- Haskell-Cafe - Haskell.org

How to combine Error and IO monads?
by Cat Dancer 13 Dec '06

13 Dec '06

I've read Jeff Newbern's tutorial on monad transformers (http://www.nomaware.com/monads/html/index.html) but I don't grok it yet and I can't tell how to get started with this particular requirement, or even if I need monad transformers for this. I have a program that performs a series of IO operations, each which can result in an error or a value. If a step returns a value I usually want to pass that value on to the next step, if I get an error I want to do some error handling but usually want to skip the remaining steps. Thus I have a lot of functions with return types like IO (Either String x), where x might be (), Integer, or some other useful value type, and a lot of case statements like a :: Integer -> IO (Either String String) (a :: Either String Integer) <- some_io_action_returning_integer_or_error case a of -- got to get from (Either String Integer) to (Either String String) Left message -> return $ Left message -- continue on Right i -> do_more i If I was using just Either I know how to get rid of the case statements: import Control.Monad.Error p :: Integer -> Either String Integer q :: Integer -> Either String String r :: Either String String r = do a <- p 3 b <- q a return b which of course is much nicer. But how do I combine the Error and IO monads so that r could have the type IO (Either String String)?

4 6

Why so slow?
by Lyle Kopnicky 13 Dec '06

13 Dec '06

The code below is using way more RAM than it should. It seems to only take so long when I build the 'programs' list - the actual reading/parsing is fast. For a 5MB input file, it's using 50MB of RAM! Any idea how to combat this? Thanks, Lyle {-# OPTIONS_GHC -fglasgow-exts #-} -- linear_importer.hs import Control.Monad (unless) import Data.ByteString.Char8 (ByteString) import qualified Data.ByteString.Char8 as BS import Text.Regex as RE import Data.Time.Calendar import Data.Time.LocalTime import System.Environment -- Consider adding strictness as necessary for performance. -- STB time channel data ChannelChange = ChannelChange Int LocalTime Int deriving Show -- channel program_no start end network_no data ScheduleProgram = ScheduleProgram Int Int LocalTime LocalTime (Maybe Int) deriving Show main = do fileNames <- getArgs ensure (length fileNames == 2) usageMessage let [eventFileName,programFileName] = fileNames putStrLn ("Reading program schedule file from '" ++ programFileName ++ "'...") text <- BS.readFile programFileName programs <- sequence $ map parseScheduleProgram (BS.lines text) print (take 20 programs) return () usageMessage = "Usage: linear_importer <channel change file> <schedule file>" parseScheduleProgram :: ByteString -> IO ScheduleProgram parseScheduleProgram s = do let fields = BS.split '|' s ensure (length fields == 7) ("Wrong number of fields in schedule program: " ++ BS.unpack s) let [_,channelNoText,programNoText,_,startTimeText,endTimeText,networkNoText] = fields let channelNo = read $ BS.unpack channelNoText programNo = read $ BS.unpack programNoText startTime <- parseProgramTime startTimeText endTime <- parseProgramTime endTimeText let networkNo = if BS.null networkNoText then Nothing else Just (read (BS.unpack networkNoText)) return $ ScheduleProgram channelNo programNo startTime endTime networkNo parseProgramTime :: ByteString -> IO LocalTime parseProgramTime s = do let parts = BS.split 'T' s ensure (length parts == 2) ("Expected exactly one T in eventChannelChange time: " ++ BS.unpack s) let [datePart,timePart] = parts ensure (BS.length datePart == 8) ("Expected 8 digits in date part of eventChannelChange time: " ++ BS.unpack s) let (yearPart, monthDayPart) = BS.splitAt 4 datePart (monthPart, dayPart) = BS.splitAt 2 monthDayPart year = read $ BS.unpack yearPart month = read $ BS.unpack monthPart day = read $ BS.unpack dayPart let date = fromGregorian year month day ensure (BS.length timePart == 6) ("Expected 6 digits in time part of eventChannelChange time: " ++ BS.unpack s) let (hoursPart,minutesSecondsPart) = BS.splitAt 2 timePart (minutesPart,secondsPart) = BS.splitAt 2 minutesSecondsPart hours = read $ BS.unpack hoursPart minutes = read $ BS.unpack minutesPart seconds = read $ BS.unpack secondsPart let time = TimeOfDay hours minutes (fromInteger seconds) return (LocalTime date time) ensure :: Bool -> String -> IO () ensure x s = unless x $ ioError (userError s)

