August 2009 - Haskell-Cafe

Re: [Haskell-cafe] Functional Dependencies and multiparametertypeclasses
by Harald ROTTER 24 Aug '09

24 Aug '09

Thanks for the quick response. Your explanation is so much clearer than ghc's error messages :-)) Harald. |---------+------------------------------> | | Miguel Mitrofanov | | | <miguelimo38@yandex| | | .ru> | | | | | | | | | 24.08.2009 13:29 | | | Delivered date: | | | 24.08.2009 13:31 | | | | |---------+------------------------------> >------------------------------------------------------------------------------------------------------------| | | | To: Harald ROTTER <harald.rotter(a)sagem.com> | | cc: haskell-cafe(a)haskell.org | | Subject: Re: [Haskell-cafe] Functional Dependencies and multiparameter typeclasses | >------------------------------------------------------------------------------------------------------------| > However, I do not really understand the cause of the original problem. Why > do I need the functional dependency to make this work ? Suppose you have an instance: instance Pixel MyPixel where... instance Image MyImage MyPixel where width i = countPixels i ... And somebody (not necessarily you) makes another instance: instance Pixel CustomPixel where... instance Image MyImage CustomPixel where width i = 0 ... Now, take any image im :: MyImage Then you'd expect that width im :: Int But now there are TWO implementations of "width :: MyImage -> Int" and no way for compiler to figure out which one to use. GHC sort of foresees this problem. By adding a functional dependency you instruct it to forbid the second declaration, when the first one is present, thus removing the problem. " Ce courriel et les documents qui y sont attaches peuvent contenir des informations confidentielles. Si vous n'etes pas le destinataire escompte, merci d'en informer l'expediteur immediatement et de detruire ce courriel ainsi que tous les documents attaches de votre systeme informatique. Toute divulgation, distribution ou copie du present courriel et des documents attaches sans autorisation prealable de son emetteur est interdite." " This e-mail and any attached documents may contain confidential or proprietary information. If you are not the intended recipient, please advise the sender immediately and delete this e-mail and all attached documents from your computer system. Any unauthorised disclosure, distribution or copying hereof is prohibited."

1 0

ANNOUNCE: OAuth library in haskell
by Diego Souza 24 Aug '09

24 Aug '09

Dear Haskellers, hoauth is a library which helps you to deal with oauth protocol. Currently it supports only consumer side applications, but there are plans to add service providers support in near future. The source code can be found at [darcs]: http://projects.bitforest.org/hoauth/ and now in hackage: http://hackage.haskell.org/package/hoauth If you have any questions, comments or criticism, please get in touch with me. I'll appreciate it very much, specially because there are so many things yet to learn. Thanks, PS: This is the first piece of code I produce in Haskell since I've start learning it few months ago. I must say it has been a joy since then. -- ~dsouza yahoo!im: paravinicius gpg key fingerprint: 71B8 CE21 3A6E F894 5B1B 9ECE F88E 067F E891 651E

1 0

Where does documentation get installed with cabal?
by Colin Paul Adams 23 Aug '09

23 Aug '09

Hello, I'm trying to find the API documentation for happstack 0.3 (online is for 0.2). So I did: cabal install happstack --reinstall --enable-documentation but I can't find it anywhere within ~/.cabal - where should I look? -- Colin Adams Preston Lancashire

5 6

hslogger and syslog-ng
by Magnus Therning 23 Aug '09

23 Aug '09

I'm having some problem with logging to syslog using System.Log.Logger and friends. I have the following: > :m +System.Log.Logger > :m +System.Log.Handler.Syslog > sl <- openlog "foo" [PID] USER DEBUG > updateGlobalLogger rootLoggerName (addHandler sl) > warningM "bar" "test" *** Exception: sendTo: protocol error (Protocol wrong type for socket) What could be the cause for this? I've tried hsyslog and it reports to my instance of syslog without problems. My system is a 32-bit Arch Linux, running syslog-ng 3.0.3. /M -- Magnus Therning (OpenPGP: 0xAB4DFBA4) magnus＠therning．org Jabber: magnus＠therning．org http://therning.org/magnus identi.ca|twitter: magthe

