Streaming xml-conduit parse results

Question

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

Answer 1

Let me start by saying that the streaming helper API in xml-conduit has not be worked on in years, and could probably benefit from a reimagining given changes that have happened to conduit in the interim. I think there are likely much better ways to accomplish things.

That said, let me explain the problem you're seeing. The many function creates a list of results, and will not produce any values until it has finished processing. In your case, there are so many values that this appears to never happen. Ultimately, when the entire file has been read, the entire list of users will be displayed at once. But that's clearly not the behavior you're looking for.

Instead, what you want to do is create a stream of User values which are produced as soon as they're ready. What you want to do is basically replace the many function call with a new function which will yield a result each time it's parsed. A simple implementation of this could be:

yieldWhileJust :: Monad m
               => ConduitM a b m (Maybe b)
               -> Conduit a m b
yieldWhileJust consumer =
    loop
  where
    loop = do
        mx <- consumer
        case mx of
            Nothing -> return ()
            Just x -> yield x >> loop

Also, instead of using putStrLn $ unlines $ map show, you want to attach the entire pipeline to a consumer which will print each individually yielded User value. This can be implemented easily with Data.Conduit.List.mapM_, e.g.: CL.mapM_ (liftIO . print).

I've put together a full example based on your code. The input is an artificially generated infinite XML file, just to prove the point that it really is yielding output immediately.

{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE RankNTypes        #-}
import           Control.Applicative    ((<*))
import           Control.Concurrent     (threadDelay)
import           Control.Monad          (forever, void)
import           Control.Monad.IO.Class (MonadIO (liftIO))
import           Data.ByteString        (ByteString)
import           Data.Conduit
import qualified Data.Conduit.List      as CL
import           Data.Text              (Text)
import           Data.Text.Encoding     (encodeUtf8)
import           Data.XML.Types         (Event)
import           Text.XML.Stream.Parse

-- instead of actually including a large input data file, just for testing purposes
infiniteInput :: MonadIO m => Source m ByteString
infiniteInput = do
    yield ""
    forever $ do
        yield $ encodeUtf8
            ""
        liftIO $ threadDelay 1000000
    --yield "" -- will never be reached

data User = User {name :: Text} deriving (Show)

parseUserRow :: MonadThrow m => Consumer Event m (Maybe User)
parseUserRow = tagName "row" (requireAttr "DisplayName" <* ignoreAttrs) $ \displayName -> do
    return $ User displayName

parseUsers :: MonadThrow m => Conduit Event m User
parseUsers = void $ tagNoAttr "users" $ yieldWhileJust parseUserRow

yieldWhileJust :: Monad m
               => ConduitM a b m (Maybe b)
               -> Conduit a m b
yieldWhileJust consumer =
    loop
  where
    loop = do
        mx <- consumer
        case mx of
            Nothing -> return ()
            Just x -> yield x >> loop

main :: IO ()
main = infiniteInput
    $$ parseBytes def
    =$ parseUsers
    =$ CL.mapM_ print