- Haskell-Cafe - Haskell.org

safe lazy IO or Iteratee?
by David Leimbach 05 Feb '10

05 Feb '10

Hi everyone, This is not an attempt to start a flame war. I'm just trying to get a good feel for the advantages and disadvantages of the newer safe lazy io lib available on Hackage vs using Iteratee. It does appear to me that using something like Itereatee gives a bit of room to really tweak the low level enumerator/iteratee relationship, and, if you asked my old boss, I'm just a big fan of folds too, I use them a lot in my Erlang :-), so Iteratee really does appeal to me on many levels. Yet at the same time, I'm quite enamored with the beauty of "interact" and functions of that sort. I realize mixing the effects of the lazy IO and pure code may not be the clearest way to write code for everyone, but there is something about being able to get linewise data as interact (unlines . fmap someLineWiseFunction . lines) that is just kind of cool. Dave

4 5

Sparks created, but not used (par/pseq, ST monad)
by Michael Lesniak 05 Feb '10

05 Feb '10

Hello haskell-cafe, I currently have a problem with running par/pseq in the ST monad. The function testST is the minimal counterexample that works -- or, to be more clear, does not work as expected for me. As a remark, the tasks/function calls in my "real application" are much more computational expensive, but the code is too long to post here. > -- File ST.lhs > module Main where > import Control.Monad.ST > import Control.Parallel > main = testST > > testST :: IO () > testST = do > putStrLn "Starting" > (runST $ f 10) `pseq` putStrLn "Stopping" > where > f :: forall s. Int -> ST s () > f n = do > p $ "\nTask:" ++ show n > if n < 0 > then return () > else do > (n1,n2) <- return (n-1,n-2) > q n1 > q n2 > a <- f n1 > b <- f n2 > (a `par` b) `pseq` return a > > -- Some helper functions > p x = unsafeIOToST (putStrLn x) > q x = unsafeIOToST (print x) As far as I understand, compiling with ghc --make -O2 -threaded ST.lhs -o st -XRankNTypes and running with ./st +RTS -N -s should create sparks that could be run in parallel. The problem I have now, is that sparks are created but not converted, according to -s output: ... a lot of (uninteresting) stuff SPARKS: 232 (0 converted, 0 pruned) ... I have not found any information on this type of behaviour on the net and would be glad if someone could give me points or hints what's happening and how I can improve this. Cheers, Michael -- Dipl.-Inf. Michael C. Lesniak University of Kassel Programming Languages / Methodologies Research Group Department of Computer Science and Electrical Engineering Wilhelmshöher Allee 73 34121 Kassel Phone: +49-(0)561-804-6269

2 2

Re: [Haskell-cafe] safe lazy IO or Iteratee?
by John Lato 05 Feb '10

05 Feb '10

Hello, > From: David Leimbach <leimy2k(a)gmail.com> > > Hi everyone, > > Yet at the same time, I'm quite enamored with the beauty of "interact" and > functions of that sort. I realize mixing the effects of the lazy IO and > pure code may not be the clearest way to write code for everyone, but there > is something about being able to get linewise data as > > interact (unlines . fmap someLineWiseFunction . lines) > > that is just kind of cool. I think so too. I really like the "interact" style of code; I find it supremely elegant. However I also find it insufficient for many interesting cases. One general problem is the difficulty of embedding monadic code into interact functions. Usually "interact" requires a pure function. Although in many cases I think it would be possible to have interactM :: (String -> IO String) -> IO () in practice these functions don't usually exist, perhaps because they can have some surprising behavior. In my opinion IO is a weakness of Haskell, and many Haskell users either subconciously or actively seek out alternative approaches. I offer as evidence the frequency with which beginners ask the cafe about "handle closed before data could be read" errors, and also the proliferation of IO libraries. I know which I approach I think represents the best way forward, but I would be surprised if there was a community consensus at this point. In any case, I'm glad there are a multitude of options and each probably has a place and purpose. Cheers, John Lato

1 0

setNonBlockingFD?
by Magnus Therning 04 Feb '10

04 Feb '10

network-fancy fails to build because "Not in scope: 'setNonBlockingFD'". Any pointers to what should be used in 6.12? /M -- Magnus Therning (OpenPGP: 0xAB4DFBA4) magnus＠therning．org Jabber: magnus＠therning．org http://therning.org/magnus identi.ca|twitter: magthe

3 3

ANN: progression-0.1
by Neil Brown 04 Feb '10

04 Feb '10

Hi all, I've just uploaded the first release of Progression to Hackage. Progression is a small library that pulls together several other libraries and utilities (particularly Criterion) to support the optimisation of Haskell programs. To use it, you wrap up your benchmarks in a program that passes them to Progression. When you run this program, Progression records the benchmarks, and offers to draw you a graph comparing these latest times to a selection (specified by you) of previously recorded times. What you get is a graph that allows you to see if your changes have made enough difference over your benchmarks for you to keep the change (or whether it has made most of them worse, and you need to roll back!). Progression doesn't include any benchmarking logic; it delegates all that to the Criterion library, and plots the resulting mean (with confidence intervals). More details and an example graph are available in a blog post: http://chplib.wordpress.com/2010/02/04/progression-supporting-optimisation-… Feedback is welcome, either here or in the comments on the blog post. Thanks, Neil.

1 0

Translation of Haskell type classes
by Enrique Martín 04 Feb '10

04 Feb '10

Hello all, few days ago I made some experiments with Haskell type classes. I wrote a small Haskell program for searching in sorted lists, defining my own type classes for equality (MyEq) and order (MyOrd) so that they only have one member function: -------------------------------------------------------- class MyEq a where eq :: a -> a -> Bool class MyEq a => MyOrd a where less :: a -> a -> Bool data Nat = Z | S Nat instance MyEq Nat where eq Z Z = True eq Z (S x) = False eq (S x) Z = False eq (S x) (S y) = eq x y instance MyOrd Nat where less Z Z = False less Z (S x) = True less (S x ) Z = False less (S x) (S y) = less x y search :: MyOrd a => a -> [a] -> Bool search x [] = False search x (y:ys) = (eq x y) || (less y x && search x ys) -------------------------------------------------------- I also wrote the translation of this program using the classical approach of dictionaries that appears in "How to make ad-hoc polymorphism less ad hoc", Wadler & Blott 1989 or "Type Classes in Haskell", Cordelia V. Hall et. al. 1996. -------------------------------------------------------- -- From the definition of type class MyEq data DictMyEq a = DictMyEq (a -> a -> Bool) eq :: DictMyEq a -> (a -> a -> Bool) eq (DictMyEq x) = x -- From the definition of type class MyOrd data DictMyOrd a = DictMyOrd (DictMyEq a) (a -> a -> Bool) getMyEqFromMyOrd :: DictMyOrd a -> DictMyEq a getMyEqFromMyOrd (DictMyOrd x y) = x less :: DictMyOrd a -> (a -> a -> Bool) less (DictMyOrd x y) = y data Nat = Z | S Nat -- From the instance MyEq Nat eqNat :: Nat -> Nat -> Bool eqNat Z Z = True eqNat Z (S x) = False eqNat (S x) Z = False eqNat (S x) (S y) = eqNat x y dictMyEqNat :: DictMyEq Nat dictMyEqNat = DictMyEq eqNat -- From the instance MyOrd Nat lessNat :: Nat -> Nat -> Bool lessNat Z Z = False lessNat Z (S x) = True lessNat (S x ) Z = False lessNat (S x) (S y) = lessNat x y dictMyOrdNat :: DictMyOrd Nat dictMyOrdNat = DictMyOrd dictMyEqNat lessNat search :: DictMyOrd a -> a -> [a] -> Bool search _ x [] = False search dict x (y:ys) = (eq (getMyEqFromMyOrd dict) x y) || (less dict y x && search dict x ys) -------------------------------------------------------- I made some tests in GHC 6.8.2 and I noticed that the original program with type classes runs pretty faster than the translated program. For example, reducing the expression search (S Z) (replicate 1000000 Z) needs 2.07 seconds in the original program. However the translated expression search dictMyOrdNat (S Z) (replicate 1000000 Z) needs 3.10 seconds in the translated program, which is one more second. Surprised with the results, I repeated the test this time in Hugs Sept. 2006. I noticed that the difference was not so big: search (S Z) (replicate 100000 Z) --> (2100051 reductions, 2798068 cells, 2 garbage collections) search dictMyOrdNat (S Z) (replicate 100000 Z) --> (2200051 reductions, 2898067 cells, 3 garbage collections) My first idea was that type classes were implemented using the approach of dictionaries, but the test showed me that it is not true (mainly in GHC). Then I discovered the paper "Implementing Haskell overloading", Augustsson 1993, when he describes some ways to improve the speed of Haskell overloading. So my questions are: 1) is the enhancement obtained only using the optimizations of Augustsson's paper? 2) Could anyone tell me where I can find the translation of type classes that GHC and Hugs use? Thank you very much, Enrique M.

