bug#47711: bug#48841: bug#47711: bug#48841: bug#47711: [PATCH VERSION 2] Add new `completion-filter-completions` API and deferred highlighting

[Dmitry Gutov]

On 08/11/2023 03:24, João Távora wrote:
> On Wed, Nov 8, 2023 at 1:06 AM Dmitry Gutov <dmitry@gutov.dev> wrote:
> > On 07/11/2023 14:13, João Távora wrote:
> > > Any objections?  Seems to speed it up when flex is the preferred
> > > completion style outside icomplete.  These are average times
> > > collected from the instrumentation of the above
> > > completion-all-completions when doing a M-: (setq i TAB)
> > >
> > > I just used my normal Emacs session for this.
> > >
> > > with lazy hilit:      0.104536125
> > > without lazy hilit:   0.172522571
> >
> > IIUC the problem with the default completion-at-point UI here is that is
> > prints all completions anyway, in the buffer *Completions*. And so it
> > applies syntax highlighting to them as well, and does that eagerly (as
> > opposed to e.g. doing that via jit-lock).
> >
> > If you instrumented only the 'completion-all-completions' call, then
> > that might miss the subsequent time spent in sorting.
>
> You probably mean highlighting: that's the saving being made here,
> not sorting.

I meant both but I forgot that we moved scoring to later in both cases.

> Anyway, it _felt_ snappier, but maybe I was dreaming.  Got any
> better suggestions for places where to place `benchmark-progn`?

Perhaps around the whole minibuffer-completion-help.

> Also, I don't think *Completions* has _every_ matching completion,
> does it?  Doesn't it display more completions as you keep TABing?
> That's what I supposed was providing the speedup.

Well, I type (d, press C-M-i, then select the window showing 
*Completions* and scroll to the end of the buffer. The last completion 
displayed there starts with "dy", so that's probably all of them.

That's with the default completion-styles config, but with flex it's the 
same: all completions shown in the buffer. Sorted and highlighted.