Skip to content

Embed URL


Subversion checkout URL

You can clone with
Download ZIP
First pass at Haskell port of glibc qsort.c
module QSort1 ( qsort, quicksort, main, prop_qsort ) where
-- Ported from glibc's qsort.c
-- Copyright (C) 1991-2014 Free Software Foundation, Inc.
-- This file is part of the GNU C Library.
-- Written by Douglas C. Schmidt (
-- The GNU C Library is free software; you can redistribute it and/or
-- modify it under the terms of the GNU Lesser General Public
-- License as published by the Free Software Foundation; either
-- version 2.1 of the License, or (at your option) any later version.
-- The GNU C Library is distributed in the hope that it will be useful,
-- but WITHOUT ANY WARRANTY; without even the implied warranty of
-- Lesser General Public License for more details.
-- You should have received a copy of the GNU Lesser General Public
-- License along with the GNU C Library; if not, see
-- <>.
-- If you consider tuning this algorithm, you should consult first:
-- Engineering a sort function; Jon Bentley and M. Douglas McIlroy;
-- Software - Practice and Experience; Vol. 23 (11), 1249-1265, 1993.
import qualified Data.Vector as V
import qualified Data.Vector.Mutable as MV
import Data.Vector.Mutable ( IOVector )
import Control.Monad ( when )
import qualified Data.IORef as I
import Control.Applicative ( (<$>), (<*>) )
import Data.List ( sort )
import Test.QuickCheck ( Property, quickCheck )
import Test.QuickCheck.Monadic ( monadicIO, run, assert )
main :: IO ()
main = quickCheck prop_qsort
prop_qsort :: [Int] -> Property
prop_qsort xs = monadicIO $ do
xs' <- run (qsort xs)
assert (xs' == sort xs)
-- | Discontinue quicksort algorithm when partition gets below this size.
-- This particular magic number was chosen to work best on a Sun 4/260.
maxThresh :: Int
maxThresh = 4
qsort :: (Ord a) => [a] -> IO [a]
qsort = go . V.fromList
go v = do
mv <- V.unsafeThaw v
quicksort mv compare
V.toList <$> V.unsafeFreeze mv
-- Order size using quicksort. This implementation incorporates
-- four optimizations discussed in Sedgewick:
-- 1. Non-recursive, using an explicit stack of pointer that store the
-- next array partition to sort. To save time, this maximum amount
-- of space required to store an array of SIZE_MAX is allocated on the
-- stack. Assuming a 32-bit (64 bit) integer for size_t, this needs
-- only 32 * sizeof(stack_node) == 256 bytes (for 64 bit: 1024 bytes).
-- Pretty cheap, actually.
-- 2. Chose the pivot element using a median-of-three decision tree.
-- This reduces the probability of selecting a bad pivot value and
-- eliminates certain extraneous comparisons.
-- 3. Only quicksorts TOTAL_ELEMS / MAX_THRESH partitions, leaving
-- insertion sort to order the MAX_THRESH items within each partition.
-- This is a big win, since insertion sort is faster for small, mostly
-- sorted array segments.
-- 4. The larger of the two sub-partitions is always pushed onto the
-- stack first, with the algorithm then concentrating on the
-- smaller partition. This *guarantees* no more than log (total_elems)
-- stack size is needed (actually O(1) in this case)!
quicksort :: IOVector a -- ^ vector to sort
-> (a -> a -> Ordering) -- ^ comparator
-> IO ()
quicksort v cmp
| totalElems == 0 =
-- Avoid lossage with unsigned arithmetic below.
return ()
| otherwise = do
when (totalElems > maxThresh) $ do
lo <- alloc 0
hi <- alloc endVal
top <- alloc [(0, 0)]
while (stackNotEmpty top) $ do
-- Select median value from among LO, MID, and HI. Rearrange
-- LO and HI so the three values are sorted. This lowers the
-- probability of picking a pathological pivot value and
-- skips a comparison for both the LEFT_PTR and RIGHT_PTR in
-- the while loops.
mid <- getPair lo hi >>= \(l, h) -> alloc (l + (h - l) `div` 2)
