September 2009 - Haskell-Cafe

Is it safe to use unsafePerformIO here?
by Cristiano Paris 20 Sep '09

20 Sep '09

Hi Cafè, I've the following problem: I have a (possibly very long) list of files on disk. Each file contains some metadata at the beginning and continues with a (possibly very large) chunk of data. Now, the program I'm writing can be run in two modes: either read a specific file from the disk and show the whole chunk of data on screen, or read all the files' metadata, sort the file list based on the metadata, and display a summary of those without reading the chunk of data from each file. I've factored out the file access machinery in a single module so as to use it indifferently under the two scenarios. At first, I wrote a piece of code which, in spirit, works like the following reduced case: ------ module Main where import System.IO import Control.Applicative import Data.List import Data.Ord import Debug.Trace data Bit = Bit { index :: Integer, body :: String } readBit fn = withFile fn ReadMode $ \h -> Bit <$> (hGetLine h >>= return . read) <*> readBody where readBody = withFile fn ReadMode $ \h -> do b <- hGetContents h seq b $ trace ("Read body from: " ++ fn) $ return b main = do bl <- mapM readBit ["file1.txt","file2.txt"] mapM_ (putStrLn . show . index) $ sortBy (comparing index) bl ---- which is very expressive as it's written in applicative style. Each file is like the following: ---- file1.txt ---- 1 foo ---- I've created a separate IO action for reading the body in the hope that it wouldn't get executed when the file list is sorted. But, to my surprise, it didn't work as the trace message gets written for each file before the summary is displayed. Thinking about this, I came to the conclusion that the IO Monad is enforcing proper IO ordering so that the IO action for file1's body must be executed right before IO action for file2's index one. If this is true, the only solution that came into my mind was to wrap the IO action for reading the body in an unsafePerformIO call. I actually ran the program with this modification and it works properly. So, as using unsafePerformIO is a Great Evil (TM), I'm wondering if there's a different way to do this which doesn't rely on retyping body as an IO action returning a String, which would break my pure code manipulating the files. My opinion is that using unsafePerformIO here is like ensuring the compiler that there're no observable side effects in running the IO action for reading the body and that no other side effects would impact this IO action. Thank you for any thoughts. Cristiano

7 31

TFM09: Call for Participation (FMWeek, Eindhoven, November 2009)
by J.N. Oliveira 20 Sep '09

