3

我必须解析一些 xml，我决定使用 xml-conduit 来完成该任务并使用它的流式传输部分。

xml 的结构由包含元素及其出现频率的 xsd 文件给出。但不是他们预期的顺序。

如何使用 xml 结构的子级解析所有可能的重新排序Text.XML.Stream.Parse？

问题

假设我们有一个 xml 描述，例如

    Root 
    /  \
   A    B

那么两者<Root><A>atext</A><B>btext</B></Root>和 <Root><B>btext</B><A>atext</A></Root>都是此 xml 结构的有效实例。但是在流设置中解析需要一个顺序才能成功。

我想过使用类似的东西，parseRoot1 <|> parseRoot2但后来我必须实现Alternative实例并手动编写所有可能性，我真的不想这样做。

这是一个最小的示例 haskell 程序。

Example.hs

{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE Rank2Types #-}

module Main where

import           Control.Exception
import           Control.Monad.Trans.Resource (MonadThrow)
import           Text.XML.Stream.Parse
import           Data.Monoid ((<>))
import           Data.Maybe
import           Data.Text (Text)
import           Data.XML.Types (Event)
import           Data.Conduit (ConduitM, Consumer, yield, ($=), ($$))

data Root = Root {a :: A, b :: B} deriving (Show, Eq)
data A = A Text deriving (Show, Eq)
data B = B Text deriving (Show, Eq)

ex1, ex2 :: Text
ex1 = "<Root>"<>
        "<A>Atest</A>"<>
        "<B>Btest</B>"<>
      "</Root>"
ex2 = "<Root>"<>
        "<B>Btest</B>"<>
        "<A>Atest</A>"<>
      "</Root>"

ex :: Root
ex = Root {a = A "Atest", b = B "Btest"}

parseA :: MonadThrow m => ConduitM Event o m (Maybe A)
parseA = tagIgnoreAttrs "A"
            $ do result <- content
                 return (A $ result)

parseB :: MonadThrow m => ConduitM Event o m (Maybe B)
parseB = tagIgnoreAttrs "B"
            $ do result <- content
                 return (B result)

parseRoot1 :: MonadThrow m => ConduitM Event o m (Maybe Root)
parseRoot1 = tagIgnoreAttrs "Root" $ do
                 a' <- fromMaybe (error "error parsing A") <$> parseA
                 b' <- fromMaybe (error "error parsing B") <$> parseB
                 return $ Root{a = a', b = b'}

parseRoot2 :: MonadThrow m => ConduitM Event o m (Maybe Root)
parseRoot2 = tagIgnoreAttrs "Root" $ do
                 b' <- fromMaybe (error "error parsing B") <$> parseB
                 a' <- fromMaybe (error "error parsing A") <$> parseA
                 return $ Root{a = a', b = b'}

parseTxt :: Consumer Event (Either SomeException) (Maybe a)
                          -> Text 
                          -> Either SomeException (Maybe a)
parseTxt p inTxt = yield inTxt
                  $= parseText' def
                  $$ p

main :: IO ()
main = do putStrLn "Poor Mans Test Suite"
          putStrLn "===================="
          putStrLn "test1 Root -> A - B " -- works
          print $ parseTxt parseRoot1 ex1
          putStrLn "test1 Root -> B - A " -- fails
          print $ parseTxt parseRoot1 ex2
          putStrLn "test2 Root -> A - B " -- fails
          print $ parseTxt parseRoot2 ex1
          putStrLn "test2 Root -> B - A " -- works again
          print $ parseTxt parseRoot2 ex2

笔记

example.cabal

[...]
  build-depends: base >=4.8 && <4.9
               , conduit
               , resourcet
               , text
               , xml-conduit
               , xml-types
[...]
4

1 回答 1

1

这是我的想法...

首先是一些定义：

 {-# LANGUAGE OverloadedStrings, MultiWayIf #-}

 import Control.Monad.Trans.Resource
 import Data.Conduit
 import Data.Text (Text, unpack)
 import Data.XML.Types
 import Text.XML.Stream.Parse

 data SumType = A Text | B Text | C Text

我们从一个接受 A 或 B 标签的管道开始，忽略属性并返回名称和内容：

 parseAorB :: MonadThrow m => ConduitM Event o m (Maybe (Name, Text))
 parseAorB =
   tag (\n -> if (n == "A" || n == "B") then Just n else Nothing) -- accept either A or B
       (\n -> return n)                                           -- ignore attributes
       (\n -> do c <- content; return (n,c))                      -- extract content

然后我们用它来编写一个解析两个标签的管道，确保一个是 A，另一个是 B：

 parseAB :: MonadThrow m => ConduitM Event o m (Maybe (SumType, SumType))
 parseAB = do
   t1 <- parseAorB
   case t1 of
     Nothing      -> return Nothing
     Just (n1,c1) -> do
       t2 <- parseAorB
       case t2 of
         Nothing -> return Nothing
         Just (n2,c2) -> do
           if | "A" == n1 && "B" == n2 -> return $ Just (A c1, B c2)
              | "A" == n2 && "B" == n1 -> return $ Just (A c2, B c1)
              | otherwise              -> return Nothing

更新

parseAB您可以使用MaybeT变压器减少样板：

import Control.Monad.Trans.Maybe
import Control.Monad.Trans

parseAB' :: MonadThrow m => MaybeT (ConduitM Event o m) (SumType, SumType)
parseAB' = do
 (n1, c1) <- MaybeT parseAorB
 (n2, c2) <- MaybeT parseAorB
 if | "A" == n1 && "B" == n2 -> return (A c1, B c2)
    | "A" == n2 && "B" == n1 -> return (A c2, B c1)
    | otherwise              -> MaybeT $ return Nothing

如果你有几个构造函数，我会考虑做这样的事情：

allkids = do
  kids <- many parseAorB
  let sorted = sort kids -- automatically sorts by name
  if map fst kids == [ "A", "B", "C", "D", "E", "F", "G", "H"]
    then let [ca, cb, cc, cd, ce, cf, cg, ch] = map snd kids
         in return (A ca, B cb, C cc, D cd, E ce, F cf, G cg, H ch)
    else ...error...

组合器many来自Tet.XML.Stream.Parse。

于 2015-11-10T18:47:42.017 回答