http://www.haskell.org/haskellwiki/State_Monad的具体示例非常有助于理解如何使用monad编写实际代码(另请参阅stackoverflow / 9014218)。但是我们大多数新生来自OO背景,因此将OO程序映射到haskell将有助于演示如何编写等效的haskell代码。 (是的,这两种范式完全不同,将OO风格的代码直接翻译成haskell是不明智的,但这只是一次作为教程。)
这是一个OO风格的代码,它创建一个对象的2个实例,然后调用修改其各自成员变量的成员函数,最后打印它们。我们如何使用haskell state monads写这个?
class A:
int p;
bool q;
A() { p=0; q=False;} // constructor
int y() { // member function
if(q) p++; else p--;
return p;
}
bool z() { // member function
q = not q;
return q;
}
main:
// main body - creates instances and calls member funcs
a1 = A; a2 = A; // 2 separate instances of A
int m = a1.y();
m = m + a1.y();
bool n = a2.z();
print m, n, a1.p, a1.q, a2.p, a2.q;
答案 0 :(得分:5)
直接翻译类似于:
module Example where
import Control.Monad.State
data A = A { p :: Int, q :: Bool }
-- constructor
newA :: A
newA = A 0 False
-- member function
y :: State A Int
y = do
b <- getQ
modifyP $ if b then (+1) else (subtract 1)
getP
-- member function
z :: State A Bool
z = do
b <- gets q
modifyQ not
getQ
main :: IO ()
main = do
let (m,a1) = flip runState newA $ do
m <- y
m <- (m +) `fmap` y
return m
let (n,a2) = flip runState newA $ do
n <- z
return n
print (m, n, p a1, q a1, p a2, q a2)
-- general purpose getters and setters
getP :: State A Int
getP = gets p
getQ :: State A Bool
getQ = gets q
putP :: Int -> State A ()
putP = modifyP . const
putQ :: Bool -> State A ()
putQ = modifyQ . const
modifyP :: (Int -> Int) -> State A ()
modifyP f = modify $ \a -> a { p = f (p a) }
modifyQ :: (Bool -> Bool) -> State A ()
modifyQ f = modify $ \a -> a { q = f (q a) }
我可能不会打扰手动getter / setter而只使用lenses。
{-# LANGUAGE TemplateHaskell, FlexibleContexts #-}
module Main where
import Control.Applicative
import Control.Monad.State
import Data.Lenses
import Data.Lenses.Template
data A = A { p_ :: Int, q_ :: Bool } deriving Show
$( deriveLenses ''A )
-- constructor
newA :: A
newA = A 0 False
-- member function
y :: MonadState A m => m Int
y = do
b <- q get
if b then p $ modify (+1) else p $ modify (subtract 1)
p get
-- member function
z :: MonadState A m => m Bool
z = do
q $ modify not
q get
data Main = Main { a1_ :: A, a2_ :: A, m_ :: Int, n_ :: Bool } deriving Show
$( deriveLenses ''Main )
main :: IO ()
main = do
-- main body - creates instances and calls member funcs
print $ flip execState (Main undefined undefined undefined undefined) $ do
a1 $ put newA ; a2 $ put newA -- 2 separate instances of A
m . put =<< a1 y
m . put =<< (+) <$> m get <*> a1 y
n . put =<< a2 z
但这不是我真正写的东西,因为我正在向后弯曲Haskell以试图模仿OO风格。所以它看起来很尴尬。
对我来说,面向对象代码的真正目的是编程到一个接口。当我使用这些类型的对象时,我可以依赖它们来支持这些方法。所以,在haskell中,我会使用类型类来做到这一点:
{-# LANGUAGE TemplateHaskell, FlexibleContexts #-}
module Main where
import Prelude hiding (lookup)
import Control.Applicative
import Control.Monad.State
import Data.Lenses
import Data.Lenses.Template
