April 2011 - Haskell-Cafe

Unboxing CFloat and GLfloat
by Jason Dagit 03 Apr '11

03 Apr '11

First a bit of background. Sometimes I would like to be able to create unboxed arrays of CFloats and GLfloats. Both of these are defined as newtypes around types that can be unboxed (e.g., Float). I'm assuming GHC here. As far as I can tell, the newtyping is there to help programmers pick a representation that matches the C compiler's representation on their platform without having to use CPP in their Haskell code. Would it be possible to add instances like these? instance MArray (STUArray s) GLfloat (ST s) instance MArray IOUArray GLfloat IO instance MArray (STUArray s) CFloat (ST s) instance MArray IOUArray CFloat IO I would also want instances for the other C types that map to Haskell types that are unboxed and the other GL types too. For example, in my program I imported all the right libraries and added this instance as an experiment: instance MArray (STUArray s) GLfloat (ST s) where {-# INLINE getBounds #-} getBounds (STUArray l u _ _) = return (l,u) {-# INLINE getNumElements #-} getNumElements (STUArray _ _ n _) = return n {-# INLINE unsafeNewArray_ #-} unsafeNewArray_ (l,u) = unsafeNewArraySTUArray_ (l,u) fLOAT_SCALE {-# INLINE newArray_ #-} newArray_ arrBounds = newArray arrBounds 0 {-# INLINE unsafeRead #-} unsafeRead (STUArray _ _ _ marr#) (I# i#) = ST $ \s1# -> case readFloatArray# marr# i# s1# of { (# s2#, e# #) -> (# s2#, unsafeCoerce (F# e#) #) } {-# INLINE unsafeWrite #-} unsafeWrite (STUArray _ _ _ marr#) (I# i#) e = ST $ \s1# -> let F# e# = unsafeCoerce e in case writeFloatArray# marr# i# e# s1# of { s2# -> (# s2#, () #) } I had to use an unsafeCoerce because the constructor for GLfloat is not exposed, and realToFrac has terrible performance because the following rules (suggested by Andy Gill) are missing from the OpenGL bindings: {-# RULES "realToFrac/a->GLfloat" realToFrac = \x -> GLfloat (realToFrac x) #-} {-# RULES "realToFrac/GLfloat->a" realToFrac = \(GLfloat x) -> realToFrac x #-} Even if I had those rules, I would still need to get access to the underlying representation (the e#) to have the right kind. GLfloat is defined in terms of CFloat which is defined in terms of other things. On my machine, CFloat happens to correspond to Float, although in general this need not be the case. So to do this instance correctly I would need to know, in general, what the underlying type is for CFloat. I think that means I would need to use the same macros as GHC, although I don't think that requires this instance to appear in base. I could just borrow the macros once I figure out the corresponding instance for CFloat. While the GLfloat instances don't need to be in base, the CFloat instance probably should be in base. Is this something I should make a library proposal for and submit patches? Is there an easier way to get these instances? Has someone already done this? Thanks, Jason

2 1

Call for contributions: Second edition of the Typeclassopedia
by Brent Yorgey 03 Apr '11

03 Apr '11

Hi all, It's time for a second edition of the Typeclassopedia [1], and I'd love to have your suggestions and contributions. You can send me suggestions and darcs patches [2] by email. For more detailed information, see my blog post [3]. -Brent [1] http://www.haskell.org/wikiupload/8/85/TMR-Issue13.pdf [2] http://code.haskell.org/~byorgey/typeclassopedia/ [3] http://byorgey.wordpress.com/2011/04/03/call-for-contributions-second-editi…

1 0

error to Fail Conversion
by Anwar Bari 03 Apr '11

03 Apr '11

HI Cafe I have problem with this function value :: Theorm-> [(String,Int)] -> Tem -> Int value _ env (Var s) = maybe (error "Unknown variable.") id (lookup s env) value mdl env (OP s _ _ l) = maybe (error "Function not defined.") id (lookup (map (value mdl env) l) (S.toList $ maybe (error "Uknown function.") id (fmap snd (multiLookup s (operations mdl))))) This function is working fine, but I want to change error function to fail function to be able use it in my GUI application. I can't change the output of value function to IO Int,any idea will help. Thanks Nour.

