5

我有一些评估原始程序的代码。程序是一个语句列表(表达式、块、返回语句)。评估结果是最后评估的表达式。评估者也应该正确对待return陈述(即在第一次出现后停止评估return)。

为了实现这个逻辑,我传递了特殊的回调函数 ( NextStep),它在当前语句之后进行下一步评估。处理返回语句时我不调用下一步:

data Statement = 
      Expr Int
    | Block [Statement]
    | Return Int
    deriving (Show, Eq)

data Value = 
      Undefined
    | Value Int
    deriving (Show, Eq)

type NextStep = Value -> Value

evalStmt :: Statement -> NextStep -> Value
evalStmt (Expr val) next = 
    let res = Value val
    in next res
evalStmt (Block stmts) next = evalBlock stmts next
evalStmt (Return val) next = Value val

evalBlock :: [Statement] -> NextStep -> Value
evalBlock [] next = next Undefined
evalBlock [st] next = evalStmt st next
evalBlock (st:rest) next = evalStmt st $ \ _ -> evalBlock rest next

evalProgram stmts = evalBlock stmts id

prog1 = [Expr 1, Block [Return 3, Expr 2], Expr 4] 
evalProg1 = evalProgram prog1 -- result will be Value 3

问题是如何用 continuation monad 重写这段代码?我想摆脱显式传递的NextStep回调evalStmtevalBlock函数。可能吗?

4

1 回答 1

7

翻译相当机械。

请记住,在延续单子中,return将值输入延续。

evalStmt :: Statement -> Cont Value Value
evalStmt (Expr val) = 
    let res = Value val
    in return res
evalStmt (Block stmts) = evalBlock stmts
evalStmt (Return val) = cont $ \_ -> Value val

evalBlock :: [Statement] -> Cont Value Value
evalBlock [] = return Undefined
evalBlock [st] = evalStmt st
evalBlock (st:rest) = evalStmt st >> evalBlock rest

evalProgram :: [Statement] -> Value
evalProgram stmts = runCont (evalBlock stmts) id

为了模拟早期回报,我们只是忽略给予的延续Return val,只返回我们拥有的价值。

于 2014-05-11T15:36:04.323 回答