Re: [PATCH] `completing-read`: Add `group-function` support to completion metadata

[Stefan Monnier]

>>> So it's better if at least icomplete-mode is enabled, preferably with an
>>> option which shows the completions right away with no input. And the
>>> vertical style should be even better.
>> Maybe we could accommodate this in the default UI by adding an option to
>> eagerly popup the *Completions* buffer (and keep it updated as long as
>> it's displayed) and to keep it disabled by default but make it possible
>> to enable it via completion-category-(overrides|defaults).
> I don't know if it's enough. And I have my doubts (expressed previously
> already) about a *Completions* buffer that is impossible to dismiss: it
> might hide a window showing a buffer you care about.

I haven't looked at it closely, but I assume it shouldn't be too hard
to tweak the current code such that we can refrain from popping down the
*Completions* automatically, while still allowing the user to pop it down.

> Ultimately, it can be a good change/feature, but it should be championed by
> somebody actually willing to use the resulting UI on a regular basis.

Of course.

> My ideal UI would probably be more like this:
> https://emacs-lsp.github.io/lsp-ui/images/lsp-xref.gif (though perhaps with
> a slightly different layout). Meaning, a posframe popup showing the list of
> all locations, vertically, and a preview popup by its side or below it,
> displaying the current selected location.

I was thinking only of tweaking the existing UI to better support the
case of "selection" compared to "completion".  What you suggest here is
nice and it goes much further but I don't think the two are competing.

If anything they work together: the idea behind my tweak would be to let
more callers of `completing-read` presume a "selection" kind of UI
(without having to impose a non-default UI), which in turn will make it
more often useful/beneficial to use a UI like the one you propose.


        Stefan