3 2

Nested Types stack size ...
by Günther Schmidt 04 Feb '10

04 Feb '10

Hi, I'm using deeply nested types in my app and exceed some stack-size in ghci. I can't remember the runtime option I have to pass to ghci anymore, it's been about a year since I last pursued this approach. Is it also possible to set this stack-size via pragma in the source code? Günther

4 6

functional references and HList?
by Günther Schmidt 04 Feb '10

04 Feb '10

Hi all, I wonder if there is some a "field of use" overlap between HList and functional references aka accessors. Do both tackle the same problem? Günther

3 3

HList darcs repo
by oleg＠okmij.org 04 Feb '10

04 Feb '10

HList (and OOHaskell, for that matter) are now hosted on community.haskell.org, in the directories /srv/code/HList /srv/code/OOHaskell The repositories are also available at these URLs: http://code.haskell.org/HList/ http://code.haskell.org/OOHaskell/ Many thanks to Ian Lynagh for moving the repositories. Ian has kindly set up the redirection so that the old, used in many places URL http://darcs.haskell.org/HList/ still works. It now redirects to code.haskell.org.

1 0

Re: [Haskell-cafe] matrix question
by John Lato 03 Feb '10

03 Feb '10

> From: Ivan Lazar Miljenovic <ivan.miljenovic(a)gmail.com> > > man <manuel.a.castro(a)gmail.com> writes: >> http://hackage.haskell.org/package/hmatrix-0.8.1.1 >> >> it wraps gls, blas and lapack (so you need to install the libraries). > > There's also the blas package if you just want blas support. I've been using blas recently with very good results. It's great if all you need are simple matrix operations. I just wanted to point out that it doesn't install on ghc-6.12.1 because the ieee dependency provides a type class instance for CLDouble, which was omitted from ghc-6.12.1. I got around this by commenting out the instance in CLDouble in the ieee package, then everything installed and runs properly (using OSX 10.6, compiled with -faccelerate). John

1 0

The errorCalls and ioErrors in extensible exceptions way
by Gracjan Polak 03 Feb '10

03 Feb '10

Hi all, I have base==3.* code that uses errorCalls and ioErrors to intercept either ErrorCall or IOError that may arise in deeper code. I'd like to convert this code to base==4.* new exceptions. -- | Evaluate the argument and catch error call exceptions errorToErr :: Monad m => a -> Err m a errorToErr x = let e = unsafePerformIO (tryJust errorCalls (evaluate x)) in ErrorT (return e) -- | Catch IO errors in the error monad. ioErrorToErr :: IO a -> Err IO a ioErrorToErr = ErrorT . liftM (either (Left . show) Right) . tryJust ioErrors Look here for more context: http://hackage.haskell.org/packages/archive/haxr/3000.5/doc/html/Network-Xm… I know that just importing OldException will do the trick for now. But I'd like to know how to do such a trick in extensible exceptions way. Also that unsafePerformIO looks a bit scary. Isn't there a better way to achieve purpose? -- Gracjan

1 0

Multi-Class monadic type?
by Alexander Treptow 03 Feb '10

03 Feb '10

Hi, i got a little problem and don't know how to solve that. Hope you can help me. code: ---------- module Test where {-# LANGUAGE Rank2Types, RankNTypes #-} import Control.Monad.Trans data TestType = TestType {tst :: (Conf a, MonadIO m) => m a} class Conf a where get :: MonadIO m => m a testFunc :: (forall a. Conf a, MonadIO m => m a) -> TestType testFunc = TestType -------------- error: -------------- Test.hs:11:23 Class `Conf' used as a type In the type signature for `testFunc': testFunc :: (forall a. Conf a, (MonadIO m) => m a) -> TestType -------------- explanation: I need a data type that creates a record with a member that has no fixed type, because its not known at compile-time. The Rank2Types language extension fits that need, but i figured out that i ll need to make that time monadic to avoid the use of unsafePerformIO in the program that uses this lib. Thanks and greetings, Alex

4 5

Hackage Package version ranges
by Neil Mitchell 03 Feb '10

03 Feb '10

Hi, I maintain the Haskell package HLint. HLint depends on haskell-src-exts, cpphs, hscolour and uniplate, plus things which are shipped with GHC. For each of the external library dependencies, I have to specify a version constraint. For example, I developed HLint against cpphs-1.10 so I can write one of: 1) cpphs == 1.10.* 2) cpphs >= 1.10 I know (1) works, since I've tested it. However, every time I use (1) some distro (usually Arch) emails me to say they need wider ranges so they can get all their packages consistent. I then usually flip to (2). Then, sometime later, someone releases a new version of cpphs that changes the interface in a way that breaks HLint, and I get emails from users, and tend to go back to (1). Flip flopping between version constraints isn't a great idea. What should I do? Alternatively, if I could sign up to be emailed when something went wrong, I'd happily fix it. i.e. I'd like an email either when my package fails to compile against the latest version of all packages but within my constrained range, or when the latest version falls outside my constraint range. Thanks, Neil

3 5

GHC/base API documentation has been generated incorrectly
by John Millikin 03 Feb '10

03 Feb '10

Looking at (for example) the docs for 'bracket'[1], the parameter documentation has been "shifted" up by one. The source code[2] looks correct, so perhaps Haddock is parsing it incorrectly? Docs for other functions, such as 'showSigned'[3], exhibit the same error. [1] http://www.haskell.org/ghc/docs/latest/html/libraries/base/Control-Exceptio… [2] http://www.haskell.org/ghc/docs/latest/html/libraries/base/src/Control-Exce… [3] http://www.haskell.org/ghc/docs/latest/html/libraries/base/Numeric.html#v%3…