20 Sep '09

TFM2009 2nd Int. FME Conference on Teaching Formal Methods Friday, November 6th 2009, co-located with FM2009 : 16th Int. Symposium on Formal Methods Eindhoven, the Netherlands, November 2 - November 6, 2009 CALL FOR PARTICIPATION (URL: http://www.di.uminho.pt/tfm09) 1. About the conference ----------------------- Ten years after the First World Formal Methods Congress (FM'99) in Toulouse, formal methods communities from all over the world will once again have an opportunity to come together. As part of the First Formal Methods Week event surrounding the FM2009 conference in Eindhoven, Formal Methods Europe will be organizing TFM2009, the Second International Conference on Teaching Formal Methods. The conference will serve as a forum to explore the successes and failures of Formal Methods (FM) education, and to promote cooperative projects to further education and training in FMs. TFM2009 will include a panel discussion on the idea of building a 'Guide to the Formal Methods Body of Knowledge' (FMBoK), inspired by similar efforts for software engineering (SWEBoK) and for project management (PMBoK); such a resource would provide guidance to teachers, managers, and developers on what should be expected from a comprehensive, balanced programme of education in FMs. 2. Invited speaker ------------------ Jeff Kramer (Imperial College London, UK) 3. Accepted papers ------------------ * Teaching Concurrency: Theory in Practice (Luca Aceto, Anna Ingolfsdottir, Kim Guldstrand Larsen, Jiri Srba) * Integrated and Tool-Supported Teaching of Testing, Debugging, and Verification (Wolfgang Ahrendt, Richard Bubel, Reiner Haehnle) * What Top-Level Software Engineers Tackle after Learning Formal Methods - Experiences from the Top SE Project (Fuyuki Ishikawa, Kenji Taguchi, Nobukazu Yoshioka, Shinichi Honiden) * Teaching program specification and verification using JML and ESC/Java2 (Erik Poll) * Chief Chefs of Z to Alloy: Using A Kitchen Example to Teach Alloy with Z (Sureyya Tarkan, Vibha Sazawal) * Teaching Formal Methods based on Rewriting Logic and Maude (Peter Olveczky) * Which Mathematics for the Information Society? (Joao Ferreira, Alexandra Mendes, Roland Backhouse, Luis Barbosa) * How to explain mistakes (Stefan Hallerstede, Michael Leuschel) * On Teaching Formal Methods: Behavior Models and Code Analysis (Jan Kofron, Ondrej Sery, Pavel Parizek) * Teaching Formal Methods for the Unconquered Territory (Nestor Catano, Camilo Rueda) 4. Sponsors -------------- * Formal Methods Europe Association (FME) * Software Improvement Group (SIG) , Amsterdam, Netherlands 5. Programme Committee ---------------------- Izzat Alsmadi (North Dakota State University, USA) Dines Bjorner (IIMM Institute, Denmark) Eerke Boiten (University of Kent, UK) Raymond Boute (Universiteit Gent, Belgium) Andrew Butterfield (Trinity College, Dublin) Jim Davies (University of Oxford, UK) David Duce (Oxford Brookes University, UK) John Fitzgerald (University of Newcastle upon Tyne, UK) Jeremy Gibbons (University of Oxford, UK) Randolph Johnson (National Security Agency, USA) Michael Mac an Airchinnigh (Trinity College, Dublin) Dino Mandrioli (Politecnico di Milano, Italy) Jose Oliveira (Universidade do Minho, Portugal) Kees Pronk (Technische Universiteit Delft, NL) Bernhard Schaetz (Tecnical University of Munique, Germany) Wolfgang Schreiner (Johannes Kepler University Linz, Austria) Simao Melo de Sousa (Universidade da Beira Interior, Portugal) Kenji Taguchi (National Institute of Informatics, Japan) Jeannette Wing (Carnegie-Mellon University, USA)

1 0

Ambiguous type variable with subclass instance (also: is there a better way to do this?)
by Andy Gimblett 19 Sep '09

19 Sep '09

Hi all. This email is in literate Haskell; you should be able to load it into ghci and verify what I'm saying (nb: it won't compile without alteration: see below). I'm trying to do something which may anyway be stupid / not the best approach to what I'm trying to achieve; however, it's not working and I can't see why not. So I'm asking for help on two fronts: 1) Why is this failing? 2) Maybe more usefully, how should I actually be doing this? It seems an ugly approach; a voice in my head is saying "scrap your boilerplate", but I've no idea yet if that's actually applicable here; should I look at it? On with the show... I need these for "subclass" stuff later on... > {-# LANGUAGE FlexibleInstances #-} > {-# LANGUAGE OverlappingInstances #-} > {-# LANGUAGE UndecidableInstances #-} > module Ambig where I wish to define a number of algebraic data types with the ability to turn Int values into instances of those types. So I define a typeclass saying this is possible. I use Maybe so I can encode the existence of out-of-range Int values, which will vary from target type to target type. > class Target a where > convert :: Int -> Maybe a E.g. here's a type Foo which only wants values between 1 and 10: > data Foo = Foo Int deriving (Show) > instance Target Foo where > convert n | n `elem` [1..10] = Just $ Foo n > | otherwise = Nothing (That's a simple example; some are rather more complex. How to do this isn't what I'm asking about, really.) So we have, for example: *Ambig> (convert 1) :: Maybe Foo Just (Foo 1) *Ambig> (convert 11) :: Maybe Foo Nothing Now, some of those algebraic data type types happen to be enumerations; in this case, my idea is to list the constructors, with the rule that each constructor's position in the list is the Int which gets converted into that constructor. > class Enumerated a where > constructors :: [a] E.g. here's a type Bar with three constructors: > data Bar = X | Y | Z deriving (Show) > instance Enumerated Bar where > constructors = [X, Y, Z] (This is certainly ugly. Any suggestions?) Now we get to the crux. If a type is an instance of Enumerated, it should also be a Target, because we should be able to convert from Int just by list lookup. But we include a bounds check, naturally... > instance (Enumerated a) => Target a where > convert n | n `elem` [0..len-1] = Just $ constructors !! n > | otherwise = Nothing > where len = length constructors So I would _hope_ that then, e.g., we'd have: *Ambig> (convert 0) :: Maybe Bar Just X *Ambig> (convert 1) :: Maybe Bar Just Y *Ambig> (convert 3) :: Maybe Bar Nothing Sadly, this function doesn't compile, dying with an "Ambiguous type variable" error: Ambig.lhs:75:29: Ambiguous type variable `a' in the constraint: `Enumerated a' arising from a use of `constructors' at Ambig.lhs:74:29-40 Probable fix: add a type signature that fixes these type variable(s) If we replace "length constructors" with "3" (say), it compiles (but is useless). Adding a type signature doesn't help: it's "misplaced" in that context. If I break it out of the instance declaration so I can add one, I still get the same problem: > convert' :: (Enumerated a, Target a) => Int -> Maybe a > convert' n | n `elem` [0..len-1] = Just $ constructors !! n > | otherwise = Nothing > where len = length constructors I guess I see roughly what's going on; the question is "which constructors instance is meant?", right? In the "Just" part it's OK, because it can be inferred from the function's return type (right?). But in the guard we don't have that help, so it could be any Enumerated instance? Any advice appreciated! Particularly if this is just a dumb approach. For context, this is related to deserialisation of binary data (they'll actually be Word8's, not Int's) into a variety of data structures. Hmmm, maybe I should just be using Data.Binary... Many thanks, -Andy

