packages feed

vector-0.13.2.0: src/Data/Vector/Generic/Mutable.hs

{-# LANGUAGE BangPatterns #-}
{-# LANGUAGE CPP #-}
{-# LANGUAGE FlexibleContexts #-}
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE ScopedTypeVariables #-}
{-# LANGUAGE TypeFamilies #-}
-- |
-- Module      : Data.Vector.Generic.Mutable
-- Copyright   : (c) Roman Leshchinskiy 2008-2010
--                   Alexey Kuleshevich 2020-2022
--                   Aleksey Khudyakov 2020-2022
--                   Andrew Lelechenko 2020-2022
-- License     : BSD-style
--
-- Maintainer  : Haskell Libraries Team <libraries@haskell.org>
-- Stability   : experimental
-- Portability : non-portable
--
-- Generic interface to mutable vectors.

module Data.Vector.Generic.Mutable (
  -- * Class of mutable vector types
  MVector(..),

  -- * Accessors

  -- ** Length information
  length, null,

  -- ** Extracting subvectors
  slice, init, tail, take, drop, splitAt,
  unsafeSlice, unsafeInit, unsafeTail, unsafeTake, unsafeDrop,

  -- ** Overlapping
  overlaps,

  -- * Construction

  -- ** Initialisation
  new, unsafeNew, replicate, replicateM, generate, generateM, clone,

  -- ** Growing
  grow, unsafeGrow,
  growFront, unsafeGrowFront,

  -- ** Restricting memory usage
  clear,

  -- * Accessing individual elements
  read, readMaybe, write, modify, modifyM, swap, exchange,
  unsafeRead, unsafeWrite, unsafeModify, unsafeModifyM, unsafeSwap, unsafeExchange,

  -- * Folds
  mapM_, imapM_, forM_, iforM_,
  foldl, foldl', foldM, foldM',
  foldr, foldr', foldrM, foldrM',
  ifoldl, ifoldl', ifoldM, ifoldM',
  ifoldr, ifoldr', ifoldrM, ifoldrM',

  -- * Modifying vectors
  nextPermutation, nextPermutationBy,
  prevPermutation, prevPermutationBy,

  -- ** Filling and copying
  set, copy, move, unsafeCopy, unsafeMove,

  -- * Internal operations
  mstream, mstreamR,
  unstream, unstreamR, vunstream,
  munstream, munstreamR,
  transform, transformR,
  fill, fillR,
  unsafeAccum, accum, unsafeUpdate, update, reverse,
  unstablePartition, unstablePartitionBundle, partitionBundle,
  partitionWithBundle,
  -- * Re-exports
  PrimMonad, PrimState, RealWorld
) where

import           Data.Vector.Generic.Mutable.Base
import qualified Data.Vector.Generic.Base as V

import qualified Data.Vector.Fusion.Bundle      as Bundle
import           Data.Vector.Fusion.Bundle      ( Bundle, MBundle, Chunk(..) )
import qualified Data.Vector.Fusion.Bundle.Monadic as MBundle
import           Data.Vector.Fusion.Stream.Monadic ( Stream )
import qualified Data.Vector.Fusion.Stream.Monadic as Stream
import           Data.Vector.Fusion.Bundle.Size
import           Data.Vector.Fusion.Util        ( delay_inline )
import           Data.Vector.Internal.Check

import Control.Monad.Primitive ( PrimMonad(..), RealWorld, stToPrim )

import Prelude
  ( Ord, Monad, Bool(..), Int, Maybe(..), Either(..), Ordering(..)
  , return, otherwise, flip, const, seq, min, max, not, pure
  , (>>=), (+), (-), (<), (<=), (>), (>=), (==), (/=), (.), ($), (=<<), (>>), (<$>) )
import Data.Bits ( Bits(shiftR) )

#include "vector.h"


-- ------------------
-- Internal functions
-- ------------------

unsafeAppend1 :: (PrimMonad m, MVector v a)
        => v (PrimState m) a -> Int -> a -> m (v (PrimState m) a)
{-# INLINE_INNER unsafeAppend1 #-}
    -- NOTE: The case distinction has to be on the outside because
    -- GHC creates a join point for the unsafeWrite even when everything
    -- is inlined. This is bad because with the join point, v isn't getting
    -- unboxed.
unsafeAppend1 v i x
  | i < length v = do
                     unsafeWrite v i x
                     return v
  | otherwise    = do
                     v' <- enlarge v
                     checkIndex Internal i (length v') $ unsafeWrite v' i x
                     return v'

unsafePrepend1 :: (PrimMonad m, MVector v a)
        => v (PrimState m) a -> Int -> a -> m (v (PrimState m) a, Int)
{-# INLINE_INNER unsafePrepend1 #-}
unsafePrepend1 v i x
  | i /= 0    = do
                  let i' = i-1
                  unsafeWrite v i' x
                  return (v, i')
  | otherwise = do
                  (v', j) <- enlargeFront v
                  let i' = j-1
                  checkIndex Internal i' (length v') $ unsafeWrite v' i' x
                  return (v', i')

mstream :: (PrimMonad m, MVector v a) => v (PrimState m) a -> Stream m a
{-# INLINE mstream #-}
mstream v = v `seq` n `seq` Stream.unfoldrM get 0
  where
    n = length v

    {-# INLINE_INNER get #-}
    get i | i < n     = do x <- unsafeRead v i
                           return $ Just (x, i+1)
          | otherwise = return Nothing

fill :: (PrimMonad m, MVector v a)
     => v (PrimState m) a -> Stream m a -> m (v (PrimState m) a)
{-# INLINE fill #-}
fill v s = v `seq` do
                     n' <- Stream.foldM put 0 s
                     return $ unsafeSlice 0 n' v
  where
    {-# INLINE_INNER put #-}
    put i x = do
                checkIndex Internal i (length v) $ unsafeWrite v i x
                return (i+1)

transform
  :: (PrimMonad m, MVector v a)
  => (Stream m a -> Stream m a) -> v (PrimState m) a -> m (v (PrimState m) a)
{-# INLINE_FUSED transform #-}
transform f v = fill v (f (mstream v))

mstreamR :: (PrimMonad m, MVector v a) => v (PrimState m) a -> Stream m a
{-# INLINE mstreamR #-}
mstreamR v = v `seq` n `seq` Stream.unfoldrM get n
  where
    n = length v

    {-# INLINE_INNER get #-}
    get i | j >= 0    = do x <- unsafeRead v j
                           return $ Just (x,j)
          | otherwise = return Nothing
      where
        j = i-1

fillR :: (PrimMonad m, MVector v a)
      => v (PrimState m) a -> Stream m a -> m (v (PrimState m) a)
{-# INLINE fillR #-}
fillR v s = v `seq` do
                      i <- Stream.foldM put n s
                      return $ unsafeSlice i (n-i) v
  where
    n = length v

