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 02/11/2023 11:48, João Távora wrote:
> On Wed, Nov 1, 2023 at 10:45 PM Dmitry Gutov <dmitry@gutov.dev> wrote:
>
> > > If orderless (which I've never tried), does some kind of scoring of
> > > completions, it probably also needs the same complications of flex.
> >
> > Turns out, Orderless doesn't do any scoring or sorting. Only filtering.
>
> Interesting, so if I M-x d i f f with orderless I don't get results in
> any particular order?

Right.

And input "diff" actually results in prefix-only matching. But if the 
input contains space (or any other pre-configured delimiter character), 
then it's translated to a set of substring matches, performed in any 
order. So "diff mode" does not translate to ".*diff.*mode", but a more 
complex regexp with all permutations.

IOW, this whole approach results in stricter matching with fewer 
results, so a smarter sort isn't that necessary.

> > > > > > Anyway, have you looked into what it would take to solve it?
> > > > >
> > > > > No, naively, I just think it's a similar situation of display and business
> > > > > logic being mixed up.  Presumably the quoted stuff is just for insertion
> > > > > (and display?), and the unquoted stuff is what patterns/scoring should
> > > > > operate on.
> > > >
> > > > Apparently it's good for insertion, but according to that comment inside
> > > > the function, the unquoted stuff might actually be better for display.
> > >
> > > No idea what the unquoted stuff is for, so I haven't really tested it.
> >
> > A couple of scenarios:
>
> Thanks.  Then I think it is working OK, but it would be safer if you were
> to double-check yourself, as I really never use this functionality.

They look okay, yes. I imagine Daniel also tested the 
'completion--unquoted' solution before offering his patches.

> I addressed all your docstring suggestions, fixed a bug and significantly
> simplified the code in the latest version of the patch.  I also
> removed the instrumentation in icomplete.el.  Patch attached here
> and pushed to feature/completion-lazy-hilit.

Thank you. I don't see that the bug was, but deferring the scoring even 
in the eager highlighting case looks sensible (why not, indeed, if the 
performance is no worse).