- Haskell-Cafe - Haskell.org

Re: [Haskell] ANNOUNCE: vorbiscomment 0.0.1
by Antoine Latter 21 Jul '10

21 Jul '10

On Wed, Jul 21, 2010 at 10:16 AM, Chris Wagner <christopher.t.wagner(a)gmail.com> wrote: > I'm happy to announce that I have packaged up my first Haskell library, > vorbiscomment, which allows for reading of "Vorbis comments" from Ogg > Vorbis files. The code is rugged and surely buggy, but any feedback, > patches, etc. are more than welcome. > > I do have a specific question in releasing this, and you can respond to > my personal email address if this mailing list is not appropriate for > discussion, but I'm wondering where my library's functionality should > live within the module hierarchy. At present, I have simply placed it > in a top-level "Vorbis" module, but this is surely not the proper place > for it. > Hi Chris, I have 5 packages which populate under the Codec.* hierarchy, so I would try to stake something out underneath Codec.Audio.*. Are these comments unique to Vorbis, or are they for anything contained in Ogg? Antoine (also, moving discussion to -cafe)

1 0

Problem with reloading modules in GHC API
by Hongmin Fan 21 Jul '10

21 Jul '10

Hi, I'm using GHC API to dynamically load some module, and evaluate it; and later change the content of the module, and re-evaluate it. But I found unless I delete the object file created by previous compilation, the module seems not reloaded. I have set ghcLink = LinkInMemory as an older post suggested To illustrate what I'm saying, here is a piece of code (sorry for any naivety in the code, new to Haskell too) import System.IO (IOMode(..),hClose,hPutStr,openFile) import Directory (removeFile) import GHC import GHC.Paths import DynFlags import Unsafe.Coerce src_file = "Target.hs" obj_file = "Target.o" main = do writeTarget "arg" func0 <- compileTarget putStrLn $ show $ func0 2 writeTarget "arg*2" func1 <- compileTarget putStrLn $ show $ func1 2 writeTarget input = do -- removeFile obj_file `catch` (const $ return ()) -- uncomment this line to have correct results h <- openFile src_file WriteMode hPutStr h "module Target (Target.target) where\n" hPutStr h "target::Double -> Double\n" hPutStr h "target arg = \n " hPutStr h input hClose h compileTarget = defaultErrorHandler defaultDynFlags $ do func <- runGhc (Just libdir) $ do -- setup dynflags dflags <- getSessionDynFlags setSessionDynFlags dflags { ghcLink = LinkInMemory } -- load target module target <- guessTarget src_file Nothing setTargets [target] r <- load LoadAllTargets case r of Failed -> error "Compilation failed" Succeeded -> do m <- findModule (mkModuleName "Target") Nothing -- set context and compile setContext [] [m] value <- compileExpr ("Target.target") do let value' = (unsafeCoerce value) :: Double -> Double return value' return func The code basically write to a Haskell source file twice with different content, and hoping to get different results, but unless I uncomment the line with removeFile, the output of 2 runs are the same; using 'touch' to touch the source file being written between 2 runs also gives the correct results. So maybe caused by some caching mechanism? I'm using GHC 6.12.1 in Ubuntu 10.04. I have this workaround of deleting the obj file, but I'm wondering the "correct" way of doing it. Did some search on GHC API, but never got something relevant. Thanks, Hongmin

3 6

Re: [Haskell] ANNOUNCE: vorbiscomment 0.0.1
by Henk-Jan van Tuyl 21 Jul '10

21 Jul '10

Redirecting to Haskell-café On Wed, 21 Jul 2010 19:16:16 +0200, Chris Wagner <christopher.t.wagner(a)gmail.com> wrote: > I'm happy to announce that I have packaged up my first Haskell library, > vorbiscomment, which allows for reading of "Vorbis comments" from Ogg > Vorbis files. The code is rugged and surely buggy, but any feedback, > patches, etc. are more than welcome. > > I do have a specific question in releasing this, and you can respond to > my personal email address if this mailing list is not appropriate for > discussion, but I'm wondering where my library's functionality should > live within the module hierarchy. The "Hierarchical module names" page [1] on the HaskellWiki gives some guidelines; I suppose the name Sound.Vorbis.Comment would be a good one. Regards, Henk-Jan van Tuyl [1] http://www.haskell.org/haskellwiki/Hierarchical_module_names -- http://Van.Tuyl.eu/ http://members.chello.nl/hjgtuyl/tourdemonad.html --