    {-# INLINE_INNER put #-}
    put i x = do
                unsafeWrite v j x
                return j
      where
        j = i-1

transformR
  :: (PrimMonad m, MVector v a)
  => (Stream m a -> Stream m a) -> v (PrimState m) a -> m (v (PrimState m) a)
{-# INLINE_FUSED transformR #-}
transformR f v = fillR v (f (mstreamR v))

-- | Create a new mutable vector and fill it with elements from the 'Bundle'.
-- The vector will grow exponentially if the maximum size of the 'Bundle' is
-- unknown.
unstream :: (PrimMonad m, MVector v a)
         => Bundle u a -> m (v (PrimState m) a)
-- NOTE: replace INLINE_FUSED by INLINE? (also in unstreamR)
{-# INLINE_FUSED unstream #-}
unstream s = munstream (Bundle.lift s)

-- | Create a new mutable vector and fill it with elements from the monadic
-- stream. The vector will grow exponentially if the maximum size of the stream
-- is unknown.
munstream :: (PrimMonad m, MVector v a)
          => MBundle m u a -> m (v (PrimState m) a)
{-# INLINE_FUSED munstream #-}
munstream s = case upperBound (MBundle.size s) of
               Just n  -> munstreamMax     s n
               Nothing -> munstreamUnknown s

munstreamMax :: (PrimMonad m, MVector v a)
             => MBundle m u a -> Int -> m (v (PrimState m) a)
{-# INLINE munstreamMax #-}
munstreamMax s n
  = do
      v <- checkLength Internal n $ unsafeNew n
      let put i x = do
                       checkIndex Internal i n $ unsafeWrite v i x
                       return (i+1)
      n' <- MBundle.foldM' put 0 s
      return $ checkSlice Internal 0 n' n
             $ unsafeSlice 0 n' v

munstreamUnknown :: (PrimMonad m, MVector v a)
                 => MBundle m u a -> m (v (PrimState m) a)
{-# INLINE munstreamUnknown #-}
munstreamUnknown s
  = do
      v <- unsafeNew 0
      (v', n) <- MBundle.foldM put (v, 0) s
      return $ checkSlice Internal 0 n (length v')
             $ unsafeSlice 0 n v'
  where
    {-# INLINE_INNER put #-}
    put (v,i) x = do
                    v' <- unsafeAppend1 v i x
                    return (v',i+1)


-- | Create a new mutable vector and fill it with elements from the 'Bundle'.
-- The vector will grow exponentially if the maximum size of the 'Bundle' is
-- unknown.
vunstream :: (PrimMonad m, V.Vector v a)
         => Bundle v a -> m (V.Mutable v (PrimState m) a)
-- NOTE: replace INLINE_FUSED by INLINE? (also in unstreamR)
{-# INLINE_FUSED vunstream #-}
vunstream s = vmunstream (Bundle.lift s)

-- | Create a new mutable vector and fill it with elements from the monadic
-- stream. The vector will grow exponentially if the maximum size of the stream
-- is unknown.
vmunstream :: (PrimMonad m, V.Vector v a)
           => MBundle m v a -> m (V.Mutable v (PrimState m) a)
{-# INLINE_FUSED vmunstream #-}
vmunstream s = case upperBound (MBundle.size s) of
               Just n  -> vmunstreamMax     s n
               Nothing -> vmunstreamUnknown s

vmunstreamMax :: (PrimMonad m, V.Vector v a)
              => MBundle m v a -> Int -> m (V.Mutable v (PrimState m) a)
{-# INLINE vmunstreamMax #-}
vmunstreamMax s n
  = do
      v <- checkLength Internal n $ unsafeNew n
      let {-# INLINE_INNER copyChunk #-}
          copyChunk i (Chunk m f) =
            checkSlice Internal i m (length v) $ do
              f (basicUnsafeSlice i m v)
              return (i+m)

      n' <- Stream.foldlM' copyChunk 0 (MBundle.chunks s)
      return $ checkSlice Internal 0 n' n
             $ unsafeSlice 0 n' v

vmunstreamUnknown :: (PrimMonad m, V.Vector v a)
                 => MBundle m v a -> m (V.Mutable v (PrimState m) a)
{-# INLINE vmunstreamUnknown #-}
vmunstreamUnknown s
  = do
      v <- unsafeNew 0
      (v', n) <- Stream.foldlM copyChunk (v,0) (MBundle.chunks s)
      return $ checkSlice Internal 0 n (length v')
             $ unsafeSlice 0 n v'
  where
    {-# INLINE_INNER copyChunk #-}
    copyChunk (v,i) (Chunk n f)
      = do
          let j = i+n
          v' <- if basicLength v < j
                  then unsafeGrow v (delay_inline max (enlarge_delta v) (j - basicLength v))
                  else return v
          checkSlice Internal i n (length v') $ f (basicUnsafeSlice i n v')
          return (v',j)


-- | Create a new mutable vector and fill it with elements from the 'Bundle'
-- from right to left. The vector will grow exponentially if the maximum size
-- of the 'Bundle' is unknown.
unstreamR :: (PrimMonad m, MVector v a)
          => Bundle u a -> m (v (PrimState m) a)
-- NOTE: replace INLINE_FUSED by INLINE? (also in unstream)
{-# INLINE_FUSED unstreamR #-}
unstreamR s = munstreamR (Bundle.lift s)

-- | Create a new mutable vector and fill it with elements from the monadic
-- stream from right to left. The vector will grow exponentially if the maximum
-- size of the stream is unknown.
munstreamR :: (PrimMonad m, MVector v a)
           => MBundle m u a -> m (v (PrimState m) a)
{-# INLINE_FUSED munstreamR #-}
munstreamR s = case upperBound (MBundle.size s) of
               Just n  -> munstreamRMax     s n
               Nothing -> munstreamRUnknown s

munstreamRMax :: (PrimMonad m, MVector v a)
              => MBundle m u a -> Int -> m (v (PrimState m) a)
{-# INLINE munstreamRMax #-}
munstreamRMax s n
  = do
      v <- checkLength Internal n $ unsafeNew n
      let put i x = do
                      let i' = i-1
                      checkIndex Internal i' n
                        $ unsafeWrite v i' x
                      return i'
      i <- MBundle.foldM' put n s
      return $ checkSlice Internal i (n-i) n
             $ unsafeSlice i (n-i) v

munstreamRUnknown :: (HasCallStack, PrimMonad m, MVector v a)
                  => MBundle m u a -> m (v (PrimState m) a)
{-# INLINE munstreamRUnknown #-}
munstreamRUnknown s
  = do
      v <- unsafeNew 0
      (v', i) <- MBundle.foldM put (v, 0) s
      let n = length v'
      return $ checkSlice Internal i (n-i) n
             $ unsafeSlice i (n-i) v'
  where
    {-# INLINE_INNER put #-}
    put (v,i) x = unsafePrepend1 v i x