3 2

Lectureships and Chair at St Andrews
by Kevin Hammond 03 Feb '10

03 Feb '10

We have vacancies for one chair (full professor) and two lectureships (assistant professor) in computer science at the University of St Andrews. It would be good to have some applications from our community. http://www.jobs.ac.uk/job/AAK643/2-lectureships-/ http://www.jobs.ac.uk/job/AAK633/chair-in-human-computer-interaction-/ Please do not be put off that the chair advert specifies "Human-Computer Interaction". Candidates with an excellent research background will always be considered. Best Wishes, Kevin -------- Kevin Hammond, Professor of Computer Science, University of St Andrews T: +44-1334 463241 F: +44-1334-463278 W: http://www.cs.st-andrews.ac.uk/~kh In accordance with University policy on electronic mail, this email reflects the opinions of the individual concerned, may contain confidential or copyright information that should not be copied or distributed without permission, may be of a private or personal nature unless explicitly indicated otherwise, and should not under any circumstances be taken as an official statement of University policy or procedure (see http://www.st-and.ac.uk) The University of St Andrews is a charity registered in Scotland : No SC013532

1 0

Game of life in haskell.
by Lyndon Maydwell 03 Feb '10

03 Feb '10

Hi Cafe. I've made a basic game of life implementation with Haskell and OpenGL: https://github.com/sordina/Life/ The basic premise is to generate a random snapshot, then iterate the successor function on it to create an infinite list of snapshots, then output them using OpenGL. I haven't really run into any major issues aside from not being able to figure out how to automatically rerun the display function, however I believe this is an OpenGL problem, and not related to the Haskell side (press 'n' to move to the next snapshot for now). To run the game: cabal configure && cabal build && ./dist/build/life/life <size of square life matrix> I'm intending to improve the performance, and add more advanced features, but I thought I'd get some feedback first. Can anyone see a way to make this code more idiomatic, or any optimizations I might have missed? I'm still fairly new to Haskell and I haven't really come to grips with monad-transformers and the like yet, so if any advanced techniques are applicable here it would really help me link what I think I understand to reality :-) Also, is something like this worth uploading to Hackage, or should I leave it on github? Thanks guys.

11 22

Re: [Haskell-cafe] Game of life in haskell.
by Jon Harrop 03 Feb '10

03 Feb '10

On Wednesday 03 February 2010 01:08:33 Richard O'Keefe wrote: > On Feb 3, 2010, at 3:04 PM, Jon Harrop wrote: > > Fair enough. If you want to model either an infinite board or a very > > sparse one then a sparse data structure will be asymptotically faster. But > > they won't be simpler and asymptotically efficient mutable data > > structures will be faster again. > > Did I say a word about mutability? The whole point of my message > was that the original message WASN'T about mutable or immutable > data structures but about dense (2d array) or sparse data structures. When Serguey said that arrays were not idiomatic Haskell because they are "hard to update functionally" I assumed he was talking about mutable vs immutable data structures. > >> A simple Haskell analogue, which has the right asymptotic cost, > > > > That will be a lot faster than using Map or Set but you're still > > paying a lot for the indirections between cons cells. Mutating an array > > in-place would be significantly faster and no more difficult to code > > correctly because this is such a trivial algorithm. > > Again, you seem to be responding to something that wasn't said. > No claim was made that the list of (row,[col...]) pairs approach > was *FAST*, Were you not talking about performance when you said "the right asymptotic cost"? > or indeed that the code to manipulate it was notable for simplicity, only > that the data structure in itself is simple. Ok. > I once had Prolog code running faster than the Fortran code it was > based on. Mutable arrays are not in themselves a guarantee of > efficiency, and while I don't imagine that you believe that, the > paragraph I quoted came uncomfortably close to saying it. My statements were specifically about Conway's Game of Life. > In fact mutating an array in place probably would be a bit trickier > because you _can't_ mutate the row you are just calculating; you will > need the old version for the next row. Yes, of course. You would read from and write to different mutable data structures, swapping over their roles. Like double buffering or Cheney's to and from spaces. > And in a sparse-compressed representation, the new and old rows will seldom > be the same size. Just use a conventional resizeable array and it will amortize all those costly allocations for you without complicating the code at all. > For what it's worth, the sparse-compressed code that someone pointed > to, and which I alluded to, is written in C. Remember, the original > message didn't say *arrays* were a poor choice, He said: "Arrays are not fully "idiomatic" for Haskell as they are hard to update functionally." Is that not saying that arrays were a poor choice? > but that the "obvious" *2d array of boolean* was a poor choice. And that's > the (only) claim I defended. The OP was solving a dense problem on a finite board. Is a 2d array of booleans really such a bad choice in that context? Even if you change the problem to a sparse one, how large must it be before a solution like Serguey's becomes competitively efficient in practice? -- Dr Jon Harrop, Flying Frog Consultancy Ltd. http://www.ffconsultancy.com/?e

1 0

matrix question
by 조광래 03 Feb '10

03 Feb '10

define functions type Matrix=[[Double]] multMM :: Matrix -> Matrix -> Matrix --multiplies two matrices det :: Matrix -> Double --computes the determinant of a matrix inv :: Matrix -> Matrix --inverts a matrix i stuck on those problems can any one help me out?

7 6

About code style ?
by zaxis 03 Feb '10

03 Feb '10

For me i like C style instead of layout. For example, func1 a = do -- ... a * 2 -- ... I always write it as: func1 a = do { -- ...; a * 2; -- ...; } However, i donot know how to write pure function using C style. func1 a = { -- ...; a * 2; -- ...; } will not compile without `do`. Sincerely! ----- fac n = foldr (*) 1 [1..n] -- View this message in context: http://old.nabble.com/About-code-style---tp27414627p27414627.html Sent from the Haskell - Haskell-Cafe mailing list archive at Nabble.com.

8 9

Re: [Haskell-cafe] Game of life in haskell.
by Tony Finch 03 Feb '10

03 Feb '10

On Tue, 2 Feb 2010, edgar(a)ymonad.com wrote: > http://dotat.at/prog/life/hslife.hs Er, yes, I didn't link to that in my earlier message because it's a half-completed attempt. I think I got to the stage of realising that I needed to write a monad or a monad transformer to thread the hash cons state through the code without it getting in the way. For those interested in Hash Life whether or not it's written in Haskell, there's a description of how it works in http://dotat.at/prog/life/hashlife.c though that code is also incomplete. Tony. -- f.anthony.n.finch <dot(a)dotat.at> http://dotat.at/ GERMAN BIGHT HUMBER: SOUTHWEST 5 TO 7. MODERATE OR ROUGH. SQUALLY SHOWERS. MODERATE OR GOOD.

