RE: new-flex-completion-style (was: [Emacs-diffs] scratch/ 2c75775 2/2: Score, sort and annotate flex-style completions according to match tightness)

[Drew Adams]

FWIW -

I guess you're arguing for an ability to sort completion candidates at multiple 
places/levels, including far upstream from interactive use.

My concern/interest is more about interactive use, where sorting can usefully 
interact with progressive filtering.  I like late binding. ;-)

In my context (Icicles) users can filter completion candidates incrementally in 
various ways, and sorting can act either on "raw" completion candidates (e.g. 
alist entries, including their cdrs) or on displayed candidates (e.g. text in 
*Completions*).

Unlike vanilla Emacs, the same sort order applies (intentionally) for 
displaying matching completions and for cycling among those candidates.

Users can switch among any number of sort orders (predicates) interactively.

The default set of sort orders available for this switching is defined by a 
user option.  But the code context (e.g. code that calls `completing-read') can 
change this set.  So different commands can make different sets of sort orders 
available.

Users can define new sort orders with a macro, which adds them to those made 
available by the user option.

Maybe some such capabilities would be relevant to your context; dunno.

https://www.emacswiki.org/emacs/Icicles_-_Sorting_Candidates