-- Length
-- ------

-- | Length of the mutable vector.
length :: MVector v a => v s a -> Int
{-# INLINE length #-}
length = basicLength

-- | Check whether the vector is empty.
null :: MVector v a => v s a -> Bool
{-# INLINE null #-}
null v = length v == 0

-- Extracting subvectors
-- ---------------------

-- | Yield a part of the mutable vector without copying it. The vector must
-- contain at least @i+n@ elements.
slice :: (HasCallStack, MVector v a)
      => Int  -- ^ @i@ starting index
      -> Int  -- ^ @n@ length
      -> v s a
      -> v s a
{-# INLINE slice #-}
slice i n v = checkSlice Bounds i n (length v) $ unsafeSlice i n v

-- | Take the @n@ first elements of the mutable vector without making a
-- copy. For negative @n@, the empty vector is returned. If @n@ is larger
-- than the vector's length, the vector is returned unchanged.
take :: MVector v a => Int -> v s a -> v s a
{-# INLINE take #-}
take n v = unsafeSlice 0 (min (max n 0) (length v)) v

-- | Drop the @n@ first element of the mutable vector without making a
-- copy. For negative @n@, the vector is returned unchanged. If @n@ is
-- larger than the vector's length, the empty vector is returned.
drop :: MVector v a => Int -> v s a -> v s a
{-# INLINE drop #-}
drop n v = unsafeSlice (min m n') (max 0 (m - n')) v
  where
    n' = max n 0
    m  = length v

-- | /O(1)/ Split the mutable vector into the first @n@ elements
-- and the remainder, without copying.
--
-- Note that @'splitAt' n v@ is equivalent to @('take' n v, 'drop' n v)@,
-- but slightly more efficient.
splitAt :: MVector v a => Int -> v s a -> (v s a, v s a)
{-# INLINE splitAt #-}
splitAt n v = ( unsafeSlice 0 m v
              , unsafeSlice m (max 0 (len - n')) v
              )
    where
      m   = min n' len
      n'  = max n 0
      len = length v

-- | Drop the last element of the mutable vector without making a copy.
-- If the vector is empty, an exception is thrown.
init :: MVector v a => v s a -> v s a
{-# INLINE init #-}
init v = slice 0 (length v - 1) v

-- | Drop the first element of the mutable vector without making a copy.
-- If the vector is empty, an exception is thrown.
tail :: MVector v a => v s a -> v s a
{-# INLINE tail #-}
tail v = slice 1 (length v - 1) v

-- | Yield a part of the mutable vector without copying it. No bounds checks
-- are performed.
unsafeSlice :: MVector v a => Int  -- ^ starting index
                           -> Int  -- ^ length of the slice
                           -> v s a
                           -> v s a
{-# INLINE unsafeSlice #-}
-- See NOTE: [Strict indexing] in D.V.Generic
unsafeSlice !i !n v = checkSlice Unsafe i n (length v)
                    $ basicUnsafeSlice i n v

-- | Same as 'init', but doesn't do range checks.
unsafeInit :: MVector v a => v s a -> v s a
{-# INLINE unsafeInit #-}
unsafeInit v = unsafeSlice 0 (length v - 1) v

-- | Same as 'tail', but doesn't do range checks.
unsafeTail :: MVector v a => v s a -> v s a
{-# INLINE unsafeTail #-}
unsafeTail v = unsafeSlice 1 (length v - 1) v

-- | Unsafe variant of 'take'. If @n@ is out of range, it will
-- simply create an invalid slice that likely violate memory safety.
unsafeTake :: MVector v a => Int -> v s a -> v s a
{-# INLINE unsafeTake #-}
unsafeTake n v = unsafeSlice 0 n v

-- | Unsafe variant of 'drop'. If @n@ is out of range, it will
-- simply create an invalid slice that likely violate memory safety.
unsafeDrop :: MVector v a => Int -> v s a -> v s a
{-# INLINE unsafeDrop #-}
unsafeDrop n v = unsafeSlice n (length v - n) v

-- Overlapping
-- -----------

-- | Check whether two vectors overlap.
overlaps :: MVector v a => v s a -> v s a -> Bool
{-# INLINE overlaps #-}
overlaps = basicOverlaps

-- Initialisation
-- --------------

-- | Create a mutable vector of the given length.
new :: (HasCallStack, PrimMonad m, MVector v a) => Int -> m (v (PrimState m) a)
{-# INLINE new #-}
new n = checkLength Bounds n $ stToPrim
      $ unsafeNew n >>= \v -> basicInitialize v >> return v

-- | Create a mutable vector of the given length. The vector content
-- should be assumed to be uninitialized. However, the exact semantics depend
-- on the vector implementation. For example, unboxed and storable
-- vectors will create a vector filled with whatever the underlying memory
-- buffer happens to contain, while boxed vector's elements are
-- initialized to bottoms which will throw exception when evaluated.
--
-- @since 0.4
unsafeNew :: (PrimMonad m, MVector v a) => Int -> m (v (PrimState m) a)
{-# INLINE unsafeNew #-}
unsafeNew n = checkLength Unsafe n $ stToPrim $ basicUnsafeNew n

-- | Create a mutable vector of the given length (0 if the length is negative)
-- and fill it with an initial value.
replicate :: (PrimMonad m, MVector v a) => Int -> a -> m (v (PrimState m) a)
{-# INLINE replicate #-}
replicate n x = stToPrim $ basicUnsafeReplicate (delay_inline max 0 n) x

-- | Create a mutable vector of the given length (0 if the length is negative)
-- and fill it with values produced by repeatedly executing the monadic action.
replicateM :: (PrimMonad m, MVector v a) => Int -> m a -> m (v (PrimState m) a)
{-# INLINE replicateM #-}
replicateM n m = munstream (MBundle.replicateM n m)

-- | /O(n)/ Create a mutable vector of the given length (0 if the length is negative)
-- and fill it with the results of applying the function to each index.
-- Iteration starts at index 0.
--
-- @since 0.12.3.0
generate :: (PrimMonad m, MVector v a) => Int -> (Int -> a) -> m (v (PrimState m) a)
{-# INLINE generate #-}
generate n f = stToPrim $ generateM n (return . f)

-- | /O(n)/ Create a mutable vector of the given length (0 if the length is
-- negative) and fill it with the results of applying the monadic function to each
-- index. Iteration starts at index 0.
--
-- @since 0.12.3.0
generateM :: (PrimMonad m, MVector v a) => Int -> (Int -> m a) -> m (v (PrimState m) a)
{-# INLINE generateM #-}
generateM n f = munstream (MBundle.generateM n f)

-- | Create a copy of a mutable vector.
clone :: (PrimMonad m, MVector v a) => v (PrimState m) a -> m (v (PrimState m) a)
{-# INLINE clone #-}
clone v = do
            v' <- unsafeNew (length v)
            unsafeCopy v' v
            return v'

