You are suggesting later in this thread that the old ListT could be used to solve my initial problem. I don't see how, so I'm wondering if you'd mind sorting some things out for me? On Tue, Jan 23, 2007 at 11:59:58 +0200, Yitzchak Gale wrote: [..]
test = do a <- liftIO getChar guard $ a /= 'q' return $ a `mplus` test
This piece has type problems. I couldn't get ghci to accept it without making some changes: test :: ListT IO Char test = do a <- liftIO getChar guard $ a /= 'q' (return a) `mplus` test [..]
test2 = runListT . mapM (liftIO print)
:t liftIO
:t print
There are some type problems here too, and I don't really see how to fix them. liftIO :: (MonadIO m) => IO a -> m a print :: (Show a) => a -> IO ()
:t (liftIO print) Couldn't match expected type `IO a' against inferred type `a1 -> IO ()' In the first argument of `liftIO', namely `print'
I also don't quite see how 'mapM foo` can be applied to a ListT since 'mapM foo' would have the type [a] -> m [b] (a & b depending on foo). Perhaps you didn't mean mapM but rather mapListT? But using mapListT in a similar way:
runListT.mapListT (\ m -> m >>= mapM putChar) $ test
does not result in the desired interleaving. Changing to using runListT first also gives the wrong behaviour:
runListT test >>= mapM_ putChar
I am really confused about your statement that ListT as found in GHC 6.6 can be used to solve my problem. I'd like to avoid using unsafeInterleaveIO if there's a nice solution for it. IMHO a solution with ListT would be nice, so I'd really like to understand how to make it work the way I want it. /M -- Magnus Therning (OpenPGP: 0xAB4DFBA4) magnus@therning.org Jabber: magnus.therning@gmail.com http://therning.org/magnus