2 1

SYB/Generics documentation inaccessible
by Arne D Halvorsen 02 Feb '10

02 Feb '10

I have tried to access a few of the documents on generics that are referred to from wikipedia, from the GHC documentation, from hackage and from the wiki. These documents (on vu.nl, subdirectory boilerplate and strafunski) are apparently forbidden to me, and have been for some weeks. I tried to mail a webmaster and some other mentioned mail addresses, but the mails bounced. Could somebody make the docs accessible, or should it be re-hosted on Hackage? Regards Arne D Halvorsen

5 5

Failure to load hmatric
by dlb＠patriot.net 02 Feb '10

02 Feb '10

OK, I'm working on matrix stuff in Haskell now (I've been trying to get the professor to approve that) and when I use cabal install to install hmatrix, it fails at HUnit with: ----------- Configuring HUnit-1.2.2.1... Preprocessing library HUnit-1.2.2.1... Preprocessing executables for HUnit-1.2.2.1... Building HUnit-1.2.2.1... [1 of 5] Compiling Test.HUnit.Terminal ( Test/HUnit/Terminal.hs, dist/build/Test/HUnit/Terminal.o ) [2 of 5] Compiling Test.HUnit.Lang ( Test/HUnit/Lang.hs, dist/build/Test/HUnit/Lang.o ) [3 of 5] Compiling Test.HUnit.Base ( Test/HUnit/Base.hs, dist/build/Test/HUnit/Base.o ) [4 of 5] Compiling Test.HUnit.Text ( Test/HUnit/Text.hs, dist/build/Test/HUnit/Text.o ) [5 of 5] Compiling Test.HUnit ( Test/HUnit.hs, dist/build/Test/HUnit.o ) /usr/bin/ar: creating dist/build/libHSHUnit-1.2.2.1.a Test/HUnit/Lang.hs:22:1: lexical error at character 'i' cabal: Error: some packages failed to install: HUnit-1.2.2.1 failed during the building phase. The exception was: exit: ExitFailure 1 ---------------- Short term question: what the heck do I do with this? Long term question: how can I handle cabal failures of this type without bothering the list every time? Dave Barton University of Toronto

2 1

Tracer for Haskell showing substitutions
by Ulrik Rasmussen 02 Feb '10

02 Feb '10

Hi all, I was wondering if someone has written a tracer/debugger that shows you how a given Haskell expression is evaluated, by generating all the intermediate states of the expression until it is in normal form? For instance, given the following code: > take' 0 xs = [] > take' n (x:xs) = x : take' (n-1) xs > exp = take' 2 [1,2,3,4,5,6] the trace of 'exp' would generate something like this: > exp = take' 2 [1,2,3,4,5,6] > exp = (\n (x:xs) -> x : take' (n-1) xs) 2 [1,2,3,4,5,6] > exp = 1 : take' (2-1) [2,3,4,5,6] > exp = 1 : take' 1 [2,3,4,5,6] > exp = 1 : (\n (x:xs) -> x : take' (n-1) xs) 1 [2,3,4,5,6] > exp = 1 : 2 : take' (1-1) [3,4,5,6] > exp = 1 : 2 : take' 0 [3,4,5,6] > exp = 1 : 2 : (\0 xs -> []) 0 [3,4,5,6] > exp = 1 : 2 : [] > exp = [1,2] That is, all the substitutions performed when evaluating 'exp' from left to right. I was thinking that something like this could be rather useful when teaching or learning Haskell. Thanks, Ulrik

2 1

want to post haskell question
by 조광래 02 Feb '10

02 Feb '10

hi I have a question about haskell coding. Matrices One of many ways to define matrices in Haskell is the list of matrix rows, where a row is a list of double precision oating point numbers: type Matrix=[[Double]] Using this type you should define the following functions (definition of dim is given as an inspirational example). dim :: Matrix -> (Int,Int) --returns matrix dimension (number of rows and columns) dim m = if (not.isMatrix) m then error "Not a matrix" else (length m, length (head m)) isMatrix :: Matrix -> Bool --checks whether all rows have the same length isMatrix m = and [length (head m) == length a | a <- m] I have done upto here and I am stuck on multSM. multSM :: Double -> Matrix -> Matrix --multiplies a scalar and a matrix when I do: multSM d m = [[(b*a)| b<-[d], a<-(head m)]] or [map (*d) (head m)] ...I get (using Hugs): Main> multSM 3 [[2,3,4],[1,3,4],[6,3,0]] [[6.0,9.0,12.0]] So I solve 3 [2,3,4] but I do not know how to get the rest 3 [[1,3,4],[6,3,0]. Please help me on this problem.

3 2

Re: [Haskell-cafe] Very imperfect hash function
by John Lato 02 Feb '10

02 Feb '10

> From: Hans Aberg <haberg(a)math.su.se> > > On 28 Jan 2010, at 20:07, Steve Schafer wrote: > >> The data are currently in a large lookup table. To save space, I'd >> like >> to convert that into a sort of hash function: >> >> hash :: key -> value >> >> My question is this: Is there any kind of generic approach that can >> make >> use of the knowledge about the internal redundancy of the keys to come >> up with an efficient function? > > There are minimal perfect hash functions; there are some libraries > mentioned here, though they are not in Haskell code: > http://en.wikipedia.org/wiki/Perfect_hash_function > > This is suitable when you do a lot of lookups with few key updates. An > alternative might be Data.Map, where lookups have time complexity > O(log n), n = size of map. The minimal perfect hash function looks like the real algorithmic solution, but I would just use Data.IntMap for this. Cheers, John

3 7

Very imperfect hash function
by Steve Schafer 02 Feb '10

02 Feb '10

I'm looking for some algorithmic suggestions: I have a set of several hundred key/value pairs. The keys are 32-bit integers, and are all distinct. The values are also integers, but the number of values is small (only six in my current problem). So, obviously, several keys map to the same value. For some subsets of keys, examining only a small portion of the key's bits is enough to determine the associated value. For example, there may be 250 keys that all have the same most-significant byte, and all 250 map to the same value. There are also keys at the other extreme, where two keys that differ in only one bit position map to different values. The data are currently in a large lookup table. To save space, I'd like to convert that into a sort of hash function: hash :: key -> value My question is this: Is there any kind of generic approach that can make use of the knowledge about the internal redundancy of the keys to come up with an efficient function? Steve Schafer Fenestra Technologies Corp. http://www.fenestra.com/

5 4

Re: Non-termination of type-checking
by oleg＠okmij.org 01 Feb '10

01 Feb '10

Here is a bit more simplified version of the example. The example has no value level recursion and no overt recursive types, and no impredicative polymorphism. The key is the observation, made earlier, that two types c (c ()) and R (c ()) unify when c = R. Although the GADTs R c below is not recursive, when we instantiate c = R, it becomes recursive, with the negative occurrence. The trouble is imminent. We reach the conclusion that an instance of a non-recursive GADT can be a recursive type. GADT may harbor recursion, so to speak. The code below, when loaded into GHCi 6.10.4, diverges on type-checking. It type-checks when we comment-out absurd. {-# LANGUAGE GADTs, EmptyDataDecls #-} data False -- No constructors data R c where -- Not recursive R :: (c (c ()) -> False) -> R (c ()) -- instantiate c to R, so (c (c ())) and R (c ()) coincide -- and we obtain a recursive type -- mu R. (R (R ()) -> False) -> R (R ()) cond_false :: R (R ()) -> False cond_false x@(R f) = f x absurd :: False absurd = cond_false (R cond_false)

7 9

very strange utf8 problem
by Günther Schmidt 01 Feb '10

01 Feb '10

Hi all, I know this sounds daft but I do have good reason to ask. Is it possible that GHC's core itself has a problem with a particular Umlaut only? HDBC-ODBC won't read in data from an SQLite database as soon as it comes accross a *lowercase* U-Umlaut ("ü") ghci crashes. Other Umlauts ("ä", "ö" and "ß") pass however. This is the error message: readUTF8Char: illegal UTF-8 character 252 As I said, other Umlauts do pass. I got this very message quite often, usually when working with databases and not always using HDBC but also Takusen. I kept blaming it on HDBC-ODBC or the input data but I might have been wrong there. Günther

2 1

OT: Literature on translation of lambda calculus to combinators
by Dušan Kolář 01 Feb '10

01 Feb '10

Dear cafe, Could anyone provide a link to some paper/book (electronic version of both preferred, even if not free) that describes an algorithm of translation of untyped lambda calculus expression to a set of combinators? Preferably SKI or BCKW. I'm either feeding google with wrong question or there is no link available now... Thanks, Dušan

10 11

Determining application directory
by Matveev Vladimir 01 Feb '10

01 Feb '10

Hi, I'm writing cross-platform application in Haskell which should be running under Windows and Linux. Under Linux configuration is stored in the /etc directory, and under Windows configuration is meant to be in the application directory. So, is there a way to get an application directory path under Windows? I remember that there is a way to do this using WinAPI, but how to do this Haskell?