2 1

Can't connect to local yackage server
by Yitzchak Gale 03 Apr '11

03 Apr '11

On Mac OS X 10.6.7, after upgrading to yackage 0.1.0.1: After running yackage -l, I cannot connect to it even directly from localhost. I get the message: This Yackage server only talks to local clients On a Linux server it seems to work fine though. On the Mac, when I do "telnet localhost" I get: Trying ::1... Could it be that yackage 0.1.0.1 doesn't recognize the IPv6 notion of localhost? Thanks, Yitz

2 2

7th Ghent Functional Programming Group meeting on Tuesday, April 26
by Jasper Van der Jeugt 03 Apr '11

03 Apr '11

Dear all, We are very glad to announce an exciting program for the 7th meeting of the Ghent Functional Programming Group, especially since we are celebrating our first year of existence today. Our program features no less than three interesting functional programming languages: Erlang, Haskell and Scheme. The meeting will take place on Tuesday, April 26 in the Technicum building of Ghent University (Sint-Pietersnieuwstraat 41, 9000 Gent) at 19h30. As before, to enter the building, you should go to the automatic sliding door on the far left of the building and dial the phone number that is provided on the note taped to the door. Someone will then open the door for you. Our program is as follows: 1. Tom Van Custem - Experiments with MapReduce in Erlang MapReduce is a programming model for large data processing popularized by, and in daily use at Google. The MapReduce model builds strongly on key tenets of functional programming such as higher-order functions and side-effect free execution. In this talk, we summarize this programming model and describe a didactic implementation in Erlang. Invented at Ericsson's research labs, Erlang is known for its massively concurrent programming model, and itself builds on a functional core language. The talk will not focus on Erlang as such, but we will describe its key features as needed to understand the MapReduce abstraction. 2. Tom Schrijvers - How you could have won the VPW 2011 contest with Haskell We all know that Functional Programming is great for writing concise solutions for programming problems. With some skill this can even be done quickly! Yet, there was little evidence of this at the 3rd edition of the Flemish Programming Contest (VPW 2011) that took place on March 23. Not so before the contest: The jury stress-tested all questions by writing various solutions in different languages. Haskell was used to solve most problems and invariably produced short solutions. In this talk I present my own Haskell solutions to several of this year's problems and discuss alternative solution strategies with the audience. After the talk you will be all set for winning next year's edition -- or at least enjoying it -- using Haskell. 3. Pieter Audenaert - Functional Geometry and a Graphical Language We will discuss a simple language for drawing images. During the presentation we will illustrate the power of data abstraction and algebraic closure, meanwhile using higher order procedures in an essential manner. The language has been designed to easy experimenting with patterns such as those appearing in typical M.C. Escher drawings where the artist repeats the pattern both moving it across the drawing and scaling it when applicable. In the language we use procedures to represent the data objects that will be combined in the final drawing and we make sure that all operations conducted on these procedures are algebraically closed. These features allow generating patterns of any complexity. For our implementation, we use the LISP functional programming language -- more accurately, the Scheme dialect. The presentation is based on "Structure and Interpretation of Computer Programs", Abelson & Sussman Hope to see you there! The GhentFPG organizing committee, Andy Georges Jeroen Janssen Jasper Van der Jeugt

2 2

using IO monad in Iteratee
by Dmitry Olshansky 03 Apr '11

03 Apr '11

Dear collegues, I use Iteratee with Monad Transformer stack in this way: type MyMonad a b = Iteratee a (StateT StateType IO) b I've wrote some Enumeratees using this type. Then I want to compose it with standard enumerator like Data.Enumerator.Binary.enumFile. But enumFile use IO monad instead of MonadIO class. I didn't see (maybe I'm blind...) a way to compose my Enumeratees with enumFile. How I can do that? Is it possible to change enumFile to using MonadIO class? Best regards, Dmitry

