Re: [elpa] master 567ea7e 1/3: Ivy-mode now works better with `find-file'

Re: [elpa] master 567ea7e 1/3: Ivy-mode now works better with `find-file'

[Stefan Monnier]

> +  (cond ((or (functionp collection)
> +             (vectorp collection))
> +         (when (eq collection 'read-file-name-internal)
> +           (setq ivy--directory default-directory)
> +           (setq initial-input nil))
> +         (setq collection (all-completions "" collection predicate)))
> +        ((hash-table-p collection)
> +         (error "Hash table as a collection unsupported"))
> +        ((listp (car collection))
> +         (setq collection (all-completions "" collection predicate))))

I suggest you also try

   C-h i
   g (em TAB


-- Stefan

Re: [elpa] master 567ea7e 1/3: Ivy-mode now works better with `find-file'

[Oleh Krehel]

On Fri, Apr 17, 2015 at 4:49 PM, Stefan Monnier
<monnier@iro.umontreal.ca> wrote:
>> +  (cond ((or (functionp collection)
>> +             (vectorp collection))
>> +         (when (eq collection 'read-file-name-internal)
>> +           (setq ivy--directory default-directory)
>> +           (setq initial-input nil))
>> +         (setq collection (all-completions "" collection predicate)))
>> +        ((hash-table-p collection)
>> +         (error "Hash table as a collection unsupported"))
>> +        ((listp (car collection))
>> +         (setq collection (all-completions "" collection predicate))))
>
> I suggest you also try
>
>    C-h i
>    g (em TAB

Info-read-node-name-1 takes STRING, PREDICATE, and CODE and says to
look in the `completing-read' doc.  But there's nothing abount CODE
there. Where do I get it?

Oleh

Re: [elpa] master 567ea7e 1/3: Ivy-mode now works better with `find-file'

[Stefan Monnier]

> Info-read-node-name-1 takes STRING, PREDICATE, and CODE and says to
> look in the `completing-read' doc.  But there's nothing abount CODE
> there. Where do I get it?

I think you're better off reading about it in the Elisp manual.
The third argument (called ACTION usually) is basically the *method* to
use on the "object implemented as a function".  It can be nil (which
means `try-completion'), t (which means `all-completions'), lambda
(which means `test-completion'), or `boundaries', or `metadata'.


        Stefan

Re: [elpa] master 567ea7e 1/3: Ivy-mode now works better with `find-file'

[Oleh Krehel]

Stefan Monnier <monnier@IRO.UMontreal.CA> writes:

>> Info-read-node-name-1 takes STRING, PREDICATE, and CODE and says to
>> look in the `completing-read' doc.  But there's nothing abount CODE
>> there. Where do I get it?
>
> I think you're better off reading about it in the Elisp manual.
> The third argument (called ACTION usually) is basically the *method* to
> use on the "object implemented as a function".  It can be nil (which
> means `try-completion'), t (which means `all-completions'), lambda
> (which means `test-completion'), or `boundaries', or `metadata'.

Seems very complex for just returning a (short) list of strings.
Why doesn't this work for `Info-read-node-name-1':

    (setq collection (all-completions "" collection predicate))

while this works within the topic, but not in the directory:

    (setq collection (funcall collection "" nil t))

If I understood the info page correctly, both should be equivalent.

Oleh

Re: [elpa] master 567ea7e 1/3: Ivy-mode now works better with `find-file'

[Stefan Monnier]

> Seems very complex for just returning a (short) list of strings.

The set of files or the set of Info nodes (which can appear in any file)
can't conveniently be treated as "a (short) list of strings".

> Why doesn't this work for `Info-read-node-name-1':
>     (setq collection (all-completions "" collection predicate))

I don't understand the question.  AFAIK it does work.  It just doesn't
return every possible node name, since there's pretty much an infinite
supply (if you include those that can be accessed via Tramp, for example).

>     (setq collection (funcall collection "" nil t))

Indeed (funcall collection "" pred t) should be equivalent to
(all-completions "" collection predicate) when `collection' is a function.
If you see otherwise, we have a problem.