2 6

Monad woes
by Levi Greenspan 23 Aug '09

23 Aug '09

Hi all, I try to create a simple monad using a stack of Reader and IO but when using it, I encounter some problems. The Monad is defined as M a: {-# LANGUAGE GeneralizedNewtypeDeriving #-} module Main where import Control.Monad.Reader import Control.Concurrent newtype M a = M { unM :: ReaderT String IO a } deriving (Monad, MonadIO, MonadReader String) runM :: String -> M a -> IO a runM s m = runReaderT (unM m) s loop :: (String -> M ()) -> M () loop f = forever $ f "hello" I then define a callback function to be invoked by 'loop': callback :: String -> M () callback s = liftIO $ print s >> threadDelay 1000000 So far so good. Then I test it like this: test1 :: IO () test1 = runM "foo" $ do loop callback liftIO $ print "here" -- OK. Never reached Still works fine. 'loop' never returns. In a real life application 'loop' is an event loop and I'd like to fork it into a new thread like this: test3 :: IO () test3 = runM "foo" $ liftIO $ do forkIO $ do return $ loop callback return () print "here" threadDelay 2000000 This not only looks ugly, it also doesn't work. For 'loop callback' to pass the type checker, it had to be returned into the IO monad. But I guess due to laziness, 'loop' will never be called. This can be confirmed without forkIO: test2 :: IO () test2 = runM "foo" $ liftIO $ do return $ loop callback print "here" Again, 'loop callback' will not be invoked. Now, given that I must find a way to combine IO and my M monad I don't know what to try next. Prima facie it seems I must somehow force 'loop callback' to be evaluated, but how? Not to mention all the liftIO clutter. I would greatly appreciate some help here. Thank you very much! Cheers, Levi

3 5

STM problems
by Michael Speer 23 Aug '09

23 Aug '09

I am playing a writing a simple STM hashtable. I expected the following code to alter the TVars in it, but they appear immutable. I assume I am simply doing something wrong, but I am not sure what it is. Any pointers would be greatly appreciated. import Control.Concurrent.STM import Control.Concurrent import Array data HashTable a b = HashTable (a -> Int) (STM (Array Int (TVar (Maybe b)))) mkHashTable :: ( a -> Int ) -> Int -> HashTable a b mkHashTable hashfn numBuckets | numBuckets > 0 = do HashTable hashfn $ do vs <- mapM (\v -> newTVar Nothing >>= \tv -> return (v,tv) ) [1..numBuckets] return $ array (1,numBuckets) vs readHashTable :: HashTable a b -> a -> STM (Maybe b) readHashTable hashtable key = let (HashTable hashfn table) = hashtable in table >>= \tb -> readTVar (tb ! hashfn key) writeHashTable :: HashTable a b -> a -> (Maybe b) -> STM () writeHashTable hashtable key value = let (HashTable hashfn table) = hashtable in table >>= \tb -> writeTVar (tb ! hashfn key) value >> return () main = let ht = mkHashTable id 10 :: HashTable Int String in do v <- atomically $ do mapM_ (\n -> writeHashTable ht n (Just $ "hello " ++ show n)) [1..10] mapM (\n -> readHashTable ht n) [1..10] print v

2 2

ANN: haskell-src-exts-1.1.3
by Niklas Broberg 23 Aug '09

23 Aug '09

Fellow Haskelleers, I'm pleased to announce the release of haskell-src-exts-1.1.3. * On hackage: http://hackage.haskell.org/package/haskell-src-exts * Via darcs: darcs get http://code.haskell.org/haskell-src-exts * Report bugs: http://trac.haskell.org/haskell-src-exts haskell-src-exts is a package for Haskell source code manipulation. In particular it defines an abstract syntax tree representation, and a parser and pretty-printer to convert between this representation and String. It handles (almost) all syntactic extensions to the Haskell 98 standard implemented by GHC, and the parsing can be parametrised on what extensions to recognise. haskell-src-exts-1.1.3 is a highly experimental release, which (apart from a small bug fix, see below) doesn't do anything at all to the current stable part of haskell-src-exts. But the release is far from as boring as all that may sound, as it includes a whole new set of modules implementing a new and more accurate syntax tree where all nodes are adorned with annotations. Together with this comes a parser that retains exact source information, stored in the aforementioned annotations. There is also support for various precision for source information, including SrcSpans. The new modules mirror the ones in the stable part of haskell-src-exts but instead live in Language.Haskell.Exts.Annotated{.Syntax,.Parser,...} (except that SrcLoc and friends now have their own module). The point of this release is that those of you that for one reason or other are waiting for these features can help me look, feel, test and comment these things in isolation, and hopefully report a lot of bugs, before I integrate this stuff with the main haskell-src-exts body of code. Note that that coming integration is intended to be backwards-compatible, but a lot of the behind-the-scenes will be switched over to the new and more general machinery. Feature-wise, with this accurate source tree in hand, the next step is to write a printer that uses the full location information to allow full and exact round-tripping, i.e. (print . parse == id). == haskell-src-exts-1.1.3 == * Adds support for more accurate, and fully annotated, abstract syntax, see above. * Fixes a bug that forced 'rec' statement groups to end with a qualifier (like 'do'). Cheers and Happy Haskelling, /Niklas

