Teach completion to insert a common prefix

Teach completion to insert a common prefix

[Spencer Baugh]

Example in emacs -Q:

C-h v -matching-lines- TAB

All possible completions for -matching-lines- start with
list-matching-lines-.  So there would be no harm in inserting "list" at
the start of the completion input.  But Emacs does not do this.

Is it possible to teach Emacs completion to insert a common prefix if
there is one, like in this situation?

Two reasons:

- It would look nicer

- It would make completion work better if the desired completion
alternative is itself a prefix of some other alternative.  For example,
if both list-matching-lines-buffer and list-matching-lines-buffer-face
are available completion alternatives, then it is hard to use completion
to input list-matching-lines-buffer, since it only completes to
-matching-lines-buffer which is not an actual alternative.

This is *not* a matter of completion-styles, variable completion already
uses the substring completion style, otherwise -matching-lines- wouldn't
even validly complete to list-matching-lines-something.

This is purely a matter of teaching completion to eagerly insert the
common prefix when there is one.

Is this possible?

Re: Teach completion to insert a common prefix

[Yuri Khan]

On Tue, 8 Aug 2023 at 03:39, Spencer Baugh <sbaugh@janestreet.com> wrote:

> Example in emacs -Q:
>
> C-h v -matching-lines- TAB
>
> All possible completions for -matching-lines- start with
> list-matching-lines-.  So there would be no harm in inserting "list" at
> the start of the completion input.  But Emacs does not do this.

It does do that if you put the point at the start of line.

Re: Teach completion to insert a common prefix

[Pierre Rouleau]

> All possible completions for -matching-lines- start with
> > list-matching-lines-.  So there would be no harm in inserting "list" at
> > the start of the completion input.  But Emacs does not do this.
>
> It does do that if you put the point at the start of line.
>
> Note that older versions of Emacs was also doing that even when point was
at the end of the line, as in Emacs 26.3

Do you know when that changed and why?

-- 
/Pierre

Re: Teach completion to insert a common prefix

[sbaugh]

Pierre Rouleau <prouleau001@gmail.com> writes:

>> All possible completions for -matching-lines- start with
>> > list-matching-lines-.  So there would be no harm in inserting "list" at
>> > the start of the completion input.  But Emacs does not do this.
>>
>> It does do that if you put the point at the start of line.
>>
>> Note that older versions of Emacs was also doing that even when point was
> at the end of the line, as in Emacs 26.3
>
> Do you know when that changed and why?

I think this is just a bug, maybe introduced in Emacs 26.  I fixed it in
bug#65137: 29.1; completion-substring-try-completion doesn't return the
longest common substring.