Re: [PATCH] `completing-read`: Add `group-function` support to completion metadata (REVISED PATCH VERSION 2)

[Juri Linkov]

>> Aren't affixations guarded by completions-detailed
>> and grouping guarded by completions-group orthogonal?
>> Then it should be fine to have separate options.
>
> Yes, they are orthogonal. I agree we can keep it as is. I still wanted
> to point it out. And note the difference - the completions-group guard
> variable is checked in minibuffer.el, while the completions-detailed
> variable is checked in the completion table in help-fns.el. Maybe
> completions-detailed should also be checked in the minibuffer.el instead?

I think both completions-detailed and completions-group should be
checked only by the API user like in help-fns.el.  Otherwise,
there is duplication that you can see in my previous patch
for read-char-by-name:

  (let* ((enable-recursive-minibuffers t)
         (completion-ignore-case t)
         (completion-tab-width 4)
         (completions-group read-char-by-name-group)
                            =======================
         (input
          (completing-read
           prompt
           (lambda (string pred action)
             (if (eq action 'metadata)
                 `(metadata
                   (group-function
                    . ,(when read-char-by-name-group
                             =======================
                         #'mule--ucs-names-group))

The same user option read-char-by-name-group is checked twice.
It should suffice to leave only the latter.