3 3

Help starting a Haskell blog
by Peter Verswyvelen 23 Aug '09

23 Aug '09

I'm going start my very first blog, documenting my everyday struggle to switch my old imperative mind to the lazy functional setting, with a focus on FRP. Although you can find a lot of articles that provide help to get started with general blogging, it might be useful to pick a blog in which presenting Haskell code is easy (e.g. like hpaste that does the syntax coloring for you), and where users can give feedback, providing code, also with syntax coloring preferably. It would also be nice to allow hyperlinking every function in the code to the standard Haskell library docs or to the docs on Hackage. Googling for "how to start a Haskell blog" just revealed a lot of Haskell blogs. Could you share your experiences with me about starting a blog? BTW: I'm on Windows. Thanks a lot, Peter Verswyvelen

6 8

ANN: ByteString Nums
by Jason Dusek 23 Aug '09

23 Aug '09

This is a simple package for relatively careless parsing of numbers from ByteStrings. It works to parse out integer strings, floating point strings and hex strings. http://hackage.haskell.org/package/bytestring-nums For integer strings, an initial sign is detected and non-digit characters are ignored: Prelude> :load Data.ByteString.Nums.Careless.Int *Data.ByteString.Nums.Careless.Int> let strings = ["121.3040", "1o2", "-a1", "0xff", "-0x30"] *Data.ByteString.Nums.Careless.Int> fmap (int . pack) strings :: [Int] [1213040,12,-1,0,-30] *Data.ByteString.Nums.Careless.Int> fmap (int . pack) strings :: [Double] [1213040.0,12.0,-1.0,0.0,-30.0] For hexadecimal strings, no signs are recognized and characters that aren't hex digits are ignored. One consequence of this is that an initial `0x` is skipped. Prelude> :load Data.ByteString.Nums.Careless.Hex *Data.ByteString.Nums.Careless.Hex> let strings = ["121.3040", "1o2", "-a1", "0xff", "-0x30"] *Data.ByteString.Nums.Careless.Hex> fmap (hex . pack) strings :: [Int] [18952256,18,161,255,48] The float parser recognizes the last group of digits behind a comma or period as the fractional part; all preceding digits are treated as the integral part. Naturally, signs are handled. Prelude> :load Data.ByteString.Nums.Careless.Float *Data.ByteString.Nums.Careless.Float> let strings = ["121.3040", "2.100,32", "1,o2", "-a1", "0xff", "-0x30"] *Data.ByteString.Nums.Careless.Float> fmap (float . pack) strings :: [Double] [122.304,2100.32,1.2,-1.0,0.0,-30.0] *Data.ByteString.Nums.Careless.Float> fmap (float . pack) strings :: [Rational] [15163 % 125,52508 % 25,6 % 5,(-1) % 1,0 % 1,(-30) % 1] I have not applied any `SPECIALIZE` or `INLINE` directives at this time; it'd be great to have some insight in to optimizing collections of small functions like this library. -- Jason Dusek

1 0

Strange console IO behavior (at least on Windows)
by Peter Verswyvelen 23 Aug '09

23 Aug '09

The following program http://hpaste.org/fastcgi/hpaste.fcgi/view?id=8445#a8445 should echo the ASCII code of each character the user types, on the fly, with no line buffering whatsoever. But it doesn't. At least not under Windows's cmd, and even less with msys. Under Windows, I have to press enter before anything is printed to the screen. It then always prints 71 (that's upper G, no idea where that comes from), and then blocks. Pressing enter again prints the codes of all characters typed, except the first one, following by the two codes of the linefeeds. So e.g. when I type "abc" ENTER, I get 71, then hitting ENTER again, I get 98 99 10 10. Under msys, it just ignores both the hSetBuffering and hSetEcho, but when hitting enter, it at least prints the codes correctly. How does it behave under OSX and Linux? Is this a bug or am I doing something wrong?

4 5