7 9

Greencard/c2hs + cabal
by Maciej Piechotka 01 Feb '10

01 Feb '10

I have a problem with integrating cabal with greencard. 1. When I run build it creates: src/Path/To/Module_stub_ffi.c src/Path/To/Module_stub_ffi.h instead of dist/build/Path/To/Module_stub_ffi.c dist/build/Path/To/Module_stub_ffi.h 2. I get the unknown symbols errors for symbols defined in stubs. The problem is that I need to import the macros constants c2hs puts #c/#endc in header which is not compiled (and therefore suffers from the same problems). Is there any way to get the value of macro without having additional c file? Regards

1 0

HWN Issue 148 -- unicode error
by Joe Fredette 31 Jan '10

31 Jan '10

Indeed, the Unicode only made it through my system, and not the mailer itself. In fact, the name should look something like "Ćwikłowski", and not the garbled mess that made it through. Apologies again, Bartek! /Joe

1 0

Haskell Weekly News: Issue 148 - January 31, 2010
by jfredett＠gmail.com 31 Jan '10

31 Jan '10

--------------------------------------------------------------------------- Haskell Weekly News http://sequence.complete.org/hwn/20100131 Issue 148 - January 31, 2010 --------------------------------------------------------------------------- Welcome to issue 148 of HWN, a newsletter covering developments in the [1]Haskell community. Hello Haskellers, this week begins with a correction. Last week, I noted a blogpost from one 'Bartek Paczesiowa', I was rapidly informed of the complete wrongness of this citation. In fact, Bartek's last name is, and I hope the Unicode makes it through here, 'wikowsk,' and Paczesiowa is his nickname. Apologies to Bartek, as penance I've reread your [2]excellent post on your blog. That said, this week's HWN if packed full of awesome potential naming errors, so I'll let you get to it. Thus, Haskellers, Your Haskell Weekly News! Announcements haskell-src-exts-1.8.0. Niklas Broberg [3]announced a new release of haskell-src-exts. Elerea 1.2.3 with some enhancements. Patai Gergely [4]announced his addition of some new features to the experimental branch of his FRP library, Elerea. hakyll-1.3. Jasper Van der Jeugt [5]announced the release of hakyll 1.3, including several new improvments and changes. ThreadScope 0.1. Satnam Singh [6]announced the release of Threadscope 0.1, the premier graphical thread profiler. The Monad.Reader Issue 15. Brent Yorgey [7]announced the most recent issue of the Monad.Reader, a monthly publication of Haskell Related expostition and discusson articles. afv-0.1.0. Tom Hawkins [8]announced a afv-0.1.0, an infinite state model checker for verifying assertions about embedded C programs. adaptive-tuple 0.1.0. John Lato [9]announced the initial release of adaptive-tuple, his library for combining the space-efficient properties of tuples with the utility of lists. Job opportunities at Citrix Systems (Cambridge, UK). Matthias Goergens [10]announced a opportunity available at Citrix Systems in Cambridge, UK. Discussion OT: Literature on translation of lambda calculus to combinators. Dusan Kolar [11]asked about texts regarding translating the untyped lambda calculus to a combinator calculus such as SKI or BCKW. Linguistic hair-splitting. Andrew Coppin [12]asked an interesting offtopic linguistic question about what we call a number, a field, an element, and a monad. Adopting hpodder? John Goerzen [13]asked if there were any interested adoptive maintainers for his hpodder project. This is an excellent opportunity for a Haskell Neophyte to help the community and learn about project management hopefully we can find hpodder a new maintainer! Blog noise [14]Haskell news from the [15]blogosphere. Blog posts from people new to the Haskell community are marked with >>>, be sure to welcome them! * Holumbus: [16]HolHac at FH-Wedel. * ++Arch Haskell Team++: [17]Arch Linux updates to GHC 6.12. * Neil Brown: [18]Exploring a Communicating Sequential Processes EDSL. * The GHC Team: [19]Yielding more improvements in parallel performance. Quotes of the Week * Berengal: I'm going to write a module Hmm with a (.) operator in it, so I can go 'Hmm..' in my code * clarkb,: in CS they dont teach you to program...You learn Data Structures, Algorithms, Logic, Discrete Math, Language theory, etc and happen to pick up programming on the way * dons: hey i love core. i dream about unboxes * cale: Differential geometry is the study of manifolds under change of notation. * kmc: the irony being, the abstraction that gets the most complaining and general noise [from imperative programmers] is the one that captures imperative programming * kmc: i am Jack's monad operator * arw: ...and a basic law of haskell is, 50% of all documentation has to be monad tutorials :) * Cale: Here [#haskell], we feed trolls until they explode. * bartek: It took me 2 years of studying teachings of Oleg Kiselyov (who was raised among types, where he learned to speak their language), but finally, I have the solution. * kmc: I think 250 milliolegs is enough to kill an elephant olsner: ... to kill an elephant - in the type system! * syntaxglitch: every time I have a cool idea about something that might work in Haskell, I go check Oleg's stuff and find that 1) he already did it 2) thought it out better 3) did it incidentally while working on something way more interesting * DRMacIver:: I dread to think what category theory would look like after the software engineering world had got their grubby paws on it. Enterprise variant functors. Commutative UML diagrams. About the Haskell Weekly News New editions are posted to [20]the Haskell mailing list as well as to [21]the Haskell Sequence and [22]Planet Haskell. [23]RSS is also available, and headlines appear on [24]haskell.org. To help create new editions of this newsletter, please see the information on [25]how to contribute. Send stories to jfredett . at . gmail . dot . com. The darcs repository is available at darcs get [26]http://patch-tag.com/r/jfredett/HWN2/pullrepo HWN2 . References 1. http://haskell.org/ 2. http://paczesiowa.blogspot.com/2010/01/pure-extensible-exceptions-and-self.… 3. http://article.gmane.org/gmane.comp.lang.haskell.cafe/69749 4. http://article.gmane.org/gmane.comp.lang.haskell.cafe/69745 5. http://article.gmane.org/gmane.comp.lang.haskell.cafe/69731 6. http://article.gmane.org/gmane.comp.lang.haskell.cafe/69612 7. http://article.gmane.org/gmane.comp.lang.haskell.cafe/69535 8. http://article.gmane.org/gmane.comp.lang.haskell.cafe/69517 9. http://article.gmane.org/gmane.comp.lang.haskell.general/17764 10. http://article.gmane.org/gmane.comp.lang.haskell.general/17763 11. http://thread.gmane.org/gmane.comp.lang.haskell.cafe/69643 12. http://thread.gmane.org/gmane.comp.lang.haskell.cafe/69620 13. http://thread.gmane.org/gmane.comp.lang.haskell.cafe/69528 14. http://planet.haskell.org/ 15. http://haskell.org/haskellwiki/Blog_articles 16. http://holumbus.fh-wedel.de/blog/?p=32 17. http://archhaskell.wordpress.com/2010/01/31/arch-linux-updates-to-ghc-6-12/ 18. http://chplib.wordpress.com/2010/01/27/exploring-a-communicating-sequential… 19. http://ghcmutterings.wordpress.com/2010/01/25/yielding-more-improvements-in… 20. http://www.haskell.org/mailman/listinfo/haskell 21. http://sequence.complete.org/ 22. http://planet.haskell.org/ 23. http://sequence.complete.org/node/feed 24. http://haskell.org/ 25. http://haskell.org/haskellwiki/HWN 26. http://patch-tag.com/r/jfredett/HWN2/pullrepo%20HWN2