4 3

Ninety Nine Haskell Problems
by Andrew Wagner 13 Dec '06

13 Dec '06

Hello, fellow Haskellers! jcreigh, from #haskell, has found a great set of exercises to work through. Check out the new wiki pages indexed at http://haskell.org/haskellwiki/99_Haskell_exercises . These have all been done in lisp, but of COURSE they can be done at least as well in Haskell, if not better. This is a great way for relatively new haskell programmers to contribute something, since most of the problems are relatively simple. So have at them! The template is already there, there's existing lisp code, and examples; all you have to do is write the canonical haskell version. Please follow the instructions for 'reserving' a problem so we're not all duplicating each others' work. We should be able to knock these problems out in a matter of hours! Have fun! Andrew

2 1

Re: [Haskell] Waiting on multiple Chan-nels
by dons＠cse.unsw.edu.au 13 Dec '06

13 Dec '06

sebastian.setzer.ext: > Hi, > How do you wait on multiple channels, but read only from one of them > (the first that gets an entry)? Is there a library-function I missed > which already does this? > > What do you think of this solution? Feels to heavy :) An alternative solution that strikes me, is to use a single chan, with a sum type to tag the elements to be received by each thread. Then lazily generate a list of filtered values for each thread. Then the reading threads need do no actual IO to read. In particular, doing the chan reading IO under a lazy stream greatly simplifies the reader threads. For example: -- -- ghc chantest.hs -o chantest -- Usage: Enter "1", "2" or "0" to exit. -- import System.IO import Control.Concurrent import Control.Concurrent.Chan import Text.Printf -- -- Rather tha wait on multiple channels, it strikes me that you could -- have a single channel, and tag each thread's input. then lazily -- stream the chans contents through a filter, passing the *pure* list -- of filtered elements to a consuming thread -- type Pipe = Chan (Either String String) main :: IO () main = do chan <- newChan :: IO Pipe s <- getChanContents chan -- lazy list of chan elements c1Thread <- forkIO $ reader "c1" (catLeft s) -- read only Lefts c2Thread <- forkIO $ reader "c2" (catRight s) -- read only Rights writer chan where catLeft ls = [x | Left x <- ls] catRight ls = [x | Right x <- ls] writer :: Pipe -> IO () writer chan = loop where loop = getChar >>= command command '0' = print "done" command '1' = writeChan chan (Left "main: 1") >> loop command '2' = writeChan chan (Right "main: 2") >> loop command '\n' = loop -- ignore command c = printf "Illegal: %c\n" c >> loop reader :: String -> [String] -> IO () reader name xs = mapM_ (printf "%s %s\n" name) xs {- $ ghc x.hs -threaded $ ./a.out 1 c1 main: 1 2 c2 main: 2 2 c2 main: 2 1 c1 main: 1 2 c2 main: 2 3 Illegal: 3 1 c1 main: 1 2 c2 main: 2 0 "done" -} -- Don P.s. redirected to haskell-cafe(a)haskell.org, which is more appropriate for design discussion.

1 0

Name that function =)
by Louis J Scoras 13 Dec '06

13 Dec '06

Hi all; I'm new to the list, so I just wanted to say hi and -- of course -- ask a quick question: I have some IO actions that I want to map over a list of pairs -- these are just directory names and their down-cased versions. It wasn't difficult to actually get the right behavior by just doing mapM twice. -- putDirs just outputs something like "mv fst snd" mapM_ putDirs pairs mapM_ (uncurry renameFile) pairs This bothered me though, because I suspected that this could be done in one pass. Naively I proceeded to this. mapM_ (putDirs >> (uncurry renameFile)) pairs Which didn't work. I thought about it a little more before realizing that putDirs wouldn't get any parameters this way: I needed some way to distribute the pair over both operations. Here's the higher-order function I needed: foo h f g i = h (f i) (g i) which could then be curried and we get: mapM_ (foo (>>) putDirs $ uncurry renameFile) pairs Works great. So my question: is there a established name for foo? What about foo partially applied to (>>)? This was a fun exercise, but I'd like to use the standard implementations if they exist. Thanks! -- Lou.