1 0

Data types a la carte
by Victor Nazarov 21 Jul '10

21 Jul '10

I've been experimenting with type level programming and open recursion. My aim this far is to implement very modular interpreter for lambda calculi and combinatory logic powerful enough to demonstrate steps of some term's reduction to normal form. I'd like to be able to setup the set of combinators and reduction rules and to express the set in type signature. I'd like to achieve maximum modularity of the code by all possible means. Below is the beginning of this effort. Show instances are defined completely modular. Show instance for Lambda and show instance for Combinator reuse the very same code of Show instance of ApplicativeF. However, there are problems with this code, and I'd like to receive some suggestions to cope with them. 1) UndecidableInstances are required for this code to compile. Is it possible to avoid this language extension? What is the reason for it to be required? 2) app, var, abs functions. app function is defined differently for every example. Is it possible to perform some automatic lifting so that app will work for any combination of ApplicativeF? Can I use Functor class or should I define my own type-class for this purpose? Is it possible at all? 3) The code below uses technique similar to Data Type a la Carte. Is it possible to achieve the same modularity using Finally Tagless technique? > {-# LANGUAGE TypeOperators #-} > {-# LANGUAGE FlexibleContexts #-} > {-# LANGUAGE FlexibleInstances #-} > {-# LANGUAGE UndecidableInstances #-} Some type level combinators first > newtype (:.) f g a = O (unO :: f (g a)) > newtype (:<*>) f g a = S {unS :: f a (g a)} > newtype Flip f x y = Flip {unFlip :: f y x} > newtype Mu f = Mu {unMu :: f (Mu f)} Their show instances doesn't affect representation of underlying type > instance Show (f (Mu f)) => Show (Mu f) > where showsPrec d (Mu f) = showsPrec d f > instance Show (f (g a)) => Show ((f :. g) a) > where showsPrec d (O f) = showsPrec d f > instance Show (f a (g a)) => Show ((f :<*> g) a) > where showsPrec d (S f) = showsPrec d f > instance Show (f y x) => Show (Flip f x y) > where showsPrec d (Flip f) = showsPrec d f Applicative is some structure that permits application of some objects. We use open recursion here. ApplicativeF o is a functor that represent this structure. And Mu (Flip ApplicativeF o) is a fix-point of this functor. > data ApplicativeF a o = App a a | Obj o > type Applicative o = Mu (Flip ApplicativeF o) Pareses are left associative in applicative structure > instance (Show o, Show a) => Show (ApplicativeF o a) > where showsPrec d (Obj o) = showsPrec d o > showsPrec d (App a b) = > showParen (d > 4) $ showsPrec 4 a . (' ':) . showsPrec 5 b Here is an example of applicative structure of strings > test1 :: Applicative String > test1 = obj "x" `app` obj "y" `app` obj "z" `app` (obj "w" `app` obj "v") > where obj = Mu . Obj > app a b = Mu $ App a b Lambda calculi is a calculi of some objects with applicative structure. This object may represent a variable or an abstraction of variable. LambdaObjF encodes this fact. We use all combinators defined above to define objects of lambda calculi. > data LambdaObjF name a = Var Name | Abs Name a > type Lambda name = Mu (ApplicativeF :<*> LambdaObjF name) > instance (Show a) => Show (LambdaObjF String a) > where showsPrec d (Var s) = showParen (d > 5) (showString s) > showsPrec d (Abs s a) = > showParen (d > 3) $ > ('\\':) . showString s . showString ". " . showsPrec 3 a Lets try to build an object of lambda calculi > type Name = String > test2 :: Lambda Name > -- test2 = (\x. \y. x) a b > test2 = (abs "x" (abs "y" (var "x")) `app` var "a") `app` var "b" > where var = Mu . S . Flip . Obj . Var > abs x b = Mu . S . Flip . Obj $ Abs x b > app a b = Mu . S . Flip $ App a b Combinatory logic defines some basic combinators > data CombinatorObjF a = CK | CS > type Cominator = Mu (ApplicativeF :<*> CombinatorObjF) > instance Show (CombinatorObjF a) > where showsPrec d CK = showString "K" > showsPrec d CS = showString "S" > test3 = s `app` k `app` k `app` (k `app` s) > where s = Mu . S . Flip . Obj $ CS > k = Mu . S . Flip . Obj $ CK > app a b = Mu . S . Flip $ App a b -- Victor Nazarov