-- Growing
-- -------

-- | Grow a vector by the given number of elements. The number must not be
-- negative, otherwise an exception is thrown. The semantics of this function
-- are exactly the same as of 'unsafeGrow', except that it will initialize the newly
-- allocated memory first.
--
-- It is important to note that mutating the returned vector will not affect the
-- vector that was used as a source. In other words, it does not, nor will it
-- ever have the semantics of @realloc@ from C.
--
-- > grow mv 0 === clone mv
--
-- @since 0.4.0
grow :: (HasCallStack, PrimMonad m, MVector v a)
     => v (PrimState m) a -> Int -> m (v (PrimState m) a)
{-# INLINE grow #-}
grow v by = checkLength Bounds by
          $ stToPrim
          $ do vnew <- unsafeGrow v by
               basicInitialize $ basicUnsafeSlice (length v) by vnew
               return vnew

-- | Same as 'grow', except that it copies data towards the end of the newly
-- allocated vector, making extra space available at the beginning.
--
-- @since 0.11.0.0
growFront :: (HasCallStack, PrimMonad m, MVector v a)
          => v (PrimState m) a -> Int -> m (v (PrimState m) a)
{-# INLINE growFront #-}
growFront v by = checkLength Bounds by
               $ stToPrim
               $ do vnew <- unsafeGrowFront v by
                    basicInitialize $ basicUnsafeSlice 0 by vnew
                    return vnew

enlarge_delta :: MVector v a => v s a -> Int
enlarge_delta v = max (length v) 1

-- | Grow a vector logarithmically.
enlarge :: (PrimMonad m, MVector v a)
        => v (PrimState m) a -> m (v (PrimState m) a)
{-# INLINE enlarge #-}
enlarge v = stToPrim $ do
  vnew <- unsafeGrow v by
  basicInitialize $ basicUnsafeSlice (length v) by vnew
  return vnew
  where
    by = enlarge_delta v

enlargeFront :: (PrimMonad m, MVector v a)
             => v (PrimState m) a -> m (v (PrimState m) a, Int)
{-# INLINE enlargeFront #-}
enlargeFront v = stToPrim $ do
                   v' <- unsafeGrowFront v by
                   basicInitialize $ basicUnsafeSlice 0 by v'
                   return (v', by)
  where
    by = enlarge_delta v

-- | Grow a vector by allocating a new mutable vector of the same size plus the
-- the given number of elements and copying all the data over to the new vector,
-- starting at its beginning. The newly allocated memory is not initialized and
-- the extra space at the end will likely contain garbage data or bottoms.
-- Use 'unsafeGrowFront' to make the extra space available in the front
-- of the new vector.
--
-- It is important to note that mutating the returned vector will not affect
-- elements of the vector that was used as a source. In other words, it does not,
-- nor will it ever have the semantics of @realloc@ from C. Keep in mind,
-- however, that values themselves can be of a mutable type
-- (eg. 'Foreign.Ptr.Ptr'), in which case it would be possible to affect values
-- stored in both vectors.
--
-- > unsafeGrow mv 0 === clone mv
--
-- @since 0.4.0
unsafeGrow
  :: (PrimMonad m, MVector v a)
  => v (PrimState m) a
  -- ^ mutable vector to copy from
  -> Int
  -- ^ number of elements to grow the vector by (must be non-negative, but
  -- this is not checked)
  -> m (v (PrimState m) a)
{-# INLINE unsafeGrow #-}
unsafeGrow v n = checkLength Unsafe n
               $ stToPrim
               $ basicUnsafeGrow v n

-- | Same as 'unsafeGrow', except that it copies data towards the end of the
-- newly allocated vector, making extra space available at the beginning.
--
-- @since 0.11.0.0
unsafeGrowFront :: (PrimMonad m, MVector v a)
                => v (PrimState m) a -> Int -> m (v (PrimState m) a)
{-# INLINE unsafeGrowFront #-}
unsafeGrowFront v by = checkLength Unsafe by $ stToPrim $ do
                         let n = length v
                         v' <- basicUnsafeNew (by+n)
                         basicUnsafeCopy (basicUnsafeSlice by n v') v
                         return v'

-- Restricting memory usage
-- ------------------------

-- | Reset all elements of the vector to some undefined value, clearing all
-- references to external objects. This is usually a noop for unboxed vectors.
clear :: (PrimMonad m, MVector v a) => v (PrimState m) a -> m ()
{-# INLINE clear #-}
clear = stToPrim . basicClear

-- Accessing individual elements
-- -----------------------------

-- | Yield the element at the given position. Will throw an exception if
-- the index is out of range.
--
-- ==== __Examples__
--
-- >>> import qualified Data.Vector.Strict.Mutable as MV
-- >>> v <- MV.generate 10 (\x -> x*x)
-- >>> MV.read v 3
-- 9
read :: (HasCallStack, PrimMonad m, MVector v a) => v (PrimState m) a -> Int -> m a
{-# INLINE read #-}
read v i = checkIndex Bounds i (length v)
         $ unsafeRead v i

-- | Yield the element at the given position. Returns 'Nothing' if
-- the index is out of range.
--
-- @since 0.13
--
-- ==== __Examples__
--
-- >>> import qualified Data.Vector.Strict.Mutable as MV
-- >>> v <- MV.generate 10 (\x -> x*x)
-- >>> MV.readMaybe v 3
-- Just 9
-- >>> MV.readMaybe v 13
-- Nothing
readMaybe :: (PrimMonad m, MVector v a) => v (PrimState m) a -> Int -> m (Maybe a)
{-# INLINE readMaybe #-}
readMaybe v i | i `inRange` (length v) = Just <$> unsafeRead v i
              | otherwise              = pure Nothing

-- | Replace the element at the given position.
write :: (HasCallStack, PrimMonad m, MVector v a) => v (PrimState m) a -> Int -> a -> m ()
{-# INLINE write #-}
write v i x = checkIndex Bounds i (length v)
            $ unsafeWrite v i x

-- | Modify the element at the given position.
modify :: (HasCallStack, PrimMonad m, MVector v a) => v (PrimState m) a -> (a -> a) -> Int -> m ()
{-# INLINE modify #-}
modify v f i = checkIndex Bounds i (length v)
             $ unsafeModify v f i

-- | Modify the element at the given position using a monadic function.
--
-- @since 0.12.3.0
modifyM :: (HasCallStack, PrimMonad m, MVector v a) => v (PrimState m) a -> (a -> m a) -> Int -> m ()
{-# INLINE modifyM #-}
modifyM v f i = checkIndex Bounds i (length v)
              $ unsafeModifyM v f i

-- | Swap the elements at the given positions.
swap :: (HasCallStack, PrimMonad m, MVector v a) => v (PrimState m) a -> Int -> Int -> m ()
{-# INLINE swap #-}
swap v i j = checkIndex Bounds i (length v)
           $ checkIndex Bounds j (length v)
           $ unsafeSwap v i j

