I want to use xml-conduit
, specifically Text.XML.Stream.Parse
in order to lazily extract a list of objects from a large XML file.
As a test case, I use the recently re-released StackOverflow data dumps. To keep it simple, I intend to extract all usernames from stackoverflow.com-Users.7z
. Even if the file is a .7z
, file
says it is just bzip2-compressed data (there might be some 7zip stuff at the end of the file, but right now I don't care).
A simplified version of the XML would be
<users>
<row id="1" DisplayName="StackOverflow"/>
...
<row id="2597135" DisplayName="Uli Köhler"/>
...
</users>
Based on this previous Q&A and the example on Hackage stream-reading the example XML in bz2-ed form works perfectly for me
However, when using runghc
to run the following program, it runs without printing any output:
{-# LANGUAGE OverloadedStrings #-}
import Data.Conduit (runResourceT, ($$), ($=))
import qualified Data.Conduit.Binary as CB
import Data.Conduit.BZlib
import Data.Conduit
import Data.Text (Text)
import System.IO
import Text.XML.Stream.Parse
import Control.Applicative ((<*))
data User = User {name :: Text} deriving (Show)
parseUserRow = tagName "row" (requireAttr "DisplayName" <* ignoreAttrs) $ \displayName -> do
return $ User displayName
parseUsers = tagNoAttr "users" $ many parseUserRow
main = do
users <- runResourceT $ CB.sourceFile "stackoverflow.com-Users.7z" $= bunzip2 $= parseBytes def $$ force "users required" parseUsers
putStrLn $ unlines $ map show users
I assume this issue occurs because Haskell tries to deeply evaluate the users
list before starting to print it. This theory is supported by the memory usage of the program continually growing about 2 percent per second (source: htop).
How can I "live-stream" the results to stdout? I assume this is possible by adding another conduit statement like $$ CB.sinkFile "output.txt"
at the end. This specific version however expects a Conduit
output of ByteString
. Could you point me in the right direction where to go from here?
Any help will be appreciated!
yield
the value fromrequireAttr
, you should do aT.copy
on it to ensure the larger text it came from can be GC'd – Hoogh