Rodrigo Queiro overdrigzed at gmail.com
Fri May 18 18:56:48 EDT 2007

```group collects equal elements in sublists. Thus, unique can be implemented
as:
unique = map head . group
i.e. taking the first of each group of equal elements.

(Yet) another way of doing this, is a modified quicksort:

qsort [] = []
qsort (x:xs) = qsort (filter (<x) xs) ++ [x] ++ qsort (filter (>x) xs)

At each step, this will ignore all elements equal to the pivot value.

On 18/05/07, Dan Weston <westondan at imageworks.com> wrote:
>
> OK, I looked up "group" and didn't see any Ord constraint analog. I give
> up. What is the common idiom? I was stupidly using nub even though I
> already have an Ord constraint. I have rewritten it to:
>
> unique :: Ord a => [a] -> [a]
> unique (x:y:z) = (if x < y then (x:) else id) (unique (y:z))
> unique xyz     = xyz
>
> uniqueSort = unique . sort
>
> but I would much rather use the "common idiom" than this dreck.
>
> Help a poor guy out? :)
>
> Dan
>
> Paul Johnson wrote:
> > Andrew Coppin wrote:
> >> It occurs to me that if you want a sorted list of only unique
> >> elements, it would seem (to me) to be efficient to do the sorting and
> >> the uniquing at the same time. Does any library function do this? (I
> >> imagine it wouldn't be hard to write it yourself...)
> > Yes, although it only works on instances of Ord then, because of the
> > sorting.
> >
> > Its actually quite a common idiom, and worth figuring out for yourself.
> > Hint: look at "group".
> >
> > Paul.
> > _______________________________________________
> >
> >
>
>
> _______________________________________________