-- | Replace the element at the given position and return the old element.
exchange :: (HasCallStack, PrimMonad m, MVector v a) => v (PrimState m) a -> Int -> a -> m a
{-# INLINE exchange #-}
exchange v i x = checkIndex Bounds i (length v) $ unsafeExchange v i x

-- | Yield the element at the given position. No bounds checks are performed.
unsafeRead :: (PrimMonad m, MVector v a) => v (PrimState m) a -> Int -> m a
{-# INLINE unsafeRead #-}
-- See NOTE: [Strict indexing] in D.V.Generic
unsafeRead v !i = checkIndex Unsafe i (length v)
                $ stToPrim
                $ basicUnsafeRead v i

-- | Replace the element at the given position. No bounds checks are performed.
unsafeWrite :: (PrimMonad m, MVector v a) => v (PrimState m) a -> Int -> a -> m ()
{-# INLINE unsafeWrite #-}
-- See NOTE: [Strict indexing] in D.V.Generic
unsafeWrite v !i x = checkIndex Unsafe i (length v)
                   $ stToPrim
                   $ basicUnsafeWrite v i x

-- | Modify the element at the given position. No bounds checks are performed.
unsafeModify :: (PrimMonad m, MVector v a) => v (PrimState m) a -> (a -> a) -> Int -> m ()
{-# INLINE unsafeModify #-}
-- See NOTE: [Strict indexing] in D.V.Generic
unsafeModify v f !i = checkIndex Unsafe i (length v)
                    $ stToPrim
                    $ basicUnsafeRead v i >>= \x ->
                      basicUnsafeWrite v i (f x)

-- | Modify the element at the given position using a monadic
-- function. No bounds checks are performed.
--
-- @since 0.12.3.0
unsafeModifyM :: (PrimMonad m, MVector v a) => v (PrimState m) a -> (a -> m a) -> Int -> m ()
{-# INLINE unsafeModifyM #-}
-- See NOTE: [Strict indexing] in D.V.Generic
unsafeModifyM v f !i = checkIndex Unsafe i (length v)
                     $ stToPrim . basicUnsafeWrite v i =<< f =<< stToPrim (basicUnsafeRead v i)

-- | Swap the elements at the given positions. No bounds checks are performed.
unsafeSwap :: (PrimMonad m, MVector v a) => v (PrimState m) a -> Int -> Int -> m ()
{-# INLINE unsafeSwap #-}
unsafeSwap v i j = checkIndex Unsafe i (length v)
                 $ checkIndex Unsafe j (length v)
                 $ stToPrim $ do
                     x <- unsafeRead v i
                     y <- unsafeRead v j
                     unsafeWrite v i y
                     unsafeWrite v j x

-- | Replace the element at the given position and return the old element. No
-- bounds checks are performed.
unsafeExchange :: (PrimMonad m, MVector v a) => v (PrimState m) a -> Int -> a -> m a
{-# INLINE unsafeExchange #-}
unsafeExchange v i x = checkIndex Unsafe i (length v) $ stToPrim $ do
                         y <- unsafeRead v i
                         unsafeWrite v i x
                         return y

-- Folds
-- -----

forI_ :: (Monad m, MVector v a) => v (PrimState m) a -> (Int -> m b) -> m ()
{-# INLINE forI_ #-}
forI_ v f = loop 0
  where
    loop i | i >= n    = return ()
           | otherwise = f i >> loop (i + 1)
    n = length v

-- | /O(n)/ Apply the monadic action to every element of the vector, discarding the results.
--
-- @since 0.12.3.0
mapM_ :: (PrimMonad m, MVector v a) => (a -> m b) -> v (PrimState m) a -> m ()
{-# INLINE mapM_ #-}
mapM_ f v = forI_ v $ \i -> f =<< unsafeRead v i

-- | /O(n)/ Apply the monadic action to every element of the vector and its index, discarding the results.
--
-- @since 0.12.3.0
imapM_ :: (PrimMonad m, MVector v a) => (Int -> a -> m b) -> v (PrimState m) a -> m ()
{-# INLINE imapM_ #-}
imapM_ f v = forI_ v $ \i -> f i =<< unsafeRead v i

-- | /O(n)/ Apply the monadic action to every element of the vector,
-- discarding the results. It's the same as @flip mapM_@.
--
-- @since 0.12.3.0
forM_ :: (PrimMonad m, MVector v a) => v (PrimState m) a -> (a -> m b) -> m ()
{-# INLINE forM_ #-}
forM_ = flip mapM_

-- | /O(n)/ Apply the monadic action to every element of the vector
-- and its index, discarding the results. It's the same as @flip imapM_@.
--
-- @since 0.12.3.0
iforM_ :: (PrimMonad m, MVector v a) => v (PrimState m) a -> (Int -> a -> m b) -> m ()
{-# INLINE iforM_ #-}
iforM_ = flip imapM_

-- | /O(n)/ Pure left fold.
--
-- @since 0.12.3.0
foldl :: (PrimMonad m, MVector v a) => (b -> a -> b) -> b -> v (PrimState m) a -> m b
{-# INLINE foldl #-}
foldl f = ifoldl (\b _ -> f b)

-- | /O(n)/ Pure left fold with strict accumulator.
--
-- @since 0.12.3.0
foldl' :: (PrimMonad m, MVector v a) => (b -> a -> b) -> b -> v (PrimState m) a -> m b
{-# INLINE foldl' #-}
foldl' f = ifoldl' (\b _ -> f b)

-- | /O(n)/ Pure left fold using a function applied to each element and its index.
--
-- @since 0.12.3.0
ifoldl :: (PrimMonad m, MVector v a) => (b -> Int -> a -> b) -> b -> v (PrimState m) a -> m b
{-# INLINE ifoldl #-}
ifoldl f b0 v = stToPrim $ ifoldM (\b i a -> return $ f b i a) b0 v

-- | /O(n)/ Pure left fold with strict accumulator using a function applied to each element and its index.
--
-- @since 0.12.3.0
ifoldl' :: (PrimMonad m, MVector v a) => (b -> Int -> a -> b) -> b -> v (PrimState m) a -> m b
{-# INLINE ifoldl' #-}
ifoldl' f b0 v = stToPrim $ ifoldM' (\b i a -> return $ f b i a) b0 v

-- | /O(n)/ Pure right fold.
--
-- @since 0.12.3.0
foldr :: (PrimMonad m, MVector v a) => (a -> b -> b) -> b -> v (PrimState m) a -> m b
{-# INLINE foldr #-}
foldr f = ifoldr (const f)

-- | /O(n)/ Pure right fold with strict accumulator.
--
-- @since 0.12.3.0
foldr' :: (PrimMonad m, MVector v a) => (a -> b -> b) -> b -> v (PrimState m) a -> m b
{-# INLINE foldr' #-}
foldr' f = ifoldr' (const f)