2 1

[Ann] New version of parser combinator library uu-parsinglib-2.4.0
by S. Doaitse Swierstra 21 Jul '10

21 Jul '10

I have uploaded a new version of the package uu-parsinglib, te collection of online, error-correcting parser combinators, with both an applicative and a monadic interface. * The new version behaves better when more than three corrective steps have to be taken at a specific position. In order ro do so we have added a form of abstract interpretation which computes the minimal length of each alternative, so the shortest one can be chosen (and not one of the recursive alternatives!). * Furthermore we compute for each parser whether it can recognise the empty string. This is used to provide error messages when such parsers are used in list-like constructs, on of the most commonly made mistakes by innocent parser writers. * a combinator <?> was added which can be used to label productions, and which label is used to report what was expected * some further overloadings of pSym have been added which function at a lower level, thus leading to much faster parsers, especially when using the combinator to construct scanner like parsers: e.g. spaces = pMunch (==' ') makes sure that all spaces are skipped in a single go, instead of dealing with each recognised ' ' by itself. I hope you like the extensions, and would be happy to hear from you, both if you use them successfully, or if you have any reason for not using them. Doaitse Swierstra

1 0

The where-clause and guards
by Eitan Goldshtrom 21 Jul '10

21 Jul '10

I'm trying to fit a where clause to some guards I'm using. I have the following f a b | c > 1 = ... | c < 1 = ... | otherwise = ... where c = a+b yet I'm getting a parsing error. Is this not the correct way to combine "where" with "guards"? -Eitan

2 5

List manager and duplicate copies of messsages
by José Romildo Malaquias 21 Jul '10

21 Jul '10

Hello. I have noticed that I do not receive duplicate copies of messages from haskell-cafe, although "Avoid duplicate copies of messages?" is set to "No" in the mailing list membership configuration. I want the copies because I archive all the mailing list messages with procmail based on the "X-BeenThere" header line. Has anybody else seen this behaviour, or am I missing anything? Is there any problems with the mailing list manager? Regards, Romildo -- Computer Science Department Universidade Federal de Ouro Preto, Brasil

