{-# LANGUAGE Rank2Types, MultiParamTypeClasses, FlexibleContexts, TypeFamilies, ScopedTypeVariables #-} {-# OPTIONS_HADDOCK hide #-} -- | -- Module : Data.Vector.Generic.Base -- Copyright : (c) Roman Leshchinskiy 2008-2010 -- License : BSD-style -- -- Maintainer : Roman Leshchinskiy <rl@cse.unsw.edu.au> -- Stability : experimental -- Portability : non-portable -- -- Class of pure vectors -- module Data.Vector.Generic.Base ( Vector(..), Mutable ) where import Data.Vector.Generic.Mutable ( MVector ) import qualified Data.Vector.Generic.Mutable as M import Control.Monad.Primitive -- | @Mutable v s a@ is the mutable version of the pure vector type @v a@ with -- the state token @s@ -- type family Mutable (v :: * -> *) :: * -> * -> * -- | Class of immutable vectors. Every immutable vector is associated with its -- mutable version through the 'Mutable' type family. Methods of this class -- should not be used directly. Instead, "Data.Vector.Generic" and other -- Data.Vector modules provide safe and fusible wrappers. -- -- Minimum complete implementation: -- -- * 'unsafeFreeze' -- -- * 'basicLength' -- -- * 'basicUnsafeSlice' -- -- * 'basicUnsafeIndexM' -- class MVector (Mutable v) a => Vector v a where -- | /Assume complexity: O(1)/ -- -- Unsafely convert a mutable vector to its immutable version -- without copying. The mutable vector may not be used after -- this operation. unsafeFreeze :: PrimMonad m => Mutable v (PrimState m) a -> m (v a) -- | /Assumed complexity: O(1)/ -- -- Yield the length of the vector. basicLength :: v a -> Int -- | /Assumed complexity: O(1)/ -- -- Yield a slice of the vector without copying it. No range checks are -- performed. basicUnsafeSlice :: Int -- ^ starting index -> Int -- ^ length -> v a -> v a -- | /Assumed complexity: O(1)/ -- -- Yield the element at the given position in a monad. No range checks are -- performed. -- -- The monad allows us to be strict in the vector if we want. Suppose we had -- -- > unsafeIndex :: v a -> Int -> a -- -- instead. Now, if we wanted to copy a vector, we'd do something like -- -- > copy mv v ... = ... unsafeWrite mv i (unsafeIndex v i) ... -- -- For lazy vectors, the indexing would not be evaluated which means that we -- would retain a reference to the original vector in each element we write. -- This is not what we want! -- -- With 'basicUnsafeIndexM', we can do -- -- > copy mv v ... = ... case basicUnsafeIndexM v i of -- > Box x -> unsafeWrite mv i x ... -- -- which does not have this problem because indexing (but not the returned -- element!) is evaluated immediately. -- basicUnsafeIndexM :: Monad m => v a -> Int -> m a -- | /Assumed complexity: O(n)/ -- -- Copy an immutable vector into a mutable one. The two vectors must have -- the same length but this is not checked. -- -- Instances of 'Vector' should redefine this method if they wish to support -- an efficient block copy operation. -- -- Default definition: copying basic on 'basicUnsafeIndexM' and -- 'basicUnsafeWrite'. basicUnsafeCopy :: PrimMonad m => Mutable v (PrimState m) a -> v a -> m () {-# INLINE basicUnsafeCopy #-} basicUnsafeCopy dst src = do_copy 0 where n = basicLength src do_copy i | i < n = do x <- basicUnsafeIndexM src i M.basicUnsafeWrite dst i x do_copy (i+1) | otherwise = return () -- | Evaluate @a@ as far as storing it in a vector would and yield @b@. -- The @v a@ argument only fixes the type and is not touched. The method is -- only used for optimisation purposes. Thus, it is safe for instances of -- 'Vector' to evaluate @a@ less than it would be when stored in a vector -- although this might result in suboptimal code. -- -- > elemseq v x y = (singleton x `asTypeOf` v) `seq` y -- -- Default defintion: @a@ is not evaluated at all -- elemseq :: v a -> a -> b -> b {-# INLINE elemseq #-} elemseq _ = \_ x -> x