-- | /O(n)/ Pure right fold using a function applied to each element and its index.
--
-- @since 0.12.3.0
ifoldr :: (PrimMonad m, MVector v a) => (Int -> a -> b -> b) -> b -> v (PrimState m) a -> m b
{-# INLINE ifoldr #-}
ifoldr f b0 v = stToPrim $ ifoldrM (\i a b -> return $ f i a b) b0 v

-- | /O(n)/ Pure right fold with strict accumulator using a function applied
-- to each element and its index.
--
-- @since 0.12.3.0
ifoldr' :: (PrimMonad m, MVector v a) => (Int -> a -> b -> b) -> b -> v (PrimState m) a -> m b
{-# INLINE ifoldr' #-}
ifoldr' f b0 v = stToPrim $ ifoldrM' (\i a b -> return $ f i a b) b0 v

-- | /O(n)/ Monadic fold.
--
-- @since 0.12.3.0
foldM :: (PrimMonad m, MVector v a) => (b -> a -> m b) -> b -> v (PrimState m) a -> m b
{-# INLINE foldM #-}
foldM f = ifoldM (\x _ -> f x)

-- | /O(n)/ Monadic fold with strict accumulator.
--
-- @since 0.12.3.0
foldM' :: (PrimMonad m, MVector v a) => (b -> a -> m b) -> b -> v (PrimState m) a -> m b
{-# INLINE foldM' #-}
foldM' f = ifoldM' (\x _ -> f x)

-- | /O(n)/ Monadic fold using a function applied to each element and its index.
--
-- @since 0.12.3.0
ifoldM :: (PrimMonad m, MVector v a) => (b -> Int -> a -> m b) -> b -> v (PrimState m) a -> m b
{-# INLINE ifoldM #-}
ifoldM f b0 v = loop 0 b0
  where
    loop i b | i >= n    = return b
             | otherwise = do a <- unsafeRead v i
                              loop (i + 1) =<< f b i a
    n = length v

-- | /O(n)/ Monadic fold with strict accumulator using a function applied to each element and its index.
--
-- @since 0.12.3.0
ifoldM' :: (PrimMonad m, MVector v a) => (b -> Int -> a -> m b) -> b -> v (PrimState m) a -> m b
{-# INLINE ifoldM' #-}
ifoldM' f b0 v = loop 0 b0
  where
    loop i !b | i >= n    = return b
              | otherwise = do a <- unsafeRead v i
                               loop (i + 1) =<< f b i a
    n = length v

-- | /O(n)/ Monadic right fold.
--
-- @since 0.12.3.0
foldrM :: (PrimMonad m, MVector v a) => (a -> b -> m b) -> b -> v (PrimState m) a -> m b
{-# INLINE foldrM #-}
foldrM f = ifoldrM (const f)

-- | /O(n)/ Monadic right fold with strict accumulator.
--
-- @since 0.12.3.0
foldrM' :: (PrimMonad m, MVector v a) => (a -> b -> m b) -> b -> v (PrimState m) a -> m b
{-# INLINE foldrM' #-}
foldrM' f = ifoldrM' (const f)

-- | /O(n)/ Monadic right fold using a function applied to each element and its index.
--
-- @since 0.12.3.0
ifoldrM :: (PrimMonad m, MVector v a) => (Int -> a -> b -> m b) -> b -> v (PrimState m) a -> m b
{-# INLINE ifoldrM #-}
ifoldrM f b0 v = loop (n-1) b0
  where
    loop i b | i < 0     = return b
             | otherwise = do a <- unsafeRead v i
                              loop (i - 1) =<< f i a b
    n = length v

-- | /O(n)/ Monadic right fold with strict accumulator using a function applied
-- to each element and its index.
--
-- @since 0.12.3.0
ifoldrM' :: (PrimMonad m, MVector v a) => (Int -> a -> b -> m b) -> b -> v (PrimState m) a -> m b
{-# INLINE ifoldrM' #-}
ifoldrM' f b0 v = loop (n-1) b0
  where
    loop i !b | i < 0     = return b
              | otherwise = do a <- unsafeRead v i
                               loop (i - 1) =<< f i a b
    n = length v

-- Filling and copying
-- -------------------

-- | Set all elements of the vector to the given value.
set :: (PrimMonad m, MVector v a) => v (PrimState m) a -> a -> m ()
{-# INLINE set #-}
set v = stToPrim . basicSet v

-- | Copy a vector. The two vectors must have the same length and may not
-- overlap.
copy :: (HasCallStack, PrimMonad m, MVector v a)
     => v (PrimState m) a   -- ^ target
     -> v (PrimState m) a   -- ^ source
     -> m ()
{-# INLINE copy #-}
copy dst src = check Bounds "overlapping vectors" (not (dst `overlaps` src))
             $ check Bounds "length mismatch" (length dst == length src)
             $ unsafeCopy dst src

-- | Move the contents of a vector. The two vectors must have the same
-- length.
--
-- If the vectors do not overlap, then this is equivalent to 'copy'.
-- Otherwise, the copying is performed as if the source vector were
-- copied to a temporary vector and then the temporary vector was copied
-- to the target vector.
move :: (HasCallStack, PrimMonad m, MVector v a)
     => v (PrimState m) a   -- ^ target
     -> v (PrimState m) a   -- ^ source
     -> m ()
{-# INLINE move #-}
move dst src = check Bounds "length mismatch" (length dst == length src)
             $ unsafeMove dst src

-- | Copy a vector. The two vectors must have the same length and may not
-- overlap, but this is not checked.
unsafeCopy :: (PrimMonad m, MVector v a)
           => v (PrimState m) a   -- ^ target
           -> v (PrimState m) a   -- ^ source
           -> m ()
{-# INLINE unsafeCopy #-}
unsafeCopy dst src = check Unsafe "length mismatch" (length dst == length src)
                   $ check Unsafe "overlapping vectors" (not (dst `overlaps` src))
                   $ dst `seq` src `seq` stToPrim (basicUnsafeCopy dst src)

