Re: [elpa] externals/vertico a962a8b 2/3: Implement deferred highlighting for all completion styles

[Daniel Mendler <mail@daniel-mendler.de>]

Hi Stefan!

On 4/11/21 6:37 PM, Stefan Monnier wrote:
>> +(declare-function orderless-highlight-matches "ext:orderless")
>> +(defun vertico--all-completions (&rest args)
>> +  "Compute all completions for ARGS with deferred highlighting."
>> +  (cl-letf* ((orig-pcm (symbol-function #'completion-pcm--hilit-commonality))
>> +             (orig-flex (symbol-function #'completion-flex-all-completions))
>> +             ((symbol-function #'completion-flex-all-completions)
>> +              (lambda (&rest args)
>> +                ;; Unfortunately for flex we have to undo the deferred highlighting, since flex uses
>> +                ;; the completion-score for sorting, which is applied during highlighting.
>> +                (cl-letf (((symbol-function #'completion-pcm--hilit-commonality) orig-pcm))
>> +                  (apply orig-flex args))))
>> +             ;; Defer the following highlighting functions
>> +             (hl #'identity)
>> +             ((symbol-function #'completion-hilit-commonality)
>> +              (lambda (cands prefix &optional base)
>> +                (setq hl (lambda (x) (nconc (completion-hilit-commonality x prefix base) nil)))
>> +                (and cands (nconc cands base))))
>> +             ((symbol-function #'completion-pcm--hilit-commonality)
>> +              (lambda (pattern cands)
>> +                (setq hl (lambda (x) (completion-pcm--hilit-commonality pattern x)))
>> +                cands))
>> +             ((symbol-function #'orderless-highlight-matches)
>> +              (lambda (pattern cands)
>> +                (setq hl (lambda (x) (orderless-highlight-matches pattern x)))
>> +                cands)))
>> +    (cons (apply #'completion-all-completions args) hl)))
> 
> Yuck!
> I think this should have a bug#NNN attached to make sure it's only used
> on a temporary basis (i.e. until Emacs's minibuffer.el provides the
> corresponding functionality).

I wasn't aware that you are monitoring this so closely. This is 
basically the prototype of how the deferred highlighting could look 
like, instead of my proposal 8, `completion-skip-highlighting`, which 
can hardly be generalized. It works well for Orderless 
(`orderless-skip-highlighting`), but that's about it.

One can introduce a function 
`completion-all-completions-deferred-highlighting` or alternatively add 
an optional argument to `completion-all-completions`, which signals that 
the function should return the list of completions and the deferred 
highlighting function/highlighting data. (One may also consider 
returning the base directly or is there a particular reason to put it 
into the last cdr?)

The function `completion-all-completions-deferred-highlighting` can 
return a special highlighting function, registered in 
`completion-styles-alist`.

Daniel