1 0

Recursively traversing a data structure with HXT
by Mads Lindstrøm 31 Jan '10

31 Jan '10

Hi I am trying to use HXT (8.3.2) for parsing XML. I think an example will clarify what I want to do. I want to parse XML like this: <object class="ex1"> <foo/> <foo> <object class="ex2"/> <object class="ex3"/> </foo> <object class="ex4"/> </object> and want to turn it into the following Haskell data structure: data Widget = Widget { cls :: String , children :: [Widget] } The XML above should be turned into: Widget "ex1" [Widget "ex2" [], Widget "ex3" [], Widget "ex4" []] That is, I want everything but the object-tags stripped out. And I want to keep the hierarchy of the object-tags. I thus wrote the following program: {-# LANGUAGE Arrows, NoMonomorphismRestriction #-} module Main where import Text.XML.HXT.Arrow data Widget = Widget { cls :: String , children :: [Widget] } deriving Show main = (runX (readDocument [(a_validate,v_0)] "test.xrc" >>> getObject)) >>= print getObject = deep (isElem >>> hasName "object") >>> proc x -> do cls <- getAttrValue "class" -< x cs <- listA getObject <<< getChildren -< x -- recursive call here returnA -< Widget cls cs But it do not work as intended. In stead I get the following output: [Widget {cls = "ex1", children = []},Widget {cls = "ex1", children = []},Widget {cls = "ex1", children = []},Widget {cls = "ex1", children = []},Widget {cls = "ex1", children = []},Widget {cls = "ex1", children = []},Widget {cls = "ex1", children = []}] Hopefully somebody can point me in the right direction. Greetings, Mads Lindstrøm

1 0

ANN: haskell-src-exts-1.8.0
by Niklas Broberg 31 Jan '10

31 Jan '10

Fellow Haskelleers, I'm pleased to announce the release of haskell-src-exts-1.8.0! * On hackage: http://hackage.haskell.org/package/haskell-src-exts * Via cabal: cabal install haskell-src-exts * Darcs repo: http://code.haskell.org/haskell-src-exts For users of hsx, the latest version 0.6.1 works unchanged with this release. 1.7.2 --> 1.8.0 =============== * Add an instance Show Fixity (derived). * Support for the new ANN and INLINE_CONLIKE pragmas. * Remove support for CFILES and INCLUDE pragmas. The support wasn't correct anyway, as it assumed the pragmas appeared at the top of files. As CFILES/INCLUDE pragmas can (and do) appear anywhere, there's no hope to support them in the AST. Better to remove the support altogether. Files with CFILES/INCLUDE pragmas can still be parsed of course, but those pragmas will be handled as comments. * Parsing with ignoreLinePragmas = False now correctly updates the file name. * Allow the whole SPECIALISE/INLINE family of pragmas in instance declarations. The InsInline constructor is removed, and is now represented by 'InsDecl (InlineSig ...)'. * Fix a bug with line numbering and quasi quotes, and a similar one with line numbering and CDATA. * Fix a few minor bugs in the exactPrinter. * Fix the strange handling of so called strings in LINE pragmas. Cheers, /Niklas

1 1

ANN: Elerea 1.2.3 with some enhancements
by Patai Gergely 31 Jan '10

31 Jan '10

Hello all, I added a few new features to the experimental branch of Elerea [1], my little FRP library. The previous release added random sources -- both noise signals and instantaneous values within generators -- using mersenne-random, and the current update added an extra primitive to handle external events that can occur several times between samplings, as I heard some people needed this. All the three variants of the experimental interface support these additions. Here's a simple example showing how the new externalMulti construct can be used: import Control.Monad import FRP.Elerea.Experimental.Simple main = do (gen,sink) <- externalMulti sig <- start $ do vals <- gen cnt <- stateful 10 (+1) return (liftM2 (:) cnt vals) let act = print =<< sig act sink 3 act >> act sink 8 >> sink 2 act sink 1 >> sink 7 >> sink 9 act >> act >> act sink 4 act The program prints the following: [10] [11,3] [12] [13,2,8] [14,9,7,1] [15] [16] [17,4] As you can see, all the input received since the last sampling is collected in a list in reverse order. It is up to the programmer to store these values somewhere or build further IO actions out of them. Gergely [1] http://hackage.haskell.org/package/elerea -- http://www.fastmail.fm - Does exactly what it says on the tin

1 0

Strange random choice algorithm
by michael rice 31 Jan '10

31 Jan '10

I'm not sure where I got this PICK function from, and don't understand why it's written as it is, so I wanted to test it for randomness. It seems random enough. But if I understand the algorithm correctly, instead of selecting one of the elements from the list, it eliminates all the elements but one and that's the value it returns. Seems like a roundabout way of doing it. Comments? Also, is there a more direct way of printing an array? Output below. Michael ================= import System.Random import Data.Array.IO pick :: [a] -> IO a pick [] = undefined pick [x] = do return x pick (x:xs) = pick' x xs (2 :: Int) pick' :: (Num p, Random p) => t -> [t] -> p -> IO t pick' curr [] _ = do return curr pick' curr (next:rest) prob = do r <- getStdRandom (randomR (1,prob)) let curr' = if r == 1 then next else curr pick' curr' rest (prob+1) main = do arr <- newArray (1,9) 0 :: IO (IOArray Int Int) doLoop arr [1,2,3,4,5,6,7,8,9] 0 doLoop arr z k = do p <- pick z a <- readArray arr p writeArray arr p (a+1) if k > 10000 then do v <- readArray arr 1 print v v <- readArray arr 2 print v v <- readArray arr 3 print v v <- readArray arr 4 print v v <- readArray arr 5 print v v <- readArray arr 6 print v v <- readArray arr 7 print v v <- readArray arr 8 print v v <- readArray arr 9 print v else do doLoop arr z (k+1) =============== [michael@localhost ~]$ runhaskell array1.hs 1110 1117 1080 1169 1112 1119 1137 1084 1074 [michael@localhost ~]$

5 6

C functional programming techniques?
by Maurício CA 31 Jan '10

31 Jan '10

Hi, all, Sorry if this looks weird, but do you know of experiences with functional programming, or "type programming", with C? Using macro tricks or just good specifications? I know this is not absurd to a small extent. I've heard of proof tool certificated C code on the net (although I don't understand that theory), and I was also able to use what I learned from Haskell in a few C-like tasks. [*] So, I imagine if someone has done such kind of experiment seriously, like others have done with object-oriented programming (gobject etc.). Google told me nothing. Thanks, Maurício [*] This may actually be interesting. I had to use a proprietary script language mixing the worst of FORTRAN and C to drive an application, and I tried to use what I learned from Haskell by, say, prohibiting side effects out of some isolated code areas that got well specified input and gave well specified output -- even if the inside of the block itself was a huge mess. The result was actually fun to use, and after two years without looking at that code I was asked to adapt the application to be used inside another, which I could do in half a day! Maybe I should write a "Why functional programming matters - 2" on that :) It's easy to show FP is cool with abstract problems, but it's far more cool when we can show it can save even programmers with no CS theory background from huge headaches.

