April 2010 - Haskell-Cafe

are forkIO threads event-driven?
by Aran Donohue 03 May '10

03 May '10

Hi Cafe, In GHC, if a thread spawned by forkIO blocks on some network or disk IO, is the threading system smart enough not to wake the thread until an IO event occurs on its input/output? The Control.Concurrent documentation doesn't specify, and the previous discussions I could find on this topic are out-of-date. There is a years-old GHC ticket, too, recently revived[2]. Put another way, is it possible yet to use forkIO for making a server to handle tens of thousands of concurrent network connections? If not, what is the best current Haskell/GHC way? Thanks, Aran [1] http://www.monkey.org/~provos/libevent/ [2] http://hackage.haskell.org/trac/ghc/ticket/635

8 14

Build problems (hsp, trhsx, ultimately Happstack)
by Alexander Solla 01 May '10

01 May '10

Consider the following bash session: [ ajs@kizaru:~/ ]$ which trhsx /home/ajs/.cabal/bin/trhsx [ ajs@kizaru:~/ ]$ trhsx Usage: trhsx <infile> [<outfile>] [ ajs@kizaru:~/ ]$ cabal install hsp Resolving dependencies... Configuring hsp-0.4.5... Preprocessing library hsp-0.4.5... Building hsp-0.4.5... ghc: could not execute: trhsx cabal: Error: some packages failed to install: hsp-0.4.5 failed during the building phase. The exception was: ExitFailure 1 Does anybody have any suggestions? I'm trying to install Happstack on Arch Linux (GHC 6.12.1). ~/.cabal/bin/ is in my path and trhsx runs. I've had this problem before, and I think I ran cabal as root for the HSP package. That's not working this time. I get the same failure.

7 19

Benchmarks game updated to ghc 6.12.2
by Don Stewart 01 May '10

01 May '10

The benchmarks game has been updated to use 6.12.2 Please dive in and help tweak/improve/spot any regressions. Esp. with respect to multicore flags/options/... ----- Forwarded message from Isaac Gouy ----- Subject: fyi benchmarks game updated to ghc 6.12.2 http://shootout.alioth.debian.org/u64q/haskell.php best wishes, Isaac ----- End forwarded message -----

6 9

Functional Dependencies Help
by John Creighton 01 May '10

01 May '10

I've been trying to apply some stuff I learned about functional dependencies, but I run into one of two problems. I either end up with inconsistent dependencies (OverlappingInstances doesn't seem to apply) or I end up with infinite recursion. I want to be able to do simple things like if a is a subset of b and b is a subset of c then a is a subset of c. If a is a is a subset of b and b is a c then a is a c. Before I added the equality functions I had infinite recursion. Once I put them them in then I have trouble with overlapping isntances. {-# LANGUAGE TypeOperators #-} {-# LANGUAGE OverlappingInstances #-} {-# LANGUAGE UndecidableInstances #-} {-# LANGUAGE MultiParamTypeClasses #-} {-# LANGUAGE FunctionalDependencies #-} {-# LANGUAGE FlexibleInstances #-} {-# LANGUAGE FlexibleContexts #-} data Noun = Noun deriving (Show) --10 data Verb = Verb deriving (Show) -- data Adjactive = Adjactive deriving (Show) data Animal=Animal deriving (Show) data Feline=Feline deriving (Show) data Cat = Cat deriving (Show) data Taby_Cat=Taby_Cat deriving (Show) data T=T deriving (Show) data F=F deriving (Show) --20 class Isa a b c | a b->c where isa::a->b->c instance Isa Animal Noun T where isa a b = T --25 class IsSuperSet a b c | a b->c where isSuperSet :: a->b->c instance IsSuperSet Feline Cat T where --30 isSuperSet a b=T instance IsSuperSet Animal Feline T where isSuperSet a b=T instance IsSuperSet a Animal F where isSuperSet a b=F --35 class TypeNotEq d b c | d b->c where typeNotEq :: a->b->c instance (IsSuperSet d b c, --40 IsSuperSet a d c, TypeNotEq a d T, TypeNotEq b d T, TypeEq c T T )=> IsSuperSet a b c where isSuperSet a b=undefined::c instance TypeNotEq a a c where typeNotEq a b = undefined::c --50 instance TypeNotEq a b c where typeNotEq a b = undefined::c class TypeEq a b c | a b->c where typeEq :: a->b->c instance TypeEq a a c where typeEq a b = undefined::c instance TypeEq a b c where typeEq a b = undefined::c class ToBool a where toBool :: a->Bool instance ToBool T where toBool a = True instance ToBool F where toBool a = False myCat=Cat bla=isSuperSet Animal Cat