ifM (lt mid lo) (swap mid lo)
ifM (lt hi mid) $ do
swap mid hi
ifM (lt mid lo) (swap mid lo)
leftPtr <- allocM (succ <$> get lo)
rightPtr <- allocM (pred <$> get hi)
-- Here's the famous ``collapse the walls'' section of quicksort.
-- Gotta like those tight inner loops! They are the main reason
-- that this algorithm runs much faster than others.
doWhile (getOp (<=) leftPtr rightPtr) $ do
while (lt leftPtr mid) $
inc leftPtr
while (lt mid rightPtr) $
dec rightPtr
ifElseM (getOp (<) leftPtr rightPtr)
swap leftPtr rightPtr
ifElseM (getOp (==) mid leftPtr)
(putM mid (get leftPtr))
(ifM (getOp (==) mid rightPtr) $
putM mid (get rightPtr))
inc leftPtr
dec rightPtr)
(ifM (getOp (==) leftPtr rightPtr) $ do
inc leftPtr
dec rightPtr)
-- Set up pointers for next iteration. First determine whether
-- left and right partitions are below the threshold size. If so,
-- ignore one or both. Otherwise, push the larger partition's
-- bounds on the stack and continue sorting the smaller one.
loVal <- get lo
hiVal <- get hi
leftVal <- get leftPtr
rightVal <- get rightPtr
case (rightVal - loVal, hiVal - leftVal) of
(loSize, hiSize)
| loSize <= maxThresh ->
if hiSize <= maxThresh
-- Ignore both small partitions.
then pop top lo hi
-- Ignore small left partition.
else put lo leftVal
| hiSize <= maxThresh ->
-- Ignore small right partition.
put hi rightVal
| loSize > hiSize -> do
-- Push larger left partition indices.
push top (loVal, rightVal)
put lo leftVal
| otherwise -> do
-- Push larger right partition indices.
push top (leftVal, hiVal)
put hi rightVal
-- Once the BASE_PTR array is partially sorted by quicksort the rest
-- is completely sorted using insertion sort, since this is efficient
-- for partitions below MAX_THRESH size. BASE_PTR points to the beginning
-- of the array to sort, and END_PTR points at the very last element in
-- the array (*not* one beyond it!).
tmpPtr <- alloc 0
basePtr <- alloc 0
-- Find smallest element in first threshold and place it at the
-- array's beginning. This is the smallest array element,
-- and the operation speeds up insertion sort's inner loop.
runPtr <- alloc 1
while ((<= min endVal maxThresh) <$> get runPtr) $ do
ifM (lt runPtr tmpPtr) $
putM tmpPtr (get runPtr)
inc runPtr
ifM (getOp (/=) tmpPtr basePtr) $
swap tmpPtr basePtr
-- Insertion sort, running from left-hand-side up to right-hand-side.
put runPtr 1
while (inc runPtr >> (<= endVal) <$> get runPtr) $ do
putM tmpPtr (pred <$> get runPtr)
while (lt runPtr tmpPtr) $
dec tmpPtr
inc tmpPtr
ifM (getOp (/=) tmpPtr runPtr) $ do
lo <- allocM (get runPtr)
hi <- alloc 0
while (putM hi (get lo) >> dec lo >> getOp (>=) lo tmpPtr) $
swap lo hi
totalElems = MV.length v
endVal = totalElems - 1
-- define a DSL for C-like programming in Haskell
swapPure = MV.swap v
swap aVar bVar = do
a <- get aVar
b <- get bVar
swapPure a b
alloc = I.newIORef
get = I.readIORef
allocM = (alloc =<<)
getPair aVar bVar = (,) <$> get aVar <*> get bVar
getAt var = get var >>= v
getOp op aVar bVar = do
a <- get aVar
b <- get bVar
return (op a b)
put varName x = I.writeIORef varName $! x
putM varName a = a >>= put varName
modify varName = I.modifyIORef' varName
inc varName = modify varName succ
dec varName = modify varName pred
lt aVar bVar = do
res <- cmp <$> getAt aVar <*> getAt bVar
return (LT == res)
ifM c a = c >>= \x -> when x a
ifElseM c aThen aElse = c >>= \x -> when x aThen >> when (not x) aElse
while c a = ifM c (a >> while c a)
doWhile c a = a >> while c a
-- Define a stack abstraction, similar to the macros in C but without
-- contiguous storage for simplicity
push stack = modify stack . (:)
pop stack low high = do
(a, b):rest <- get stack
put low a
put high b
put stack rest
stackNotEmpty stack =
not . null <$> get stack
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.