Added an O(n log n) version of nub

The new nub is called nub', and it replaces the old sortNub which was 
not lazy and did not retain the order between the elements.

src/compiler/GF/Compile/GeneratePMCFG.hs

@@ -22,7 +22,7 @@ import GF.Grammar.Lookup
 import GF.Grammar.Predef
 import GF.Data.BacktrackM
 import GF.Data.Operations
-import GF.Data.Utilities (updateNthM, updateNth, sortNub)
+import GF.Data.Utilities (updateNthM, updateNth)
 
 import System.IO
 import qualified Data.Map as Map

src/compiler/GF/Data/Utilities.hs

@@ -17,6 +17,7 @@ module GF.Data.Utilities where
 import Data.Maybe
 import Data.List
 import Control.Monad (MonadPlus(..),liftM)
+import qualified Data.Set as Set
 
 -- * functions on lists
 
@@ -67,15 +68,18 @@ safeInit :: [a] -> [a]
 safeInit [] = []
 safeInit xs = init xs
 
--- | Like 'nub', but more efficient as it uses sorting internally.
+-- | Like 'nub', but O(n log n) instead of O(n^2), since it uses a set to lookup previous things.
-sortNub :: Ord a => [a] -> [a]
+--   The result list is stable (the elements are returned in the order they occur), and lazy.
-sortNub = map head . group . sort
+--   Requires that the list elements can be compared by Ord.
+--   Code ruthlessly taken from http://hpaste.org/54411
+nub' :: Ord a => [a] -> [a]
+nub' = loop Set.empty
+    where loop _    []            = []
+          loop seen (x : xs)
+              | Set.member x seen = loop seen xs
+              | otherwise         = x : loop (Set.insert x seen) xs
 
--- | Like 'nubBy', but more efficient as it uses sorting internally.
+-- | Sorts and then groups elements given an ordering of the 
-sortNubBy :: (a -> a -> Ordering) -> [a] -> [a]
-sortNubBy f = map head . sortGroupBy f
-
--- | Sorts and then groups elements given and ordering of the 
 --   elements.
 sortGroupBy :: (a -> a -> Ordering) -> [a] -> [[a]]
 sortGroupBy f = groupBy (compareEq f) . sortBy f

src/compiler/GF/Speech/RegExp.hs

@@ -53,7 +53,7 @@ isEpsilon (REConcat []) = True
 isEpsilon _ = False
 
 unionRE :: Ord a => [RE a] -> RE a
-unionRE = unionOrId . sortNub . concatMap toList 
+unionRE = unionOrId . nub' . concatMap toList 
   where 
     toList (REUnion xs) = xs
     toList x = [x]

src/runtime/haskell/PGF/Optimize.hs

@@ -20,7 +20,6 @@ import qualified Data.IntMap as IntMap
 import qualified GF.Data.TrieMap as TrieMap
 import qualified Data.List as List
 import Control.Monad.ST
-import GF.Data.Utilities(sortNub)
 
 optimizePGF :: PGF -> PGF
 optimizePGF pgf = pgf{concretes=fmap (updateConcrete (abstract pgf) .