1 3

Asynchronous exception wormholes kill modularity
by Bas van Dijk 01 May '10

01 May '10

Dear all, (sorry for this long mail) When programming in the IO monad you have to be careful about asynchronous exceptions. These nasty little worms can be thrown to you at any point in your IO computation. You have to be extra careful when doing, what must be, an atomic transaction like: do old <- takeMVar m new <- f old `onException` putMVar m old putMVar m new If an asynchronous exception is thrown to you right after you have taken your MVar the putMVar will not be executed anymore and will leave your MVar in the empty state. This can possibly lead to dead-lock. The standard solution for this is to use a function like modifyMVar_: modifyMVar_ :: MVar a -> (a -> IO a) -> IO () modifyMVar_ m io = block $ do a <- takeMVar m a' <- unblock (io a) `onException` putMVar m a putMVar m a' As you can see this will first block asynchronous exceptions before taking the MVar. It is usually better to be in the blocked state as short as possible to ensure that asynchronous exceptions can be handled as soon as possible. This is why modifyMVar_ unblocks the the inner (io a). However now comes the problem I would like to talk about. What if I want to use modifyMVar_ as part of a bigger atomic transaction. As in: block $ do ... modifyMVar_ m f ... >From a quick glanse at this code it looks like asynchronous exceptions can't be thrown to this transaction because we block them. However the unblock in modifyMVar_ opens an asynchronous exception "wormhole" right into our blocked computation. This destroys modularity. Besides modifyMVar_ the following functions suffer the same problem: * Control.Exception.finally/bracket/bracketOnError * Control.Concurrent.MVar.withMVar/modifyMVar_/modifyMVar * Foreign.Marshal.Pool.withPool We can solve it by introducing two handy functions 'blockedApply' and 'blockedApply2' and wrapping each of the operations in them: > import Control.Exception > import Control.Concurrent.MVar > import Foreign.Marshal.Pool > import GHC.IO ( catchAny ) > blockedApply :: IO a -> (IO a -> IO b) -> IO b > blockedApply a f = do > b <- blocked > if b > then f a > else block $ f $ unblock a > blockedApply2 :: (c -> IO a) -> ((c -> IO a) -> IO b) -> IO b > blockedApply2 g f = do > b <- blocked > if b > then f g > else block $ f $ unblock . g Control.Exception: > finally :: IO a -> IO b -> IO a > a `finally` sequel = blockedApply a $ \a' -> do > r <- a' `onException` sequel > _ <- sequel > return r > bracket :: IO a-> (a -> IO b) -> (a -> IO c) -> IO c > bracket before after thing = blockedApply2 thing $ \thing' -> do > a <- before > r <- thing' a `onException` after a > _ <- after a > return r > bracketOnError :: IO a -> (a -> IO b) -> (a -> IO c) -> IO c > bracketOnError before after thing = blockedApply2 thing $ \thing' -> do > a <- before > thing' a `onException` after a Control.Concurrent.MVar: > withMVar :: MVar a -> (a -> IO b) -> IO b > withMVar m io = blockedApply2 io $ \io' -> do > a <- takeMVar m > b <- io' a `onException` putMVar m a > putMVar m a > return b > modifyMVar_ :: MVar a -> (a -> IO a) -> IO () > modifyMVar_ m io = blockedApply2 io $ \io' -> do > a <- takeMVar m > a' <- io' a `onException` putMVar m a > putMVar m a' > modifyMVar :: MVar a -> (a -> IO (a,b)) -> IO b > modifyMVar m io = blockedApply2 io $ \io' -> do > a <- takeMVar m > (a',b) <- io' a `onException` putMVar m a > putMVar m a' > return b Foreign.Marshal.Pool: > withPool :: (Pool -> IO b) -> IO b > withPool act = blockedApply2 act $ \act' -> do > pool <- newPool > val <- catchAny > (act' pool) > (\e -> do freePool pool; throw e) > freePool pool > return val I'm not proposing to make this change (yet) because I first would like to have some discussion on this. Thanks for reading this rather long mail, Bas

13 46

Fwd: [Haskell-cafe] Re: Simple game: a monad for each player
by Limestraël 01 May '10