2 3

Re: [Haskell-cafe] Test. Ignore.
by "José Romildo Malaquias" 21 Jul '10

21 Jul '10

> ----- Original Message ----- > From: José Romildo Malaquias <j.romildo(a)gmail.com> > To: José Romildo Malaquias <romildo(a)operamail.com> > Cc: haskell-cafe(a)haskell.org > Subject: Re: [Haskell-cafe] Test. Ignore. > Date: Wed, 21 Jul 2010 13:33:34 -0300 > > > On Wed, Jul 21, 2010 at 01:25:19PM -0300, "José Romildo Malaquias" wrote: > > > ----- Original Message ----- > > > From: "José Romildo Malaquias" <romildo(a)operamail.com> > > > To: "José Romildo Malaquias" <j.romildo(a)gmail.com> > > > Cc: haskell-cafe(a)haskell.org > > > Subject: Re: [Haskell-cafe] Test. Ignore. > > > Date: Wed, 21 Jul 2010 13:09:16 -0300 > > > > > > ----- Original Message ----- > > > > From: José Romildo Malaquias <j.romildo(a)gmail.com> > > > > To: José Romildo Malaquias <romildo(a)operamail.com> > > > > Cc: haskell-cafe(a)haskell.org > > > > Subject: Re: [Haskell-cafe] Test. Ignore. > > > > Date: Wed, 21 Jul 2010 13:01:03 -0300 > > > > > > > > > > > > On Wed, Jul 21, 2010 at 12:53:53PM -0300, "José Romildo Malaquias" wrote: > > > > > Please, ignore this message. > > > > > > > > Replying to the list and to the sender, to see if he gets both messages > > > > using a non gmail account. > > > > I confirm that I received only one copy of the above message, > > that > was sent directly to my email address at operamail.com. I > > did not > receive any copy from the mailing list manager. > > > > Romildo (at opera.com) > > > > That was expected, as by default the "Avoid duplicate copies of > > messages?" option is set to "Yes" in the mailing list membership > > configuration. Now I have just set it to "No", and I expect to receive > > both copies next time someone from the mailing list "replies to all" > > to one of my messages. > > > > Romildo at operamail.com > > Replying to all. > > Romildo at gmail.com Well, romildo at operamail.com received both copies of the message. So the list manager worked as expected. Romildo at operamail.com -- _______________________________________________ Surf the Web in a faster, safer and easier way: Download Opera 9 at http://www.opera.com

1 0

Re: [Haskell-cafe] Test. Ignore.
by "José Romildo Malaquias" 21 Jul '10

21 Jul '10

> ----- Original Message ----- > From: "José Romildo Malaquias" <romildo(a)operamail.com> > To: "José Romildo Malaquias" <j.romildo(a)gmail.com> > Cc: haskell-cafe(a)haskell.org > Subject: Re: [Haskell-cafe] Test. Ignore. > Date: Wed, 21 Jul 2010 13:09:16 -0300 > > > > ----- Original Message ----- > > From: José Romildo Malaquias <j.romildo(a)gmail.com> > > To: José Romildo Malaquias <romildo(a)operamail.com> > > Cc: haskell-cafe(a)haskell.org > > Subject: Re: [Haskell-cafe] Test. Ignore. > > Date: Wed, 21 Jul 2010 13:01:03 -0300 > > > > > > On Wed, Jul 21, 2010 at 12:53:53PM -0300, "José Romildo Malaquias" wrote: > > > Please, ignore this message. > > > > Replying to the list and to the sender, to see if he gets both messages > > using a non gmail account. > > I confirm that I received only one copy of the above message, that > was sent directly to my email address at operamail.com. I did not > receive any copy from the mailing list manager. > > Romildo (at opera.com) That was expected, as by default the "Avoid duplicate copies of messages?" option is set to "Yes" in the mailing list membership configuration. Now I have just set it to "No", and I expect to receive both copies next time someone from the mailing list "replies to all" to one of my messages. Romildo at operamail.com -- _______________________________________________ Surf the Web in a faster, safer and easier way: Download Opera 9 at http://www.opera.com

3 2

Re: [Haskell-cafe] Test. Ignore.
by "José Romildo Malaquias" 21 Jul '10

21 Jul '10

> ----- Original Message ----- > From: José Romildo Malaquias <j.romildo(a)gmail.com> > To: José Romildo Malaquias <romildo(a)operamail.com> > Cc: haskell-cafe(a)haskell.org > Subject: Re: [Haskell-cafe] Test. Ignore. > Date: Wed, 21 Jul 2010 13:01:03 -0300 > > > On Wed, Jul 21, 2010 at 12:53:53PM -0300, "José Romildo Malaquias" wrote: > > Please, ignore this message. > > Replying to the list and to the sender, to see if he gets both messages > using a non gmail account. I confirm that I received only one copy of the above message, that was sent directly to my email address at operamail.com. I did not receive any copy from the mailing list manager. Romildo (at opera.com) -- _______________________________________________ Surf the Web in a faster, safer and easier way: Download Opera 9 at http://www.opera.com

1 0