-- | Move the contents of a vector. The two vectors must have the same
-- length, but this is not checked.
--
-- If the vectors do not overlap, then this is equivalent to 'unsafeCopy'.
-- Otherwise, the copying is performed as if the source vector were
-- copied to a temporary vector and then the temporary vector was copied
-- to the target vector.
unsafeMove :: (PrimMonad m, MVector v a)
           => v (PrimState m) a   -- ^ target
           -> v (PrimState m) a   -- ^ source
           -> m ()
{-# INLINE unsafeMove #-}
unsafeMove dst src = check Unsafe "length mismatch" (length dst == length src)
                   $ dst `seq` src `seq` stToPrim (basicUnsafeMove dst src)


accum :: forall m v a b u. (HasCallStack, PrimMonad m, MVector v a)
      => (a -> b -> a) -> v (PrimState m) a -> Bundle u (Int, b) -> m ()
{-# INLINE accum #-}
accum f !v s = Bundle.mapM_ upd s
  where
    {-# INLINE_INNER upd #-}
    upd :: HasCallStack => (Int, b) -> m ()
    upd (i,b) = do
                  a <- checkIndex Bounds i n $ unsafeRead v i
                  unsafeWrite v i (f a b)
    !n = length v

update :: forall m v a u. (HasCallStack, PrimMonad m, MVector v a)
       => v (PrimState m) a -> Bundle u (Int, a) -> m ()
{-# INLINE update #-}
update !v s = Bundle.mapM_ upd s
  where
    {-# INLINE_INNER upd #-}
    upd :: HasCallStack => (Int, a) -> m ()
    upd (i,b) = checkIndex Bounds i n $ unsafeWrite v i b

    !n = length v

unsafeAccum :: (PrimMonad m, MVector v a)
            => (a -> b -> a) -> v (PrimState m) a -> Bundle u (Int, b) -> m ()
{-# INLINE unsafeAccum #-}
unsafeAccum f !v s = Bundle.mapM_ upd s
  where
    {-# INLINE_INNER upd #-}
    upd (i,b) = do
                  a <- checkIndex Unsafe i n $ unsafeRead v i
                  unsafeWrite v i (f a b)
    !n = length v

unsafeUpdate :: (PrimMonad m, MVector v a)
                        => v (PrimState m) a -> Bundle u (Int, a) -> m ()
{-# INLINE unsafeUpdate #-}
unsafeUpdate !v s = Bundle.mapM_ upd s
  where
    {-# INLINE_INNER upd #-}
    upd (i,b) = checkIndex Unsafe i n $ unsafeWrite v i b
    !n = length v

reverse :: (PrimMonad m, MVector v a) => v (PrimState m) a -> m ()
{-# INLINE reverse #-}
reverse !v = reverse_loop 0 (length v - 1)
  where
    reverse_loop i j | i < j = do
                                 unsafeSwap v i j
                                 reverse_loop (i + 1) (j - 1)
    reverse_loop _ _ = return ()

unstablePartition :: forall m v a. (PrimMonad m, MVector v a)
                  => (a -> Bool) -> v (PrimState m) a -> m Int
{-# INLINE unstablePartition #-}
unstablePartition f !v = from_left 0 (length v)
  where
    -- NOTE: GHC 6.10.4 panics without the signatures on from_left and
    -- from_right
    from_left :: Int -> Int -> m Int
    from_left i j
      | i == j    = return i
      | otherwise = do
                      x <- unsafeRead v i
                      if f x
                        then from_left (i+1) j
                        else from_right i (j-1)

    from_right :: Int -> Int -> m Int
    from_right i j
      | i == j    = return i
      | otherwise = do
                      x <- unsafeRead v j
                      if f x
                        then do
                               y <- unsafeRead v i
                               unsafeWrite v i x
                               unsafeWrite v j y
                               from_left (i+1) j
                        else from_right i (j-1)

unstablePartitionBundle :: (PrimMonad m, MVector v a)
        => (a -> Bool) -> Bundle u a -> m (v (PrimState m) a, v (PrimState m) a)
{-# INLINE unstablePartitionBundle #-}
unstablePartitionBundle f s
  = case upperBound (Bundle.size s) of
      Just n  -> unstablePartitionMax f s n
      Nothing -> partitionUnknown f s

unstablePartitionMax :: (PrimMonad m, MVector v a)
        => (a -> Bool) -> Bundle u a -> Int
        -> m (v (PrimState m) a, v (PrimState m) a)
{-# INLINE unstablePartitionMax #-}
unstablePartitionMax f s n
  = do
      v <- checkLength Internal n $ unsafeNew n
      let {-# INLINE_INNER put #-}
          put (i, j) x
            | f x       = do
                            unsafeWrite v i x
                            return (i+1, j)
            | otherwise = do
                            unsafeWrite v (j-1) x
                            return (i, j-1)

      (i,j) <- Bundle.foldM' put (0, n) s
      return (unsafeSlice 0 i v, unsafeSlice j (n-j) v)

partitionBundle :: (PrimMonad m, MVector v a)
        => (a -> Bool) -> Bundle u a -> m (v (PrimState m) a, v (PrimState m) a)
{-# INLINE partitionBundle #-}
partitionBundle f s
  = case upperBound (Bundle.size s) of
      Just n  -> partitionMax f s n
      Nothing -> partitionUnknown f s

partitionMax :: (PrimMonad m, MVector v a)
  => (a -> Bool) -> Bundle u a -> Int -> m (v (PrimState m) a, v (PrimState m) a)
{-# INLINE partitionMax #-}
partitionMax f s n
  = do
      v <- checkLength Internal n $ unsafeNew n

      let {-# INLINE_INNER put #-}
          put (i,j) x
            | f x       = do
                            unsafeWrite v i x
                            return (i+1,j)

            | otherwise = let j' = j-1 in
                          do
                            unsafeWrite v j' x
                            return (i,j')

      (i,j) <- Bundle.foldM' put (0,n) s
      check Internal "invalid indices" (i <= j)
        $ return ()
      let l = unsafeSlice 0 i v
          r = unsafeSlice j (n-j) v
      reverse r
      return (l,r)

partitionUnknown :: (PrimMonad m, MVector v a)
        => (a -> Bool) -> Bundle u a -> m (v (PrimState m) a, v (PrimState m) a)
{-# INLINE partitionUnknown #-}
partitionUnknown f s
  = do
      v1 <- unsafeNew 0
      v2 <- unsafeNew 0
      (v1', n1, v2', n2) <- Bundle.foldM' put (v1, 0, v2, 0) s
      checkSlice Internal 0 n1 (length v1')
        $ checkSlice Internal 0 n2 (length v2')
        $ return (unsafeSlice 0 n1 v1', unsafeSlice 0 n2 v2')
  where
    -- NOTE: The case distinction has to be on the outside because
    -- GHC creates a join point for the unsafeWrite even when everything
    -- is inlined. This is bad because with the join point, v isn't getting
    -- unboxed.
    {-# INLINE_INNER put #-}
    put (v1, i1, v2, i2) x
      | f x       = do
                      v1' <- unsafeAppend1 v1 i1 x
                      return (v1', i1+1, v2, i2)
      | otherwise = do
                      v2' <- unsafeAppend1 v2 i2 x
                      return (v1, i1, v2', i2+1)


partitionWithBundle :: (PrimMonad m, MVector v a, MVector v b, MVector v c)
        => (a -> Either b c) -> Bundle u a -> m (v (PrimState m) b, v (PrimState m) c)
{-# INLINE partitionWithBundle #-}
partitionWithBundle f s
  = case upperBound (Bundle.size s) of
      Just n  -> partitionWithMax f s n
      Nothing -> partitionWithUnknown f s

