April 2007 - Haskell-Cafe

A monad using IO, Reader, Writer, State and Error
by Martin Huschenbett 13 Apr '07

13 Apr '07

Hi all, for my current project I need a monad that is an instance of MonadIO, MonadReader, MonadWriter, MonadState, and MonadError. I see two ways for defining such a monad using the mtl. 1) type MyMonad = ErrorT E (RWST R W S IO) and 2) type MyMonad = RWST R W S (ErrorT E IO) I can't figure out what is the difference between these two definitions and therefore which one is more suitable for my problem. Or are the equivalent and it is unimportant which one I use? Or is it even better define a new type like newtype MyMonad a = MyMonad { runMyMonad :: R -> S -> IO (Either E a,S,W) } and declare instances for all 5 type classes? Thanks for your help in advance, Martin.

2 2

implementing try for RWST ?
by Jeremy Shaw 13 Apr '07

13 Apr '07

Hello, I defined a newtype like this (the ()s will be replace with something more useful in the future): > newtype DryRunIO a = DryRunIO { runDryRunIO :: RWST Bool () () IO a } > deriving (Monad, MonadIO, MonadError IOError, MonadFix, Functor, MonadReader Bool, MonadWriter (), MonadState ()) > > run :: Bool -> DryRunIO a -> IO a > run dryRun action = (runRWST (runDryRunIO action)) dryRun () >>= \ (a, _, _) -> return a and I want to define a function similar to |try| like this: > tryDR :: DryRunIO a -> DryRunIO (Either IOError a) > tryDR m = catchError (m >>= return . Right) (return . Left) unfortunately, when I use it, it does not work the way I want: *Main> run False (tryDR (error "cheese")) *** Exception: cheese This is because 'error' raises an |Exception|, not a |IOError|. I would like to instead define tryDR to deal with exceptions: > newtype DryRunIO a = DryRunIO { runDryRunIO :: RWST Bool () () IO a } > deriving (Monad, MonadIO, MonadError Exception, MonadFix, Functor, MonadReader Bool, MonadWriter (), MonadState ()) > > tryDR :: DryRunIO a -> DryRunIO (Either Exception a) > tryDR m = catchError (m >>= return . Right) (return . Left) But to do this, I need an different instance of MonadError for IO, namely: > instance MonadError Exception IO where > throwError = throwIO > catchError = catch -- (from Control.Exception) However, if I add that to my module I get this error: Functional dependencies conflict between instance declarations: instance MonadError Exception IO -- Defined at /tmp/DR.hs:17:0 instance MonadError IOError IO -- Defined in Control.Monad.Error Where do I go from here? Also, what is the justificiation for the current default of IOError instead of the more general Exception? thanks! j. ps. As a hack I have implemented tryDR as: > tryDR' :: DryRunIO a -> DryRunIO (Either Exception a) > tryDR' (DryRunIO m) = DryRunIO $ RWST $ \r s -> > catch (runRWST m r s >>= \(a, s, w) -> return (Right a, s, w)) -- uses catch from Control.Exception > (\e -> runRWST (return $ Left e) r s) However, I think this is buggy, because changes to 's' and 'w' will be lost if 'm' raises an exception. For example in: > tryDR' (io $ putStr "hello" >> updatePosition (length "hello") >> error "goodbye") The updatedPosition would reflect the position before the tryDR, but the cursor on the screen would be somewhere else. It is my hope that the earlier definition does not have this bug.

