> You just have to tell it. Nicely :-)

Little by little programs take control over America.

By and by software claims all of Russia.

Cold War 3.0 - AI vs AI?

-- 
underground experts united
https://dataswamp.org/~incal

Re: merging results from completion-at-point-functions

[Eric S Fraga]

On Friday, 29 Apr 2022 at 16:52, Stefan Monnier via Users list for the GNU Emacs text editor wrote:
>>> (setq-local completion-at-point-functions
>>>     (list (cape-super-capf #'cape-dabbrev #'cape-dict #'cape-keyword)))
>
> Note that this sets up these CAPF functions only for the buffer that's
> current when the code is executed, so if you put it into your .emacs
> it's probably going to affect completions in *scratch* only.

Yes, that's clear.  Thank you.

I am setting the variable in the right place to have
it set for the buffer I am writing in.  But it just doesn't do anything
when I try to complete.

To check, if I use

(setq-local completion-at-point-functions (list #'cape-dabbrev))

instead of the expression above, that particular capf does get invoked.

-- 
Eric S Fraga with org 9.5.3 in Emacs 29.0.50 on Debian 11.3

Re: merging results from completion-at-point-functions

[Stefan Monnier via Users list for the GNU Emacs text editor]

Eric S Fraga [2022-05-01 17:29:54] wrote:
> On Friday, 29 Apr 2022 at 16:52, Stefan Monnier via Users list for the GNU Emacs text editor wrote:
>>>> (setq-local completion-at-point-functions
>>>>     (list (cape-super-capf #'cape-dabbrev #'cape-dict #'cape-keyword)))
>>
>> Note that this sets up these CAPF functions only for the buffer that's
>> current when the code is executed, so if you put it into your .emacs
>> it's probably going to affect completions in *scratch* only.
>
> Yes, that's clear.  Thank you.
>
> I am setting the variable in the right place to have
> it set for the buffer I am writing in.  But it just doesn't do anything
> when I try to complete.
>
> To check, if I use
>
> (setq-local completion-at-point-functions (list #'cape-dabbrev))
>
> instead of the expression above, that particular capf does get invoked.

What happens if you go to the place where you want completion to take
place and then do:

    M-: (funcall (cape-super-capf #'cape-dabbrev #'cape-dict #'cape-keyword)) RET

[ It's not supposed to have any effect other than returning a value, so
  the question is what value does it return.  ]


        Stefan

Re: merging results from completion-at-point-functions

[Eric S Fraga]

On Sunday,  1 May 2022 at 13:22, Stefan Monnier via Users list for the GNU Emacs text editor wrote:
>     M-: (funcall (cape-super-capf #'cape-dabbrev #'cape-dict #'cape-keyword)) RET

It didn't do anything.

After a lot of experimentation and some discussions on the cape github
issue pages, I have determined that cape-super-capf does not work with
vanilla completion-at-point but does if you use corfu on top of it.

The settings I now have are:

  (global-corfu-mode)
  (corfu-history-mode)

and

(setq-local completion-at-point-functions
 (list
  (cape-super-capf #'cape-tex #'cape-dabbrev #'cape-dict #'cape-ispell)))

for org mode buffers only.

I have set up corfu to be invoked with TAB as I find the auto mode too
intrusive and it imposes a small but noticeable lag in my typing.

-- 
Eric S Fraga with org 9.5.3 in Emacs 29.0.50 on Debian 11.3

Re: merging results from completion-at-point-functions

[Eric S Fraga]

On Friday, 29 Apr 2022 at 11:26, Eric Abrahamsen wrote:
> Sounds like some kind of bug, then. I'd check the github issues (several
> of the closed issues are about super capes) and maybe open a new one.

This was a good suggestion and I'm pursuing some of the suggestions in
the issues.  If/when I get a solution, I'll post here for completeness.

Thank you,
eric

-- 
Eric S Fraga with org 9.5.3 in Emacs 29.0.50 on Debian 11.3