On Sat, May 4, 2013 at 4:58 PM, Óscar Fuentes <ofv@wanadoo.es> wrote:

> This change introduces a serious slowdown which is noticeable for large
> candidate lists (try with 10000 elements.) The slowdown happens on every
> invocation.
>
> It is obvious that having duplicate candidates makes no sense, but at
> the same time scanning the list in advance for all duplicates is
> expensive.


The only way to introduce list with dupes is ido-completing-read (i.e. it's
not an issue for files and buffers), so I think it's okay to remove dupes
just once on entry.

Patch atached.


-- 
Le