4 5

ANN: hakyll-1.3
by Jasper Van der Jeugt 30 Jan '10

30 Jan '10

Hello, Today I'd like to announce the release of hakyll-1.3. Changes since 1.2: - Categories were added (as opposed to tags). You can find some information in this tutorial[1]. - createListing and createListingWith functions were added, as a way of having a more high-level way to create listings. All tutorials have been updated to work with listings now, a short explanation is available here[2]. In short, this basically replaces the old and verbose renderAndConcat/createCustomPage workflow. Changes since 1.1: - Rewrote caching system. - Rewrote templating system. - 4x speed improvement, mostly because of the Data.Binary library. You can get the latest version from hackage[3]. All feedback/criticism/comments are welcome. Kind regards, Jasper Van der Jeugt [1]: http://jaspervdj.be/hakyll/tutorial6.html [2]: http://jaspervdj.be/hakyll/tutorial3.html#custom-pages [3]: http://hackage.haskell.org/package/hakyll

1 0

Why cannot i add the `let` declaration ?
by zaxis 30 Jan '10

30 Jan '10

find_the_day sDay 0 = sDay find_the_day sDay nDay = if (is_trade_day $ nextDay sDay) then find_the_day (nextDay sDay) (nDay - 1) else find_the_day (nextDay sDay) nDay nextDay sDay = addDays 1 sDay The above code works fine. But the following always reports `compiling error`: find_the_day sDay 0 = sDay find_the_day sDay nDay = let nextDay = addDays 1 sDay if (is_trade_day $ nextDay) then find_the_day nextDay (nDay - 1) else find_the_day nextDay nDay Any suggestion is appreciated! ----- fac n = foldr (*) 1 [1..n] -- View this message in context: http://old.nabble.com/Why-cannot-i-add-the-%60let%60-declaration---tp273818… Sent from the Haskell - Haskell-Cafe mailing list archive at Nabble.com.

4 5

where is the eros distribution
by Thomas Hartman 30 Jan '10

30 Jan '10

I was inspired by the google tech talk and would like to install and play with eros, but the http://darcs.haskell.org/packages/Eros/dist/ address pointed to at http://www.haskell.org/haskellwiki/Eros appears to be obsolete.

3 2

Re: [Haskell-cafe] begginers' questions
by Han Joosten 29 Jan '10

29 Jan '10

knyttr wrote: > > 1. suppose I have a function like this > > fun x y z = ... > fun x y [] = ... > fun x [] [] = ... > where > ... > > the "where" term will be applied just to the last definition. is it > possible to force it to all "fun" definitions? > The last two parts of your definition of fun will never be reached. The order of the definition is relevant. The first 'match' will be used. In this case, even if the third parameter of fun is an empty list, the part fun x y z =... is a match. In this case, changing the order to fun x [] [] = ... fun x y [] = ... fun x y z = ... would make more sence. however, there still is overlap. (ghc could generate a warning for this) now the where part: there are several options: fun x y z | z == [] = ... | y == [] = ... | otherwise = ... where ... or using a case expression... fun x y z = case (x,y,z) of (_ , _ , []) -> ... (_ , [] , [_:_]) -> ... (_ , [_:_], [_:_]) -> ... where ... pattern matches do not overlap this way. you could use this pattern matches also in the previous solution. Then your second question: > 2. how can I change a field in a list in the most easy way? the only > solution I found is below. I am not very happy with it - it does too many > operations just to change one field. > > changeIn array index value = (take (index-1) array)++[value]++(drop index > array) > > Remember, in haskell you do not 'change' values in variables. Thers is no such thing as a variable. You probably want a function that takes a list, an index and a value. It produces a list with all elements from the first list, only the n-th element is replaced with the given value. I would specify this as follows: f x:xs i v | i == 1 = v:xs | i > 1 = x : (f xs (i-1) v) but there are probably dozens of other ways to do this. While you are new to Haskell, do not worry too much about performance. (ghc will surly surprise you =^D) The main thing is to think functions. If you do that, you will appreciate Haskell for sure. -- View this message in context: http://old.nabble.com/begginers%27-questions-tp27377905p27378623.html Sent from the Haskell - Haskell-Cafe mailing list archive at Nabble.com.

2 1

problems installing cabal on windows
by Han Joosten 29 Jan '10

29 Jan '10

I recently upgraded to GHC 6.12.1 on my windows system. Now I have trouble getting cabal right. I read the forums, and I got the cabal.exe for windows. I think I followed the upgrade instructions. However I do not get it all to work. THis is my version of Cabal: C:>cabal --version cabal-install version 0.6.2 using version 1.6.0.2 of the Cabal library However, when I try to install any library, I get: C:>cabal install graphviz cabal: failed to parse output of 'ghc-pkg dump' I get the feeling that I miss some step during installation, but I cannot figure out what that should be. I am the only user of my windows machine, and I have admin rights. Any suggestions are welcome! Han Joosten -- View this message in context: http://old.nabble.com/problems-installing-cabal-on-windows-tp27363309p27363… Sent from the Haskell - Haskell-Cafe mailing list archive at Nabble.com.