1 0

Parsing indentation-based languages with Parsec
by George Pollard 13 Apr '07

13 Apr '07

Hi, first time list poster :) I've searched around a bit but haven't been able to find any examples of this. I want to be able to parse a language (such as Haskell, Python) which has only EOL as the 'statement' separator and has indentation levels to indicate block structure. Whilst doing this I want to use Parsec's nice library. The first thing I noticed was that Parsec's whiteSpace parser will ignore EOL as just whiteSpace, so I need to redefine that. Is this the correct way to do it? I've only been using Haskell for a week or so so I'm not too sure on the record structures and updating them... lexer :: P.TokenParser () lexer = ( P.makeTokenParser emptyDef { commentLine = "#", nestedComments = True, identStart = letter, identLetter = letter, opStart = oneOf "+*/-=", opLetter = oneOf "+*/-=", reservedNames = [], reservedOpNames = [], caseSensitive = False } ) { --update lexer fields P.whiteSpace = do --just gobble spaces many (char ' ') return () } (I got the basic code from the tutorial contained within the Parsec docs.) For handling the indented blocks I thought I would use something to hold current indentation state, as Parsec has support for threading state through all the parsers. Is this the right way to go about this? Has anyone done the 'groundwork' with parsing such languages so I don't need to reinvent this? Thanks in advance, - porges.

1 0

Re: [Haskell-cafe] Compiling GHC
by Chris Witte 12 Apr '07

12 Apr '07

On 4/11/07, Andrew Appleyard <aja(a)cse.unsw.edu.au> wrote: > The source paths for these copies are derived from the path of the > gcc binary ($GccDir). Did you call configure with '--with-gcc=C:/ > Mingw/bin/gcc.exe'? I received those errors when I forgot to do that. Yep that fixed that. Thanks. > I think that 'gettimeofday' is statically linked into the ghc > executable, so it's just that GHCi doesn't know about it. Adding a: > > Sym(gettimeofday) \ > > line to the RTS_MINGW_ONLY_SYMBOLS #define in rts/Linker.c (from line > 265) seemed to fix the problem for me. > > I'm not sure if this is a GHCi bug or something else being awry. > 'gettimeofday' was added to the mingwex library in mingw- > runtime-3.10, which was released in July last year, so I would have > expected this problem to have come up before now... I made this change but I still get the error unknown symbol `_gettimeofday' I'm using mingw 5.1.3 which as far as i can tell should use the 3.11 runtime.

2 3

xmonad window manager mailing list
by dons＠cse.unsw.edu.au 12 Apr '07

12 Apr '07

I'm pleased to announce that the xmonad window manager now has a mailing list set up: http://www.haskell.org/mailman/listinfo/xmonad For those who don't know, xmonad is a tiling window manager written in Haskell. You can find more about it at: http://xmonad.org and Lennart Kolmodin's recent blog post http://lennartkolmodin.blogspot.com/2007/04/xmonad.html Additionally, the lead xmonad developer, Spencer Janssen, has been accepted by PSU to work on XCB bindings to Haskell, for the Google Summer of Code. XCB is a simpler, smaller, threadsafe replacement for Xlib, which should simplify further the window manager code base. We hope to have the 0.1 release of xmonad out shortly, until then, hope to see you on the mailing list! -- Don

1 0

more problems with c2hs
by Ruben Zilibowitz 12 Apr '07

12 Apr '07

Hi, I've built and installed c2hs on my Mac OS X system now. I read through the research paper on c2hs available on the website for c2hs and decided I would need to try it out to learn it. But it gives me errors which I am having trouble trying to understand. Here are the error messages: c2hs file.chs powerpc-apple-darwin8-gcc-4.0.1: c: No such file or directory powerpc-apple-darwin8-gcc-4.0.1: c: No such file or directory powerpc-apple-darwin8-gcc-4.0.1: warning: '-x -x' after last input file has no effect powerpc-apple-darwin8-gcc-4.0.1: no input files c2hs: Error during preprocessing custom header file Can anyone figure out what is going on here? Is it my chs file or is it to do with c2hs? Regards, Ruben