01 May '10

Okay, I just understood that 'Prompt' was just a sort of view for 'Program'. >I'd like to make it very accessible, so please don't hesitate to report >any difficulties with finding and understanding documentation and examples! Then I think the name 'Prompt' may be misleading for those who doesn't know the MonadPrompt package. Maybe something like 'ProgramView' ? >> What have we achieved? Both players still can only access functions from >> whatever monad m turns out to be. But now each strategy can pile its own >> custom monad stack on the Player m monad! And of course, the use of >> the m Monad is completely optional. > >Of course, the custom monad stack has to provide a projection back to >the Player m a type > > runMyStackT :: MyStackT (Player m) a -> Player m a According to what Bertram said, "each strategy can pile its own custom monad stack ON the (Player m) monad". Here, you are stacking the (Player m) monad ON the custom monad stack. What is then the use of the 'm', in (Player *m*)? Is it not supposed to be a custom monad? (MonadIO for human, Identity for AI, etc.) But then, I don't see how the game function could work: > game :: Monad m => Player m () -> Player m () -> m () As it is written, it requires both players to run in the SAME monad. And if have a network player ( e.g.* Player (StateT Handle IO)* ) and an AI storing former opponent's moves ( e.g. *(Monad m) => Player (StateT [Move] m)* ), then they can't be in the same monad... * * 2010/4/14 Heinrich Apfelmus <apfelmus(a)quantentunnel.de> > Limestraël wrote: > > > I have some difficulties to see the use of PromptT, because in the > tutorial, > > this type is never mentioned, and its operations (Return and :>>=) are > > instead constructors of ProgramT... > > > > Would you have some concrete examples? Because there I'm a bit lost > (since > > the tutorial doesn't match the operational package as it is, because of > the > > type PromptT)... > > The project page > > http://projects.haskell.org/operational/ > > links to documentation that describes the differences to "The > Operational Monad Tutorial", in particular the new Prompt and PromptT > types. It also links to several examples. Two small examples are also > included in the Haddock documentation. > > > I'd like to make it very accessible, so please don't hesitate to report > any difficulties with finding and understanding documentation and examples! > > > Regards, > Heinrich Apfelmus > > -- > http://apfelmus.nfshost.com > > _______________________________________________ > Haskell-Cafe mailing list > Haskell-Cafe(a)haskell.org > http://www.haskell.org/mailman/listinfo/haskell-cafe >

3 9

How to make way into a hadoop infrastructure
by C K Kashyap 30 Apr '10

30 Apr '10

Dear Haskellers, A big part of my new job requires tuning app's on Hadoop. I was wondering if there is a way to push some Haskell code in the mix. I did some googling on "Hadoop/Haskell" and came across Holumbus - but looks like that is parallel to Hadoop. I was thinking in the lines of doing a Haskell implementation that could run in a Hadoop cluster - has anyone tried anything like that? -- Regards, Kashyap

2 2

why does Data.Text.Lazy.IO.readFile return the internal type Data.Text.Lazy.Internal.Text, when Data.Text.IO.readFile returns plain IO Data.Text.Text?
by Thomas Hartman 30 Apr '10

30 Apr '10

*Main> :t Data.Text.IO.readFile Data.Text.IO.readFile :: FilePath -> IO T.Text but *Main> :t Data.Text.Lazy.IO.readFile Data.Text.Lazy.IO.readFile :: FilePath -> IO text-0.7.1.0:Data.Text.Lazy.Internal.Text why does the lazy version use the internal type, whereas the strict version of Text IO just using plain Data.Text type? and how can I get from internal type to regular type when using Data.Text? also the internal type doesn't appear to be reflected in the haddock: http://hackage.haskell.org/packages/archive/text/0.7.1.0/doc/html/Data-Text… ghc-pkg list | grep -i text text-0.7.1.0 thanks for any help!

5 9

64-bit code on OSX?
by Gregory Crosswhite 30 Apr '10

30 Apr '10

Hey everyone, Just out of curiosity, does anyone know the story on what is holding up the ability of GHC to generate 64-bit code on OSX? Cheers, Greg

1 0

ANN: Combinatorrent v0.3.0
by Jesper Louis Andersen 30 Apr '10

30 Apr '10

We are pleased to announce Combinatorrent v0.3.0 Combinatorrent is a bittorrent client, written in Haskell and with an emphasis on concurrency. This version has the following changes from v0.2.0: - Numerous space leak fixes. The client doesn't leak memory anymore for typical runs. - A number of performance optimizations. In particular the number of CPU seconds per downloaded megabyte went down considerably compared to earlier versions. - Lowered allocation rate of the client so the garbage collector has less work to do. - Support BEP 0027 implicitly. - Add support for compiling with GHC HEAD (6.13.yyyymmdd series). - Change the parser in the incoming direction to attoparsec and use its incremental features to simplify the receiver loop. Outgoing direction still uses cereal. - Claim BEP 0006 support by implementing the Fast-extension. - Add (untested) BEP 0007 support: Handle IPv6 addresses as well as IPv4 addresses. - Fix a bug where interest updates were not considered correctly. Also lower the amount of interest communication to peers to save bandwidth. Looking ahead, the goal is to support at least the extended messaging protocol. The EMP is a dependency prerequisite for many of the other extensions, hence it is an important building block for further work. As always, the source code and issue tracker is at github: http://github.com/jlouis/combinatorrent -- J.

1 0