Haskell 包 hxt 的用法对我来说还是有点奇怪。尤其是箭头符号和产生的类型是一种魔法。
到目前为止，我无法管理以下内容：我想处理一个主要包含两个部分的 XML 文件。一个保留对象的定义，第二个保留对象的用途/目的。首先，我想编写一些 hxt 处理以在第 1 部分上获取 Haskell 数据结构，然后在第 2 部分处理，最后结合在程序的真实逻辑中读取的两种数据结构。
由于箭头教程，现在处理文件通常很好。但我希望现在有一个 do notation 执行三个步骤：读取文档（惰性），用第一个处理器处理一次结果结构，然后用第二个处理器再次处理相同的结构。我不希望像以下示例中那样调用“readDocument”两次。
import Text.XML.HXT.Core
import Data.Char(toUpper)
import Data.Tree.NTree.TypeDefs
play filename = do
results <- runX (getAllAddresses filename)
results2 <- runX (getAllAddressesUsages filename)
print results
print results2
getAllAddresses :: FilePath -> IOSArrow XmlTree [(String,NTree XNode)]
getAllAddresses filename =
readDocument [withValidate no] filename >>>
getChildren >>>
isElem >>> hasName "main" >>>
getChildren >>>
isElem >>> hasName "part1" >>>
getChildren >>>
isElem >>> hasName "address" >>>
listA(getAddress) -- create a list for each variable, so use listA
getAddress :: IOSArrow XmlTree (String,NTree XNode)
getAddress =
getChildren >>>
isElem >>>
(
neg ( hasName "location") >>> -- all elements being no "location"
getName &&& (getChildren) -- get the name and the value for each element
)
<+>
(
hasName "location" >>> -- work on all nodes within the "location" subcontainer
getChildren >>>
isElem >>>
( getName &&& (getChildren) ) -- get the name and the value for each element
)
getAllAddressesUsages :: FilePath -> IOSArrow XmlTree [(String,NTree XNode)]
getAllAddressesUsages filename =
readDocument [withValidate no] filename >>>
getChildren >>>
isElem >>> hasName "main" >>>
getChildren >>>
isElem >>> hasName "part2" >>>
getChildren >>>
listA(getAddressUsagePurpose2) -- create a list for each variable, so use listA
getAddressUsagePurpose2 :: IOSArrow XmlTree (String,NTree XNode)
getAddressUsagePurpose2 =
hasName "use_obj-names_for_purpose_2" >>> -- work on all nodes with usage 2
( getName &&& (getChildren) ) -- get the name and the value for each element
示例数据：
<main>
<part1>
<address>
<obj-name>one</obj-name>
<name>peter 1</name>
<street>streetname 1</street>
<location>
<country>Germany</country>
<state>Baden Wuerttemberg</state>
</location>
</address>
<address>
<obj-name>two</obj-name>
<name>peter 2</name>
<street>streetname 2</street>
<location>
<country>Germany</country>
<state>Nordrhein Westfalen</state>
</location>
</address>
</part1>
<part2>
<use_obj-names_for_purpose_1>
<obj-name>two</obj-name>
</use_obj-names_for_purpose_1>
<use_obj-names_for_purpose_2>
<obj-name>two</obj-name>
</use_obj-names_for_purpose_2>
</part2>
</main>
所以正式的问题是：
为了得到这样的东西，monadic 在函数 play 中的表现如何：
readXmlDocument :: String -> IOSArrow XmlTree (NTree XNode)
readXmlDocument filename = readDocument [withValidate no] filename
play filename = do
document <- readXmlDocument filename
allAddresses <- getAllAddresses document
allPurposes <- getAllAddressesUsages document
result <- processLogics allAddresses allPurposes
print result
我如何从 Monads 到 Arrows，再回到 Monads，再到纯数据再回到 Monads。
我为什么要这样做？