- Haskell-Cafe - Haskell.org

Haskell & LLVM
by Ryan Cleven 19 Aug '06

19 Aug '06

Has anyone ever investigated using LLVM as a target for a haskell compiler? Perhaps a backend for GHC? Thanks, Ryan.

1 0

Haskell genetic algorithm
by Maurício 18 Aug '06

18 Aug '06

Hi, Does anyone knows of nice genetic algorithms implementations in Haskell? Thanks, Maurício

1 0

how to generate XHTML 1.1?
by Tamas K Papp 18 Aug '06

18 Aug '06

Hi, I would like to use Haskell to generate static HTML webpages, preferably using XHTML 1.1. How can I do this? I found libraries to generate XHTML 1.0 transitional, but not 1.1. Scheme has LAML [1], which parses the DTD, uses S-expressions and validates the XHTML, I am looking for something similarly elegant (even if without the S-expressions, which I haven't seen in Haskell yet). Thanks, Tamas [1] http://www.cs.aau.dk/~normark/laml/

3 2

More threading confusion
by Creighton Hogg 18 Aug '06

18 Aug '06

Good afternoon Haskellers, So I'm trying to understand how STM works, and wrote a quick 'eating philosophers' example to see if I understood how it's supposed to work. The problem is that while it executes, it doesn't appear to *do* anything. Did I completely write things wrongheadedly or am I being bitten by something more subtle? Thanks. import Control.Concurrent.STM import Control.Concurrent import Data.Array import System.Random think :: IO () think = do ms <- randomRIO (20,1000) threadDelay ms data Philosopher = Philosopher {left::Bool,right::Bool,neighbors::(Int,Int)} deriving Show makeInitPhilosopher a = Philosopher {left=False,right=False,neighbors=a} initPhilosophers = listArray (0,4) (map makeInitPhilosopher [(1,4),(2,0),(3,1),(4,2),(0,3)]) main = do z <- atomically $ newTVar initPhilosophers mapM_ (\x -> forkIO (loop x z 0 10000)) [0,1,2,3,4] loop n tps c l | c > l = (atomically (readTVar tps)) >>= (\x -> print x) | otherwise = do think atomically $ eat n tps loop n tps (c+1) l eat :: Int -> TVar (Array Int Philosopher) -> STM () eat n tps = do takeLeft n tps takeRight n tps releaseLeft n tps releaseRight n tps takeLeft :: Int -> TVar (Array Int Philosopher) -> STM () takeLeft n tps = do ps <- readTVar tps let p = ps ! n if right (ps ! (fst $ neighbors p)) == False then (writeTVar tps $ ps // [(n,p{left=True})]) else retry takeRight :: Int -> TVar (Array Int Philosopher) -> STM () takeRight n tps = do ps <- readTVar tps let p = ps ! n if left (ps ! (snd $ neighbors p)) == False then (writeTVar tps $ ps // [(n,p{right=True})]) else retry releaseLeft n tps = do ps <- readTVar tps let p = ps ! n writeTVar tps $ ps // [(n,p{left=False})] releaseRight n tps = do ps <- readTVar tps let p = ps ! n writeTVar tps $ ps // [(n,p{right=False})]

3 2

Samba/FTP bindings
by Ivan Tarasov 17 Aug '06

17 Aug '06

Is there some Haskell library which provides Samba bindings and some FTP client library bindings (e.g. ftplib3)? I started writing the bindings for some of the functions myself for my project but it looks like it is a lot of work and probably someone has done it already. -- Best regards, Ivan Tarasov

2 1

Last statement in 'do' must be an expression error.
by Szymon Ząbkiewicz 17 Aug '06

17 Aug '06

Hi. When trying to compilke this code: {...} 8.if (a == 0) && (b == 0) 9. then do 10. nr1 <- read (prompt "enter 1. number: ") 11. nr2 <- read (prompt "enter 2. number: ") 12. else do 13. let nr1 = a 14. nr2 = b {...} The compiler tells me thats there's an error on line 10: "The last statement in a 'do' construct mesy be an expression" Could you tell me how to change it so that the "declaration" of the first nr1 and nr2 is still in the "then" block.