4 5

Refactoring recklessly
by Greg Fitzgerald 12 Dec '06

12 Dec '06

I'd like to be able to reorganize my code and then verify that I didn't change any functionality. That is, the old and new code have precisely the same meaning. Also, I'd like to be able to change a function and verify that efficiency was the only thing affected. Are either of these possible in Haskell or any language? Thanks, Greg

5 4

Haskell and .NET
by Justin Bailey 12 Dec '06

12 Dec '06

I am learning Haskell, and I really want to be able to use .NET libraries from it. I've found the Hugs98.NET binary available from Galois, but it seems very, very stale. Is there anything more recent available? Or did I miss something? My motivation for this email is a recent series of posts[1] I saw on creating games using F# and XNA Express. If F# can do it, Haskell should be able to also. Any pointers are much appreciated! Justin [1] Part 1 - http://grammerjack.spaces.live.com/blog/cns!F2629C772A178A7C!156.entry Part 2 - http://grammerjack.spaces.live.com/blog/cns!F2629C772A178A7C!157.entry Part 3 - http://grammerjack.spaces.live.com/blog/cns!F2629C772A178A7C!158.entry

1 0

Re: [Haskell-cafe] Name that function =)
by Louis J Scoras 12 Dec '06

12 Dec '06

On 12/12/06, Conal Elliott <conal(a)conal.net> wrote: > Try foo = liftM2, and check out the recent haskell thread "Cannot understand > liftM2". That's actually what I reached for first. The problem is that the arguments aren't monads: *Main> :t (uncurry renameFile) (uncurry renameFile) :: (FilePath, FilePath) -> IO () *Main> :t putDirs putDirs :: (String, String) -> IO () and liftM2 works on monadic arguments: liftM2 :: (Monad m) => (a1 -> a2 -> r) -> m a1 -> m a2 -> m r I need to somehow get the mapped elements to be applied to each of the arguments before they can be combined. Substituting liftM2 for (foo (>>)) yields: No instance for (Monad ((->) (String, String))) arising from use of `liftM2' at DCFiles.hs:30:9-14 Probable fix: add an instance declaration for (Monad ((->) (String, String))) The foo function does the job, so I assume that I'm trying to solve a different problem than liftM. I wouldn't be surprised if I was wrong, however. -- Lou.

3 4

Translating Haskell to VHDL. What approach to follow?
by Alfonso Acosta 12 Dec '06

12 Dec '06

Hi all, As I wrote in the previous HLADPSA announcement, during the following moths I'm going to write a translator from Haskell to VHDL in order to accomplish my masters thesis goal. The main requirement is managing to translate from a ForSyDe (http://www.imit.kth.se/info/FOFU/ForSyDe/ ) specification which is Haskell after all, to VHDL. In order to get part of the work for free I decided to design the translator as a compiler backend. The best option so far seems to be Yhc's Core API (http://haskell.org/haskellwiki/Yhc/API/Core ), which unfortunately lacks type information. I discarded GHC due to the current "bit-rotted" state of External Core (http://www.haskell.org/pipermail/haskell-cafe/2006-October/018786.html ) and higher complexity of its core representation (SystemFC). I might have overlooked other approaches and/or intermediate representations such as GHC's C-- and STG (is it still used?). I'd be pleased to hear any suggestions. Thanks in advance, Alfonso Acosta

3 8

Pretty printing HTML
by Neil Mitchell 12 Dec '06

12 Dec '06

Hi, I want to pretty print some text as HTML, using the standard HughesPJ combinators. The problem is that I want the "length" of a text string to be its textual context, not its character length. For example, a keyword will display as "case", but internally will be the string "<span class='keyword'>case</span>" Is there any way to acheive this? I have come up with something, but its not nice at all. It basically involves parsing the generated text and inserting the extra HTML back in. Thanks Neil

1 0