bug#42149: Substring and flex completion ignore implicit trailing ‘any’

["João Távora" <joaotavora@gmail.com>]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> Regardless, if you don't want to touch that funciton, I understand, it
>> is used in more places than just completion-pcm--hilit-commonality,
>> which really should be called
>>
>>    completion--given-that-we-know-this-matches-tell-me-where-and-how-well
>
> Hmm... until someone™ added scoring to it, this function did nothing
> more than add faces to highlight the common parts and the "first
> difference".  So I'd suggest you take it up with that someone ;-)

:-)

Grumblebgrumpbl.  I kinda did, and so I added a docstring to it (have a
read).  Anyway, what's suprising about this function is that this
re-matches PATTERN to each of COMPLETIONS which is very odd to the
reader, becasue it also asserts that PATTERN already matches
COMPLETIONS.  Why are we regexp-matching twice?! -- asks the poor soul
reading this.

This comes down to completion-regexp-list being used by
Fall_completions() directly.  If that C function recorded the match data
in all the lisp strings it was passed, then completion--given-that...
would be easier follow.  And likely faster.  Though I haven't measured
the impact, sparing a regexp match against each completion might be
worth it in terms of responsiveness, especially in flex and similar
methods.

João