如何将可变向量放入状态单子

Question

我在 haskell 中编写了一个小程序，使用带有 Vector 的 State Monad 来计算 Tree 中所有 Int 值的出现次数：

import Data.Vector
import Control.Monad.State
import Control.Monad.Identity

data Tree a = Null | Node (Tree a) a (Tree a) deriving Show
main :: IO ()
main = do 
    print $ runTraverse (Node Null 5 Null)


type MyMon a = StateT (Vector Int) Identity a

runTraverse :: Tree Int -> ((),Vector Int)
runTraverse t =  runIdentity (runStateT (traverse t) (Data.Vector.replicate 7 0))

traverse :: Tree Int -> MyMon ()
traverse Null = return ()
traverse (Node l v r) = do
    s <- get
    put (s // [(v, (s ! v) + 1)]) -- s[v] := s[v] + 1
    traverse l
    traverse r
    return ()

但是不可变向量的“更新”是以 O(n) 复杂度完成的。 我正在寻找 O(1) 中的更新和 O(1) 中的访问。 据我了解，可变向量做我想做的事。要使用它们，我需要使用 ST 或 IO。 因为我想做一些单元测试，所以我更喜欢 ST monad，但我不想在函数调用中传递那个向量。 我需要继续使用 Monad Transformers，因为我将添加 ErrorT 和 WriterT 等转换器。

问题：如何使用 Monad Transformers 将可变向量放入 State Monad？

我想出了以下无法编译的代码：

import Data.Vector
import Control.Monad.State
import Control.Monad.Identity
import qualified Data.Vector.Mutable as VM
import Control.Monad.ST
import Control.Monad.ST.Trans
type MyMon2 s a = StateT (VM.MVector s Int) (STT s Identity) a

data Tree a = Null | Node (Tree a) a (Tree a) deriving Show
main :: IO ()
main = do 
    print $ runTraverse (Node Null 5 Null)

runTraverse :: Tree Int -> ((),Vector Int)
runTraverse t = runIdentity (Control.Monad.ST.Trans.runST $ do
        emp <- VM.replicate 7 0
        (_,x) <- (runStateT (traverse t) emp)
        v <- Data.Vector.freeze x
        return ((), v)
    )
traverse :: Tree Int -> MyMon2 s ()
traverse Null = return ()
traverse (Node l v r) = do
    d <- get
    a <- (VM.read d v)
    VM.write d v (a + 1)
    put d
    return ()

编译错误是：

TranformersExample: line 16, column 16:
  Couldn't match type `s'
                  with `primitive-0.5.2.1:Control.Monad.Primitive.PrimState
                          (STT s Identity)'
      `s' is a rigid type variable bound by
          a type expected by the context: STT s Identity ((), Vector Int)
          at test/ExecutingTest.hs:15:30
    Expected type: STT s Identity (MVector s Int)
      Actual type: STT
                     s
                     Identity
                     (MVector
                        (primitive-0.5.2.1:Control.Monad.Primitive.PrimState
                           (STT s Identity))
                        Int)
    In the return type of a call of `VM.new'
    In a stmt of a 'do' block: emp <- VM.new 7
    In the second argument of `($)', namely
      `do { emp <- VM.new 7;
            (_, x) <- (runStateT (traverse t) emp);
            v <- freeze x;
            return ((), v) }'
TranformersExample: line 26, column 14:
  Couldn't match type `s'
                  with `primitive-0.5.2.1:Control.Monad.Primitive.PrimState
                          (StateT (MVector s Int) (STT s Identity))'
      `s' is a rigid type variable bound by
          the type signature for traverse :: Tree Int -> MyMon2 s ()
          at test/ExecutingTest.hs:21:13
    Expected type: MVector
                     (primitive-0.5.2.1:Control.Monad.Primitive.PrimState
                        (StateT (MVector s Int) (STT s Identity)))
                     Int
      Actual type: MVector s Int
    In the first argument of `VM.write', namely `d'
    In a stmt of a 'do' block: VM.write d v (a + 1)
    In the expression:
      do { d <- get;
           a <- (VM.read d v);
           VM.write d v (a + 1);
           put d;
           .... }

注意：我知道不检查边界。

Answer 1

当使用ST 状态时，您从不明确地传递向量（它总是隐藏在s 参数中），而是对它的引用。该引用是不可变的且不会被复制，因此您不需要State 而只需一个读者即可隐式传递它。

import Data.Vector
import Control.Monad.Reader
import qualified Data.Vector.Mutable as VM
import Control.Monad.ST

type MyMon3 s = ReaderT (VM.MVector s Int) (ST s)

data Tree a = Null | Node (Tree a) a (Tree a) deriving Show
main :: IO ()
main = do 
    print $ runTraverse (Node Null 5 Null)

runTraverse :: Tree Int -> Vector Int
runTraverse t = runST $ do
        emp <- VM.replicate 7 0
        runReaderT (traverse t) emp
        Data.Vector.freeze emp

traverse :: Tree Int -> MyMon3 s ()
traverse Null = return ()
traverse (Node l v r) = do
    d <- ask
    a <- lift $ VM.read d v
    lift $ VM.write d v (a + 1)