5 12

Why the stack overflow?
by staafmeister 19 Sep '09

19 Sep '09

Hi haskell-cafe, Why does rlist 100000 [] gives stack overflow in ghci? rlist 0 l = return l rlist n l = do {x <- randomRIO (1,maxBound::Int); let nl = x:l in nl `seq` rlist (n-1) nl} I first uses replicateM then foldM and finally an explicit function. But they give all stack overflow I don't know why 100000 is not absurd and it is tail recursive. Or is it not, due to the monad structure? greetings Gerben -- View this message in context: http://www.nabble.com/Why-the-stack-overflow--tp25520431p25520431.html Sent from the Haskell - Haskell-Cafe mailing list archive at Nabble.com.

3 2

Win32 API
by Andrew Coppin 19 Sep '09

19 Sep '09

Hi guys. Last time I looked at the Win32 bindings, it covered a few basic file I/O things (e.g., special access modes, file permissions, etc.), registry editing, and that was about it. Yesterday I took another look, and was pleasently surprised to find that GDI is now covered. Suffice it to say, yesterday I wrote my very first ever program using the Win32 API directly. (Who'd have thought Haskell would be the place to do that?) This was complicated by a small glitch: Graphics.Win32.Window exposes SendMessage() but does not expose PostMessage(). Kind of an important difference there. Fortunately, it's not actually especially hard to fix this deficiency. (Basically few the source code for the module, copy and paste the line for SendMessage(), and edit it to say PostMessage(). The type signature just happens to be identical.) Is there a reason why this is missing to start with? What other functions are missing? (I didn't see PostQuitMessage() anywhere...) Still, 85 lines of code to make a working, native-looking Windows program isn't too shabby, really... (Not that any sane person writes nontrivial programs directly like this of course.)

2 1

Compiling Yi's Dependencies on Windows
by Jeff Wheeler 19 Sep '09

19 Sep '09

Hey all, Yi has a lot of dependencies that currently make it a pain to install on Windows, so I'm wondering which of those can be streamlined for easier install on Windows. So far, I haven't succeeded at getting them to work on my own machine, and I've already spent a lot of time on it. The first major impediment is Gtk2Hs. Duncan has mostly resolved this issue with a minimal build that works on 6.10.4 [1], but I think that this should at least be mentioned on the Gtk2Hs site if it is impossible for now to get cabal-install to work by itself. Then comes bindings for things like curl (this may only be needed for Darcs, I don't remember), zlib, and regex (POSIX). If I remember correctly, zlib, or some other library, is able to ship with its *.c/*.h files and compile them as necessary on Windows. Can this be done for regex-posix and curl, too? On my system now, I've got both GnuWin32 and MSYS (MinGW) installed in attempts to get regex headers that work with regex-posix, but, at the moment, I've run into a wall, in which I get this error when loading or linking to regex-posix (it compiles fine itself): ghc.exe: C:\...\cabal\regex-posix-0.94.1\ghc-6.10.4\HSregex-posix-0.94.1.o: unknown symbol `_regerror' Loading package regex-posix-0.94.1 ... linking ... ghc.exe: unable to load package `regex-posix-0.94.1' I doubt most of these changes are significant (like linking on the gtk2hs site to the 6.10.4 build), but changing them could make many apps easier to install on Windows, not just Yi. Are there any other dependencies that have trouble on Windows which I haven't run into yet? If they're library dependencies, can the headers be included in the Hackage tarball? Jeff Wheeler

