如何简化此模式匹配?

时间:2015-08-10 03:45:18

标签: haskell pattern-matching logic

我试图在Haskell中创建一个命题逻辑模型,我需要一个函数将一些逻辑规则应用于特定的子表达式。功能"适用"获取一个列表,该列表指示树中子表达式的位置(根据左右序列),逻辑规则和逻辑表达式,并返回一个新的逻辑表达式。

data LogicExp  a = P a                              | 
                     True'                      | 
                     False'                                 | 
                     Not' (LogicExp a)                  |  
                     (LogicExp a) :&  (LogicExp a)  | 
                     (LogicExp a) :|  (LogicExp a)  | 
                     (LogicExp a) :=> (LogicExp a)    |
                     (LogicExp a) :=  (LogicExp a)
    deriving Show


type LExp = LogicExp String

data Position = L | R

deMorgan :: LExp -> LExp
deMorgan (e1 :& e2) = Not' ((Not e1) :| (Not e2))
deMorgan (e1 :| e2) = Not' ((Not e1) :& (Not e2))
deMorgan x = x

apply :: [Position] -> (LExp -> LExp) -> LExp -> LExp
apply [] f e = f e
apply (L:xs) f (e1 :& e2) = (apply xs f e1) :& e2
apply (R:xs) f (e1 :& e2) = e1 :& (apply xs f e2)
apply (L:xs) f (e1 :| e2) = (apply xs f e1) :| e2
apply (R:xs) f (e1 :| e2) = e1 :| (apply xs f e2)
apply (L:xs) f (e1 :=> e2) = (apply xs f e1) :=> e2
apply (R:xs) f (e1 :=> e2) = e1 :=> (apply xs f e2)
apply (L:xs) f (e1 := e2) = (apply xs f e1) := e2
apply (R:xs) f (e1 := e2) = e1 := (apply xs f e2)
apply (x:xs) f (Not' e) = apply xs f e

该功能正常。但是我可以使用一些数据构造函数" wildcard"有这样一个更简单的功能吗?

apply :: [Position] -> (LExp -> LExp) -> LExp -> LExp
apply [] f e = f e
apply (L:xs) f (e1 ?? e2) = (apply xs f e1) ?? e2
apply (R:xs) f (e1 ?? e2) = e1 ?? (apply xs f e2)
apply (x:xs) f (Not' e) = apply xs f e

2 个答案:

答案 0 :(得分:7)

目前,我无法回想起任何花哨的技巧。但是,您可能想要做的一件事是将LogicExp构造函数中的公共结构分解出来:

data LogicExp a
    = P a
    | True'
    | False'
    | Not' (LogicExp a) 
    | Bin' BinaryOp (LogicExp a) (LogicExp a)
    deriving Show

data BinaryOp = And' | Or' | Impl' | Equiv'
    deriving Show
apply :: [Position] -> (LExp -> LExp) -> LExp -> LExp
apply [] f e = f e
apply (L:xs) f (Bin' op e1 e2) = Bin' op (apply xs f e1) e2
apply (R:xs) f (Bin' op e1 e2) = Bin' op e1 (apply xs f e2)
apply (x:xs) f (Not' e) = apply xs f e
-- ... and the P, True' and False' cases.

通过这样做,你失去了可爱的中缀构造函数。但是,如果你真的想要它们,那就有一个奇特的技巧:view patterns(另请参阅this question了解更多示例和讨论)。

答案 1 :(得分:4)

这是使用其中一个Generics软件包的经典案例,sybuniplate

通常uniplate更快,但不如syb。幸运的是,在这种情况下,您可以使用uniplate

使用uniplate的步骤:

  1. 使用DeriveDataTypeable pragma。
  2. 自动派生DataTypeable
  3. 导入Data.Data以及类似Data.Generics.Uniplate.Data
  4. 的联合模块

    您想要的转换函数只是transform,具有相应的类型签名:

    doit :: LExp -> LExp
    doit = transform deMorgan
    

    其中deMorgan与您编写的完全相同。

    完整示例:

    {-# LANGUAGE DeriveDataTypeable #-}
    module Lib6 where
    
    import Data.Data
    import Data.Generics.Uniplate.Data
    import Text.Show.Pretty (ppShow)
    
    data LogicExp  a = P a                              |
                         True'                      |
                         False'                                 |
                         Not' (LogicExp a)                  |
                         (LogicExp a) :&  (LogicExp a)  |
                         (LogicExp a) :|  (LogicExp a)  |
                         (LogicExp a) :=> (LogicExp a)    |
                         (LogicExp a) :=  (LogicExp a)
        deriving (Show, Data, Typeable)
    
    type LExp = LogicExp String
    
    data Position = L | R
    
    deMorgan :: LExp -> LExp
    deMorgan (e1 :& e2) = Not' ((Not' e1) :| (Not' e2))
    deMorgan (e1 :| e2) = Not' ((Not' e1) :& (Not' e2))
    deMorgan x = x
    
    doit :: LExp -> LExp
    doit = transform deMorgan
    
    example = (P "a" :& P "b") :| (P "c")
    
    test = putStrLn $ ppShow (doit example)
    

    运行test会产生:

    Not' (Not' (Not' (Not' (P "a") :| Not' (P "b"))) :& Not' (P "c"))
    

    关于uniplate的简介:

    http://community.haskell.org/~ndm/darcs/uniplate/uniplate.htm