import Data.Map
class Show a => Example a where
-- constructor
new :: a
-- member function
y :: MonadState a m => m Int
-- member function
z :: MonadState a m => m Bool
data A = A { p_ :: Int, q_ :: Bool } deriving Show
$( deriveLenses ''A )
instance Example A where
new = A 0 False
y = do
b <- q get
if b then p $ modify (+1) else p $ modify (subtract 1)
p get
z = do
q $ modify not
q get
data B = B { v_ :: Int, step :: Map Int Int } deriving Show
$( deriveLenses ''B )
instance Example B where
new = B 10 . fromList $ zip [10,9..1] [9,8..0]
y = v get
z = do
i <- v get
mi <- lookup i `liftM` gets step
case mi of
Nothing -> return False
Just i' -> do
v $ put i'
return True
data Main a = Main { a1_ :: a, a2_ :: a, m_ :: Int, n_ :: Bool } deriving Show
start :: Example a => Main a
start = Main undefined undefined undefined undefined
$( deriveLenses ''Main )
run :: Example a => State (Main a) ()
run = do
-- main body - creates instances and calls member funcs
a1 $ put new ; a2 $ put new -- 2 separate instances of a
m . put =<< a1 y
m . put =<< (+) <$> m get <*> a1 y
n . put =<< a2 z
main :: IO ()
main = do
print $ flip execState (start :: Main A) run
print $ flip execState (start :: Main B) run
现在,我可以针对不同类型run
和A
重复使用相同的B
。
答案 1 :(得分:4)
State
monad不能用于模拟类。它用于模拟“粘贴”到您正在运行的代码的状态,而不是“独立”状态并驻留在面向对象的类中。
如果您不想要方法重写和继承,那么最接近Haskell中的OOP类就是使用具有相关函数的记录。在这种情况下你必须要注意的唯一区别是所有“类方法”都返回新的“对象”,它们不会修改旧的“对象”。
例如:
data A =
A
{ p :: Int
, q :: Bool
}
deriving (Show)
-- Your "A" constructor
newA :: A
newA = A { p = 0, q = False }
-- Your "y" method
y :: A -> (Int, A)
y a =
let newP = if q a then p a + 1 else p a - 1
newA = a { p = newP }
in (newP, newA)
-- Your "z" method
z :: A -> Bool
z = not . q
-- Your "main" procedure
main :: IO ()
main =
print (m', n, p a1'', q a1'', p a2, q a2)
where
a1 = newA
a2 = newA
(m, a1') = y a1
(temp, a1'') = y a1'
m' = m + temp
n = z a2
此程序打印:
(-3,True,-2,False,0,False)
请注意,我们必须创建新变量来存储m
和a1
的新版本(我每次最后都添加了'
)。 Haskell没有语言级可变变量,因此您不应该尝试使用该语言。
可以通过使用IO引用来创建可变变量。
但请注意,以下代码在Haskellers中被认为是非常糟糕的编码风格。如果我是一名老师并且有一个学生编写这样的代码,我就不会给作业通过成绩;如果我雇用一个编写这样代码的Haskell程序员,如果他没有很好的理由来编写这样的代码,我会考虑解雇他。
import Data.IORef -- IO References
data A =
A
{ p :: IORef Int
, q :: IORef Bool
}
newA :: IO A
newA = do
p' <- newIORef 0
q' <- newIORef False
return $ A p' q'
y :: A -> IO Int
y a = do
q' <- readIORef $ q a
if q'
then modifyIORef (p a) (+ 1)
else modifyIORef (p a) (subtract 1)
readIORef $ p a
z :: A -> IO Bool
z = fmap not . readIORef . q
main :: IO ()
main = do
a1 <- newA
a2 <- newA
m <- newIORef =<< y a1
modifyIORef m . (+) =<< y a1
n <- z a2
m' <- readIORef m
pa1 <- readIORef $ p a1
qa1 <- readIORef $ q a1
pa2 <- readIORef $ p a2
qa2 <- readIORef $ q a2
print (m', n, pa1, qa1, pa2, qa2)
该程序与上述程序的功能相同,但具有可变变量。同样,除了非常罕见的情况外,不要写这样的代码。