1 0

big array allocation too slow?
by Paul L 19 Sep '09

19 Sep '09

I'm trying to use newArray to allocate something that has 100M unboxed doubles. It takes quite a few seconds to do so on GHC 6.10.2. But doing the same thing (and initialize all to 0) in C returns immediately. Setting RTS heap size doesn't help. Does anybody happen to know why? -- Regards, Paul Liu Yale Haskell Group http://www.haskell.org/yale

2 2

code-build-test cycle
by Michael Mossey 18 Sep '09

18 Sep '09

I'm working on a GUI application in qtHaskell, and I have a bit of a bind. Using ghci, it launches quickly but runs slowly. On the other hand, compiling (mainly linking) takes a while---several minutes. The truth is that I can compile it much faster if I selectively import the needed modules, so figure the actual compilation/link time is more like 15 to 30 seconds. (This is Windows on a very old laptop.) I'm used to working in Python, so I'm used to a nearly instant code-build-test cycle, and GUI applications in PyQt run briskly, faster than ghci/qtHaskell. Now I'm wondering if Hugs is a faster interpreter. So during development I don't want to give up the quick cycle you get with an interpreter, but the application may be much too slow to use in any meaningful way without compilation. Any advice welcome. Maybe there is a way to speed up the interpretation. -Mike

4 3

memoization
by staafmeister 18 Sep '09

18 Sep '09

Hi, I participating in de google code jam this year and I want to try to use haskell. The following simple http://code.google.com/codejam/contest/dashboard?c=90101#s=p2 problem would have the beautiful haskell solution. import Data.MemoTrie import Data.Char import Data.Word import Text.Printf newtype ModP = ModP Integer deriving Eq p=10000 instance Show ModP where show (ModP x) = printf "%04d" x instance Num ModP where ModP x + ModP y = ModP ((x + y) `mod` p) fromInteger x = ModP (x `mod` p) ModP x * ModP y = ModP ((x * y) `mod` p) abs = undefined signum = undefined solve _ [] = 1::ModP solve [] _ = 0::ModP solve (hs:ts) t@(ht:tt) | hs==ht = solve ts tt + solve ts t | otherwise = solve ts t go (run, line) = "Case #"++show run++": "++show (solve line "welcome to code jam") main = interact $ unlines . map go . zip [1..] . tail . lines Which is unfortunately exponential. Now in earlier thread I argued for a compiler directive in the lines of {-# Memoize function -#}, but this is not possible (it seems to be trivial to implement though). Now I used memotrie which runs hopelessly out of memory. I looked at some other haskell solutions, which were all ugly and more clumsy compared to simple and concise C code. So it seems to me that haskell is very nice and beautiful until your are solving real algorithmic problems when you want to go back to some imperative language. How would experienced haskellers solve this problem? Thanks -- View this message in context: http://www.nabble.com/memoization-tp25306687p25306687.html Sent from the Haskell - Haskell-Cafe mailing list archive at Nabble.com.

10 18

∀ lexing in ghc and ghci
by Sean McLaughlin 18 Sep '09

18 Sep '09

Hi, I'm getting different behavior in ghci and ghc with the identifier ∀. In ghc I need to wrap it with parens, as in > (∀) :: Var -> Base -> Formula -> Formula > (∀) = All In ghci, I get an error this way Formula.lhs:112:2: Invalid type signature In ghci I can do > ∀ :: Var -> Base -> Formula -> Formula > ∀ = All fine. But then ghc complains. What's going on here? Thanks! Sean

3 7