2 1

Re: Why Perl is more learnable than Haskell
by Ryan Dickie 12 Apr '07

12 Apr '07

I thought I could resist this thread but I'll bite =:-() The first language i learned was basic. No real functions, simple step by step instructions. I then learned hypercard, c, c++, python, assembly, vhdl, and too many others! Now i've decided to learn haskell. I view it as a mathematicians language. I do research in the field of medical imaging.. particularly processing large cardiac data sets to figure out characteristics, diseases, etc. Why workflow generally starts out as 1) mathematical idea 2) turn that pure equation into a numerical recipe 3) implement, debug 5) analyze 6) goto step 1. I find doing my thinking in the continous domain makes things a lot easier. But here's where i differ from everyone else. I already have the mathematical relationships all nice and tidy (hopefully!) in my head before i start. I literally just implement it. I don't want to care about threading, IO, message passing, or numerical stability. I have to care about performance but only so far as it hampers my productivity. Preferably the language will do it implicitly. Your average programmer wants a language to do tasks. Having to think about the math and relationships behind it all is rather sickening to them (and me too!). I am a new haskell programmer (basically a week into it!). It is by far the hardest language i've had to pick up. A lot of my code could be structured in a functional way.. but almost all reply on looping techniques (would take a lot of work to rethink my gradient descent method and make it fast!). Regardless of actually using haskell.. i like to transfer these techniques to c++. Before i even knew of haskell i knew some FP methods and i found that using these shrunk my code, shrunk the bugs, and did nice things for performance + concurrency. In fact, I just read a google paper on their batch system. They use to functions: map and reduce. They can easily split it up over their cluster etc... these are the ideas of FP that i like! That and the set-builder notation. I also hate matlab to death. Is there any possibility of using haskell as a replacement using ghci? Mostly i care about linalg when it comes to using matlab. ps: sorry if gmail butchered this reply. I had subscribed to the digest and turns out that was a mistake :D

3 2

c2hs on mac
by Ruben Zilibowitz 12 Apr '07

12 Apr '07

Hi, I am having trouble building c2hs on Mac OS X. I noticed that Darwinports has a port for this, but I'd like to install it without using Darwinports because I already have ghc compiled and installed on my system. Does anyone know how to get c2hs to work on the mac (or why it doesn't work in the first place)? Regards, Ruben

2 6

hackage.haskell.org
by David Waern 12 Apr '07

12 Apr '07

Hi, I'd like to set up a Trac for Haddock on hackage.haskell.org. Who should I contact? /David

3 5

Re: [Haskell-cafe] k-minima in Haskell
by Nicolas Frisby 12 Apr '07

12 Apr '07

[sorry for the double, ajb] Since there seemed to be a disconnect between the expectation and the previous answers, I thought an alternative suggestion might help out. This sort of thing (haha) usually isn't my cup o' tea, so please point out any blunders. RM, is this more like what you had in mind? It leans more towards an iterative approach. If so, I encourage you to compare this to the other solutions and try understand how those solutions rely upon laziness. Stefan and Andrew, feel free to point out the drawbacks to this approach that I'm probably overlooking. kminima n l = map fst (foldr insert' (List.sort pre) suf) where (pre, suf) = (splitAt n . zip [0..]) l -- I think this insertion sort could be -- O(log k) with a better data structure. insert' x [] = [] insert' x (y : ys) | snd x < snd y = x : y : dropLast ys' | otherwise = y : insert' x ys where dropLast ys = take (length ys - 1) ys We grab the first k elements and sort them, this list is our first approximation to the k-minima. We then process the rest of the list, checking if the current element is smaller than any of the minima of the current approximation. If the current element is smaller, we improve the current approximation by inserting the new element and dropping the biggest (i.e. last) minimum from the minima accumulator. The worst behavior is: sort(k) + (n-k) * k comparisons. This could be improved (to: sort(k) + (n-k) * log k comparisons, I think) with a better data structure for the minima approximation. On 4/12/07, ajb(a)spamcop.net <ajb(a)spamcop.net> wrote: > G'day all. > > Quoting raghu vardhan <mrvr84(a)yahoo.co.in>: > > > So, any algorithm that sorts is no good (think of n as huge, and k small). > > The essence of all the answers that you've received is that it doesn't > matter if k is small, sorting is still the most elegant answer in Haskell. > > The reason is that in Haskell, a function which sort function will produce the > sorted > list lazily. If you don't demand the while list, you don't perform > the whole sort. > > Some algorithms are better than others for minimising the amount of > work if not all of the list is demanded, but ideally, producing the > first k elements will take O(n log k + k) time. > > Cheers, > Andrew Bromage > _______________________________________________ > Haskell-Cafe mailing list > Haskell-Cafe(a)haskell.org > http://www.haskell.org/mailman/listinfo/haskell-cafe >

1 0