        Stefan

Re: [elpa] master 567ea7e 1/3: Ivy-mode now works better with `find-file'

[Oleh Krehel]

Stefan Monnier <monnier@IRO.UMontreal.CA> writes:

>> Seems very complex for just returning a (short) list of strings.
>
> The set of files or the set of Info nodes (which can appear in any file)
> can't conveniently be treated as "a (short) list of strings".
>
>> Why doesn't this work for `Info-read-node-name-1':
>>     (setq collection (all-completions "" collection predicate))
>
> I don't understand the question.  AFAIK it does work.  It just doesn't
> return every possible node name, since there's pretty much an infinite
> supply (if you include those that can be accessed via Tramp, for example).
>
>>     (setq collection (funcall collection "" nil t))
>
> Indeed (funcall collection "" pred t) should be equivalent to
> (all-completions "" collection predicate) when `collection' is a function.
> If you see otherwise, we have a problem.

You can try the current ivy-mode code in elpa. It uses this for functions:

    (all-completions "" collection predicate)

And it does't work with Info. I tried also with helm-mode, it doesn't
work either.

Oleh

Re: [elpa] master 567ea7e 1/3: Ivy-mode now works better with `find-file'

[Stefan Monnier]

>> Indeed (funcall collection "" pred t) should be equivalent to
>> (all-completions "" collection predicate) when `collection' is a function.
>> If you see otherwise, we have a problem.
> You can try the current ivy-mode code in elpa. It uses this for functions:
>     (all-completions "" collection predicate)
> And it does't work with Info.

That doesn't seem related to whether it gives the same result as

   (funcall collection "" predicate t)

I'm not surprised "it doesn't work" (which of course, is poor as
a description of problem) if you expect this to give you the (infinite)
list of possible node named.

> I tried also with helm-mode, it doesn't work either.

Not surprised either.

There's a good reason why icomplete.el doesn't try to re-implement
completion, but instead piggybacks on top of the default
completion code.


        Stefan

Re: [elpa] master 567ea7e 1/3: Ivy-mode now works better with `find-file'

[Thierry Volpiatto]

Stefan Monnier <monnier@IRO.UMontreal.CA> writes:

>> Seems very complex for just returning a (short) list of strings.
>
> The set of files or the set of Info nodes (which can appear in any file)
> can't conveniently be treated as "a (short) list of strings".
>
>> Why doesn't this work for `Info-read-node-name-1':
>>     (setq collection (all-completions "" collection predicate))
>
> I don't understand the question.  AFAIK it does work.  It just doesn't
> return every possible node name, since there's pretty much an infinite
> supply (if you include those that can be accessed via Tramp, for example).
>
>>     (setq collection (funcall collection "" nil t))
>
> Indeed (funcall collection "" pred t) should be equivalent to
> (all-completions "" collection predicate) when `collection' is a function.
> If you see otherwise, we have a problem.

This is working, but the problem is that functions are made to return
the completion part of what you have in minibuffer, and this is
impossible to handle from 3rd tools.
Take the example of find-file-at-point, it returns a list of relative
file names, what should we call to return the same list prefixed by what
is in minibuffer ?

-- 
Thierry
Get my Gnupg key:
gpg --keyserver pgp.mit.edu --recv-keys 59F29997 

Re: [elpa] master 567ea7e 1/3: Ivy-mode now works better with `find-file'

[Stefan Monnier]

> This is working, but the problem is that functions are made to return
> the completion part of what you have in minibuffer, and this is
> impossible to handle from 3rd tools.
> Take the example of find-file-at-point, it returns a list of relative
> file names, what should we call to return the same list prefixed by what
> is in minibuffer ?

I have the impression that the question is aimed in my direction, but
I don't understand it enough to answer.  E.g. I don't understand what
you mean by "find-file-at-point returns a list of relative
file names" and how it's relevant to the discussion.

Maybe the answer you're looking for is something like:

   (pcase-let ((`(,start . ,_)
                (completion-boundaries <str> <collection> <pred> ""))
               (compls (all-completions <str> <collection> <pred>)))
     (if (zerop start) compls
       (let ((prefix (substring <str> 0 start)))
         (mapcar (lambda (compl) (concat prefix compl)) compls))))


-- Stefan