partitionWithMax :: (PrimMonad m, MVector v a, MVector v b, MVector v c)
  => (a -> Either b c) -> Bundle u a -> Int -> m (v (PrimState m) b, v (PrimState m) c)
{-# INLINE partitionWithMax #-}
partitionWithMax f s n
  = do
      v1 <- unsafeNew n
      v2 <- unsafeNew n
      let {-# INLINE_INNER put #-}
          put (i1, i2) x = case f x of
            Left b -> do
              unsafeWrite v1 i1 b
              return (i1+1, i2)
            Right c -> do
              unsafeWrite v2 i2 c
              return (i1, i2+1)
      (n1, n2) <- Bundle.foldM' put (0, 0) s
      checkSlice Internal 0 n1 (length v1)
        $ checkSlice Internal 0 n2 (length v2)
        $ return (unsafeSlice 0 n1 v1, unsafeSlice 0 n2 v2)

partitionWithUnknown :: forall m v u a b c.
     (PrimMonad m, MVector v a, MVector v b, MVector v c)
  => (a -> Either b c) -> Bundle u a -> m (v (PrimState m) b, v (PrimState m) c)
{-# INLINE partitionWithUnknown #-}
partitionWithUnknown f s
  = do
      v1 <- unsafeNew 0
      v2 <- unsafeNew 0
      (v1', n1, v2', n2) <- Bundle.foldM' put (v1, 0, v2, 0) s
      checkSlice Internal 0 n1 (length v1')
        $ checkSlice Internal 0 n2 (length v2')
        $ return (unsafeSlice 0 n1 v1', unsafeSlice 0 n2 v2')
  where
    put :: (v (PrimState m) b, Int, v (PrimState m) c, Int)
        -> a
        -> m (v (PrimState m) b, Int, v (PrimState m) c, Int)
    {-# INLINE_INNER put #-}
    put (v1, i1, v2, i2) x = case f x of
      Left b -> do
        v1' <- unsafeAppend1 v1 i1 b
        return (v1', i1+1, v2, i2)
      Right c -> do
        v2' <- unsafeAppend1 v2 i2 c
        return (v1, i1, v2', i2+1)

-- Modifying vectors
-- -----------------


-- | Compute the (lexicographically) next permutation of the given vector in-place.
-- Returns False when the input is the last item in the enumeration, i.e., if it is in
-- weakly descending order. In this case the vector will not get updated,
-- as opposed to the behavior of the C++ function @std::next_permutation@.
nextPermutation :: (PrimMonad m, Ord e, MVector v e) => v (PrimState m) e -> m Bool
{-# INLINE nextPermutation #-}
nextPermutation = nextPermutationByLt (<)

-- | Compute the (lexicographically) next permutation of the given vector in-place,
-- using the provided comparison function.
-- Returns False when the input is the last item in the enumeration, i.e., if it is in
-- weakly descending order. In this case the vector will not get updated,
-- as opposed to the behavior of the C++ function @std::next_permutation@.
--
-- @since 0.13.2.0
nextPermutationBy :: (PrimMonad m, MVector v e) => (e -> e -> Ordering) -> v (PrimState m) e -> m Bool
{-# INLINE nextPermutationBy #-}
nextPermutationBy cmp = nextPermutationByLt (\x y -> cmp x y == LT)

-- | Compute the (lexicographically) previous permutation of the given vector in-place.
-- Returns False when the input is the last item in the enumeration, i.e., if it is in
-- weakly ascending order. In this case the vector will not get updated,
-- as opposed to the behavior of the C++ function @std::prev_permutation@.
--
-- @since 0.13.2.0
prevPermutation :: (PrimMonad m, Ord e, MVector v e) => v (PrimState m) e -> m Bool
{-# INLINE prevPermutation #-}
prevPermutation = nextPermutationByLt (>)

-- | Compute the (lexicographically) previous permutation of the given vector in-place,
-- using the provided comparison function.
-- Returns False when the input is the last item in the enumeration, i.e., if it is in
-- weakly ascending order. In this case the vector will not get updated,
-- as opposed to the behavior of the C++ function @std::prev_permutation@.
--
-- @since 0.13.2.0
prevPermutationBy :: (PrimMonad m, MVector v e) => (e -> e -> Ordering) -> v (PrimState m) e -> m Bool
{-# INLINE prevPermutationBy #-}
prevPermutationBy cmp = nextPermutationByLt (\x y -> cmp x y == GT)

{-
http://en.wikipedia.org/wiki/Permutation#Algorithms_to_generate_permutations

The following algorithm generates the next permutation lexicographically after
a given permutation. It changes the given permutation in-place.

1. Find the largest index k such that a[k] < a[k + 1]. If no such index exists,
   the permutation is the last permutation.
2. Find the largest index l greater than k such that a[k] < a[l].
3. Swap the value of a[k] with that of a[l].
4. Reverse the sequence from a[k + 1] up to and including the final element a[n]

The algorithm has been updated to look up the k in Step 1 beginning from the
last of the vector; which renders the algorithm to achieve the average time
complexity of O(1) each call. The worst case time complexity is still O(n).
The orginal implementation, which scanned the vector from the left, had the
time complexity of O(n) on the best case.
-}

-- | Compute the (lexicographically) next permutation of the given vector in-place.
-- Here, the first argument should be a less-than comparison function.
-- Returns False when the input is the last permutation; in this case the vector
-- will not get updated, as opposed to the behavior of the C++ function 
-- @std::next_permutation@.
nextPermutationByLt :: (PrimMonad m, MVector v e) => (e -> e -> Bool) -> v (PrimState m) e -> m Bool
{-# INLINE nextPermutationByLt #-}
nextPermutationByLt lt v
  | dim < 2 = return False
  | otherwise = stToPrim $ do
      !vlast <- unsafeRead v (dim - 1)
      decrLoop (dim - 2) vlast
  where
    dim = length v
    -- find the largest index k such that a[k] < a[k + 1], and then pass to the rest.
    decrLoop !i !vi1 | i >= 0 = do
      !vi <- unsafeRead v i
      if vi `lt` vi1 then swapLoop i vi (i+1) vi1 dim else decrLoop (i-1) vi
    decrLoop _ !_ = return False
    -- find the largest index l greater than k such that a[k] < a[l], and do the rest.
    swapLoop !k !vk = go
      where
        -- binary search.
        go !l !vl !r | r - l <= 1 = do
          -- Done; do the rest of the algorithm.
          unsafeWrite v k vl
          unsafeWrite v l vk
          reverse $ unsafeSlice (k + 1) (dim - k - 1) v
          return True
        go !l !vl !r = do
          !vmid <- unsafeRead v mid
          if vk `lt` vmid
            then go mid vmid r
            else go l vl mid
          where
            !mid = l + (r - l) `shiftR` 1
  

-- $setup
-- >>> import Prelude ((*))