RE: `completion-in-region'

["Drew Adams" <drew.adams@oracle.com>]

> > TextMate's default matching (it has no name AFAIK) is the same as
> > Ido's "flex" matching and Icicles's "scatter" matching. It is
> > a poor-man's fuzzy matching (not fuzzy matching in the more usual
> > sense). (TextMate uses this matching only for file names, I 
> > believe.)
> 
> I see that ido implements it by turning "abc" into the regexp
> ".*a.*b.*c".  But matching this regexp against a string like
> "abababababab" takes time O(N^3) where N is the length of the 
> completion candidate, which makes me a bit uneasy (and in
> general matching "flex"ibly an input string of size M against
> a completion candidate of size N can take time O(N^M)).  But
> I guess experience shows it's not a real problem.
> 
> It should be pretty easy to add it as a new completion-style 
> (after all you can already get the same result with partial
> completion by typing "*a*b*c").

Yes, to all you said.

The question is whether matching like this should be the default. IMO, no.