6 5

wxFruit with OpenGL and IO action
by shelarcy 17 Aug '06

17 Aug '06

Hello. Sometimes, a few people are interested in wxFruit. http://zoo.cs.yale.edu/classes/cs490/03-04b/bartholomew.robinson/ But nobady knows current status of wxFruit. Over one and half year ago, I heard that students working on that ... . http://www.haskell.org/pipermail/haskell-cafe/2005-January/008765.html http://www.haskell.org/pipermail/haskell-cafe/2006-March/014998.html So, I'm working on that for subproject of Kamiariduki ... but progress is very slowly. I uploaded current (initial) version on Kamiariduki's file space. File name is "YaPan-initial.tgz". Because this file includes YaPan - Yet another implementation, and reimplementation of Panic (http://www.cse.unsw.edu.au/~sseefried/pan/index.html) using Yampa (and wxFruit). http://sourceforge.net/project/showfiles.php?group_id=168626 I didn't think current verion is good enough to announse [Haskell] list, so I didn't announce that and repraced "YaPan.tgz" file sometimes ... but I know that a few people downloaded my file already. So, I decide to upload initial file on that space. Of cource, "YaPan-initial.tgz" has other wxFuit example. paddle.hs is almost same as original example, except that use menu istead of button. And CFK_GL* is a OpenGL sample code. But current version has some faults. Cabal file support to build and install only YaPan, doesn't support wxFruit. Code is very messy, and YaPan doesn't support all of Panic's features. And I should notice additonal things for older version's user. Now YaPan depends on AutoForms (not for wxFruit now). http://autoforms.sourceforge.net/ And YaPan can't save static one picture, YaPan can save animation by consecutive number files only. Current version, I tested on only Windows with my build of wxWidgets 2.4.2's wxHaskell (I uplaoded Kamiariduki's file space) and afrp-0.4 environment. So I don't know wxFruit and YaPan can works well other environments. Finally, I show current problem of wxFruit with IO action. I extend that button menu allow to regist extra IO action like this. pan pixels = hBox $ proc _ -> do let start = False rec v <- boxSF $ iPre 0 -< 1 + v ev <- getEvent -< () vTrans <- boxSF $ iPre (0,0,1) -< (keyboardEvent 0.1 vTrans) ev panedPixels <- boxSF $ pictureEvent -< (pixels, vTrans, (circle v), v, gameBox) _ <- glpicture (glpsizeWithAttributes gameBox [WXCore.GL_RGBA, WXCore.GL_DOUBLEBUFFER]) -< glppic $ paintGL gameBox panedPixels _ <- menu (mtextWithDialog "&SaveImage") -< mcommandWithDialog $ (\v f -> saveImageHandler (animToDisplayFun (circle)) circle (WX.sizeW gameBox, WX.sizeH gameBox) vTrans 1 v f) v returnA -< () This design is suitable for button and menu's command. But if I want to use IO action out of these control event, modualiry is broken. Because I want to do that action out of control's event, but I have to change control's action. And I must choose using what kind of event handler. (Is interval the best?) So I tried to implement command function that intend to use IO action on reative loop wihtout control in wxFruit.hs ... but action doesn't do registered action or program is not work on my apporoach. (I commented out current command function that occur action.) If anyone have good idea how to use IO action on wxFuit, please tell me that idea. Thanks. -- shelarcy <shelarcy capella.freemail.ne.jp> http://page.freett.com/shelarcy/

1 0

Why does Haskell have the if-then-else syntax?
by Mike Gunter 17 Aug '06

17 Aug '06

I had hoped the "History of Haskell" paper would answer a question I've pondered for some time: why does Haskell have the if-then-else syntax? The paper doesn't address this. What's the story? thanks, -m

18 21

Type hackery help needed!
by Niklas Broberg 17 Aug '06

17 Aug '06

Hi fellow Haskelleers, I have a tricky problem that I just can't seem to solve. The problem is one of unresolved overloading, much like the (show . read) issue, but unlike that particular problem I feel there should be a solution to the one I'm wrestling with. I've tried to strip away all the unnecessary bits so the following will be a bit abstract, but it shows the gist of the problem. I have two overloaded functions - one I call 'build' that creates values from subparts, and one I call 'embed' that turns values into a form appropriate to call 'build' on. What they really represent is creating values of some XML tree structure, with embed making sure that the children are of the right type. Conceptually, we could imagine that these two functions to be defined as > class Build x c | x -> c, c -> x where > build :: String -> [c] -> x > > class Embed a b where > embed :: a -> b They would be used together as in e.g. > p c = build "p" [embed c] This works pretty well, the fundep x -> c tells me what b should be, assuming I infer a proper type for the result of the composition. The type of p is then p :: (Embed a c, Build x c) => a -> x where the c in the middle is determined by x via the fundep. My problems arise because I want to start nesting these to form tree structures, as in > tree = p (p "foo") Expanding the definition of p, the argument to the outer call to build is now embed $ build "p" [embed "foo"] :: (Embed String c, Build x c, Embed x x1) => x1 Through the fundep on the outer build I can infer what x1 should be, but there's no way I can infer x (without inserting an explicit type signature which is out of the question). This problem is probably unsolvable in general, but for my particular problem there are some restrictions that makes me feel there should be a clever way of working with classes and fundeps to make this all work out. I just can't seem to find one. These are the rules that the setup must obey: * A value of any type should be embeddable inside a build expression of any result type, i.e. a -> b or b -> a cannot hold on Embed in general. * The exception to the above is that if an expression using 'build' is embedded inside an outer 'build', as in 'tree' above, the inner build should have the same result type as the outer build (in a sense forming b -> a only for types instantiating Build). In other words, children of an xml element must be of the same type as their parent, even before the call to embed. * However, it would be nice, but probably not crucial, if by using explicit type signatures where I want to, I could actually embed values of "other" xml types than the outer one I am embedding them in, and letting embed convert them to xml of the correct type. I suspect this cannot be. The types of build and embed are negotiable, and certainly the type classes, as long as the two can be used together in roughly the same way as indicated above. The calls to build and embed will be autogenerated by a preprocessor from a HSP-style XML syntax, e.g. <p><% c %></p> <==> build "p" [embed c] and for this reason any solution *must* be syntactically expressible as expressions on a form similar to the above, but those expressions can be arbitrarily convoluted. For instance in my failed attempts so far I have used let expressions around build to form self-recursive versions to "pass a type" down through the evaluation, as in p c = let x = build "p" [embed x c] in x This is an ok solution syntactically, if only it solved the problem, I still can't see how to propagate it to the next level of build. :-( Is there anyone out there with the proper type class fu who can see a way out for me? Is this even possible to do at all? If yes, tell me please, and if not, I would be most interested in seeing why it cannot work. Any and all comments are welcome. Thanks for reading this long :-) /Niklas

2 5

Int->[Char] conversion
by Tamas K Papp 17 Aug '06

17 Aug '06

Hi, I am working through Hal Daume's tutorial, trying to do the exercises. I can't figure out how to output an integer with putStrLn (or any other way), I think I need an Int -> [Char] conversion but couldn't find it. Specifically, in Exercise 3.10, I have the product of numbers in pp, and would like to do putStrLn ("Product: " ++ convertnumbertostring(pp)) but I don't know which function does this... Sorry for the extremely dumb question. Thanks, Tamas

5 4