5 6

pool, persistent, persistent-sqlite: Space leak
by Ertugrul Soeylemez 03 Apr '11

03 Apr '11

Hello Michael, hello fellow Haskellers, there seems to be a space leak in either 'pool', 'persistent' or 'persistent-sqlite'. From the behaviour I suspect the bug to be in 'pool'. When I run a transaction in an infinite loop, my program keeps eating more and more memory, even if the transaction itself doesn't do anything: forever $ runSqlPool (return ()) pool This doesn't happen for non-pooled connections, i.e. the following code runs in constant space as expected: forever $ runSqlConn (return ()) conn Versions are pool-0.0.1, persistent-0.4.2, persistent-sqlite-0.4.0. It happens for both file and in-memory databases. Greets, Ertugrul -- nightmare = unsafePerformIO (getWrongWife >>= sex) http://ertes.de/

2 2

darcs repo browser
by Radoslav Dorcik 02 Apr '11

02 Apr '11

Hi, there is one haskell web application based on happstack for viewing foreign darcs repos exposed only on pure HTTP servers. I need that for checking status of some repositories when I don't have access to darcs binary/shell... maybe it can be useful for somebody else too. Service is running on following URL: http://handra.rampa.sk/dawb/changes See more comments on http://rampa.sk/static/dawb.html Regarding RSS feeds - Please send me a note If you find the hourly cron job I run on some repos to produce feeds too expensive for your server resources. It does not obey robots.txt file yet. Regards, Rado

3 2

Re: [Haskell-cafe] Asynchronous Arrows need Type Specialization - Help!
by Paul L 02 Apr '11

02 Apr '11

Sorry, forgot to CC the list. I wonder why Gmail doesn't default to reply-all. On Fri, Apr 1, 2011 at 9:48 PM, David Barbour <dmbarbour(a)gmail.com> wrote: > > If we ignore the 'delay' primitive (which lifts latency into program > logic), my model does meet all the arrow laws. Nonetheless, the issues > of physical synchronization still apply. It's important to recognize > that the Arrow Laws do not describe non-functional properties, such as > time-space performance. Well, when you throw the time stamp into the value domain, and according to your previous calculation of time stamps, (a1 *** a2) >>> first a3 will produce different result than (a1 >>> a3) *** a2. This is in direct conflict to arrow laws. Arrows by themselves do not impose physical synchronization. They are very often used to model computations about synchronous data streams, but that is a very different concept. >> In the actual implementation of such lifting (perhaps over multiple >> type classes), the calculation of the time stamp can be made precise. > > Indeed. And it is precisely the greater of the input timestamps. ;-) This is because you are using a single number as timestamp. Define your timestamp type differently, you'll get a different opinion. For example, you can use a pair of time stamps for pairs, and nested time stamps for nested values. -- Regards, Paul Liu

2 1

Setting up cabal on EC2
by Lyndon Maydwell 02 Apr '11

02 Apr '11

Hi all. I'm having some issues setting up cabal on EC2. I've installed ghc 7.0.2, however, the bootstrap.sh script for cabal-install is complaining about missing dependencies: > Linking Setup ... > Configuring Cabal-1.8.0.2... > Setup: At least the following dependencies are missing: > base >=4 && <3 && >=1 && <5, filepath >=1 && <1.2 > Setup: At least the following dependencies are missing: base >=4 && <3 && >=1 && <5, filepath >=1 && <1.2 ghc-pkg tells me that I have base and filepath installed: > /usr/local/lib/ghc-7.0.2/package.conf.d > base-4.3.1.0 > /usr/local/lib/ghc-7.0.2/package.conf.d > filepath-1.2.0.0 Could there be an environment variable I need to set somewhere?

2 4