Counting and filtering Arrow for HXT

I'm trying to parse an XML, but I want to filter and extract only a determinate number of children from a given node. For example:

<root>
    <node id="a" />
    <node id="b" />
    <node id="c" />
    <node id="d" />
</root>

And then if I execute the arrow getChildren >>> myFilter 2, I would get back only the nodes with id "a" and "b".

Intuition gives that I should use a State Arrow to keep track, but I don't know how to do that.

I tried to do it myself but it's not exactly what I want, doesn't look very elegant, and doesn't work. I try to run my chain of arrows with runSLA and a integer parameter as initial state, and then defining:

takeOnly :: IOSLA Int XmlTree XmlTree
takeOnly = changeState (\s b -> s-1)
             >>> accessState (\s b -> if s >= 0 then b else Nothing)

But of course I can't return Nothing, I need to return a XmlTree. But I don't want to return anything at all!

There's probably a better way out there. Can you help me?

Thanks for your time and help!

It would probably be more idiomatic to use the combinators in Control.Arrow.ArrowList to handle this kind of thing.

The package specifically provides (>>.) :: a b c -> ([c] -> [d]) -> a b d, which is a "combinator for converting the result of a list arrow into another list". This allows us to use the take function that we already have for lists in this context.

Here's a quick version of how you might use it:

module Main where

import Text.XML.HXT.Arrow

takeOnly :: (ArrowXml a) => Int -> a XmlTree XmlTree
takeOnly n = getChildren >>. take n 

main = do
  let xml = "<root><node id='a' /><node id='b' />\
                  \<node id='c' /><node id='d' /></root>"

  print =<< runX (readString [] xml >>> getChildren >>> takeOnly 2)

This I believe does approximately what you're looking for:

travis@sidmouth% ./ArrowTake
[NTree (XTag (LP node) [NTree (XAttr (LP id)) [NTree (XText "a") []]]) [],
 NTree (XTag (LP node) [NTree (XAttr (LP id)) [NTree (XText "b") []]]) []]

No IOSLA required. Note that I've also changed the function type a little—this version seems nicer to me, but you could easily convert it to something more like the type in your version.

Recommended topics

Hot tags