3 3

Hierachical abstraction
by Andrew Coppin 29 Jan '10

29 Jan '10

OK, this is going to be a long one... Let's start with parsers. A monadic parser library (e.g., Parsec) gives you a set of primitive parsers. It also supplies functions to construct more complex commonly-used parsers. And then it provides functions for constructing even more complex parsers. (E.g., Parsec's expression parser builder.) Let's call the most basic primitive parsers level-0. Then we have level-1 functions, which construct parsers from level-0 parsers. And later we have level-2 functions which call the level-1 functions to construct even more sophisticated parsers. And so on and so forth, to unlimited depth. (The parser library itself has a limited depth, but the library can have clients that build functions on top of functions.) And so, in the end, the user calls a level-64 function, and gets a parser that does the required task. And that's great. The only problem is, the parser you get is a black box. You can't see what's inside it or what it's going to do. You can only run it. (Or combine it with other parsers.) Why is that a problem? Well, suppose you want to pass the finished parser through some kind of optimiser to rearrange its internal structure. Ah. You can't. It's a black box. Veering off onto a tangent for a moment... It seems to me that this is an inevitable consequence for any monadic parser. After all, the bind method runs some parse primitive, but it then feeds the result to a general function, which can undertake some arbitrarily complex Turing-complete calculation to decide what parse primitive to run next. Hell, I can write a parser that parses a description of a grammar, dynamically constructs a parser for that grammer, and then runs it! There's no way in hell that some static optimiser can hope to optimise the meta-parser so that it always generates optimal parsers. Even in theory it's utterly impossible. I wonder if you can make a parser combinator library which is *not* monadic? And, if you could, in what way would that limit the kinds of things you can parse? Anyway, coming back on topic... Let's suppose that instead of parsers, we've interested in queries. Suppose I'm writing some kind of relational database engine library, and I want to optimise queries before running them. The obvious solution is to design some kind of algebraic data type which can represent every possible level-0 query primitive. I can then write level-1 functions that construct level-0 data structures, and level-2 functions that call the level-1 functions, and so on, and at the end the caller will get a (presumably complex) level-0 description, which my optimiser can then go away and chew on. Now suppose I want to perform optimisations on the higher-level structure of the query. Currently, I can't. The hierachy of abstractions gets flatterend to level-0 in the process of the call stack, so my optimiser can only see the final level-0 representation. The obvious solution would be to design an algebraic data type for every level of the hierachy you want to analyse, complete with a function for converting to the next level down. As I understand it, this is how GHC works. It starts with raw Haskell source code, which it mashes around, and eventually converts into Core. The Core representation is then fiddled around with further, and eventually turned into STG. The STG gets poked and prodded, and converted to C--. And that eventually becomes raw assembly. The thing is, these are all radically different representations of the program. You can't mix (say) Haskell and STG together. And nor would you want to, given that they're such wildly different abstractions. [I notice in passing that, apparently, you can now FFI into C--, which is interesting...] However, consider the initial Haskell input. Initially it reflects what the programmer typed in. But GHC gradually desugars this into some minimal subset of the full Haskell language, before actually converting it to Core. That means that [picking an example out of thin air] initially an expression might contain the if/then/else construct, but at some point [I presume] this is desugared into a case expression. And that must mean that there are parts of the GHC codebase where it is permissible for an expression to contain if/then/else, and other parts where this is not permissible (and the code expects this construct to not be present). Is it possible to encode this kind of expectation into the type system? Or does GHC just tacitly rely on the fact that source code takes a predetermined path through the compiler engine? (That works for a monolithic product like a compiler, but it wouldn't be so hot for a library that 3rd parties can poke and prod arbitrary data into.) I'm basically looking at a problem which is like this. There are primitive constructs, and there are more sophisticated abstractions built from these, and I would like to come up with a system where abstractions are first-class, new abstractions can be added, and for any particular data structure, you can statically guarantee which abstractions are or aren't present. Has anybody ever looked at a general solution to something like that? [I'm going to stop talking now...]

5 9

Include creation date in Haddock documentation
by Dennis Walter 29 Jan '10

29 Jan '10

Hi all, I wonder if it is possible to have Haddock include the date and time at which the documentation was generated. Ideally this should only occur on index.html, but to have it in any file (e.g. in the footer) would be OK, too. I couldn't find any information about dates in the documentation. Documentation is created via Cabal: cabal haddock --executables --hyperlink-source Right now I include a token in the Cabal "description" field of the package, and sed-replace it afterwards, which is of course ugly. Thanks!

2 1

Community.haskell.org is down
by Neil Mitchell 29 Jan '10

29 Jan '10

http://downforeveryoneorjustme.com/community.haskell.org/ Thanks, Neil

1 0

imaging textbooks?
by Dan Mead 29 Jan '10

29 Jan '10

Hey all Can anyone recommend a good textbook on computer vision or image processing? I don't care if it favors Haskell or not, I'm just trying to find a good text. -Dan

4 3

Input request: which interface for Web Application Interface
by Michael Snoyman 29 Jan '10

29 Jan '10

I've made a few more changes to WAI since my last update. Namely, request and response headers are now their own datatype. As with Method, HttpVersion and Status, they both provide constructors for "everything else" and have functions to convert to/from bytestrings. This makes the package a bit more consistent, as well as giving us some more type safety and perhaps even some (negligible) performance gains. (Thanks again to Mark for the idea.) Another feature that will be added is a wrapper module. This will provide convenience functions for accessing the request body and generating a response body via other methods. For example, it will provide a lazy I/O interface for the request body and generating a response from a lazy bytestring. (Thanks to Nicolas for the idea.) The reason that it has not been written yet is because I have two options available for the request and response body interfaces. One is a direct copy of Hyena's enumerators, sans impredicative types. The other is a simpler I/O interface. Here are the benefits of each as I see it: Hyena enumerators: Provide accumulating parameter support. Provide mechanism for early response termination. Probably makes the server side of the request body simpler to program. Simple interface: More obvious how to use it. Makes the application side of the request body simpler to program. You can see the code for each in github: Enumerator: http://github.com/snoyberg/wai/blob/enumerator/Network/Wai.hs Simple: http://github.com/snoyberg/wai/blob/master/Network/Wai.hs I'm currently leaning to the enumerator approach, since it should be simple application side to wrap the enumerator into a more tractable interface. However, I'd like to hear people's opinions on this. To get an idea of how this changes code, I'm attaching a diff of the two branches, which includes a simple HTTP server and application. I'm also aware that documentation is sorely lacking right now; once we come to a decision on this piece of the interface, I'll add the wrapper functions and comment appropriately. Commenting at this stage would be putting the cart before the horse. Michael

1 0

big problems with MS Access backends
by Günther Schmidt 29 Jan '10

29 Jan '10

Hi, I need to use MS Access as a DB-backend. To make it not too easy there are tables with Umlaut Strings in some of the columns. Using HDBC-ODBC the Umlauts get merely garbled, I believe because MS-Access delivers in some sort of UTF-16 and HDBC-ODBC seems to manage that fine safe for the Umlauts, which get garbled. And Takusen crashes outright. Are there other haskellers out there which were in the same situation, ie. MS-Acess and Umlauts (or other none-ascii) and been able to solve this? I sure tried ... Günther

2 1

Non-termination of type-checking
by Matthieu Sozeau 29 Jan '10

29 Jan '10

Dear Haskellers, while trying to encode a paradox recently found in Coq if one has impredicativity and adds injectivity of type constructors [1] I stumbled on an apparent loop in the type checker when using GADTs, Rank2Types and EmptyDataDecls. > {-# OPTIONS -XGADTs -XRank2Types -XEmptyDataDecls #-} > module Impred where The identity type > data ID a = ID a I is from (* -> *) to *, we use a partial application of [ID] here. > data I f where > I1 :: I ID The usual reification of type equality into a term. > data Equ a b where > Eqrefl :: Equ a a The empty type > data False This uses impredicativity: Rdef embeds a (* -> *) -> * object into R x :: *. > data R x where > Rdef :: (forall a. Equ x (I a) -> a x -> False) -> R x > r_eqv1 :: forall p. R (I p) -> p (I p) -> False > r_eqv1 (Rdef f) pip = f Eqrefl pip > r_eqv2 :: forall p. (p (I p) -> False) -> R (I p) > r_eqv2 f = Rdef (\ eq ax -> > case eq of -- Uses injectivity of type constructors > Eqrefl -> f ax) > r_eqv_not_R_1 :: R (I R) -> R (I R) -> False > r_eqv_not_R_1 = r_eqv1 > r_eqv_not_R_2 :: (R (I R) -> False) -> R (I R) > r_eqv_not_R_2 = r_eqv2 > rir :: R (I R) > rir = r_eqv_not_R_2 (\ rir -> r_eqv_not_R_1 rir rir) Type checking seems to loop here with ghc-6.8.3, which is a bit strange given the simplicity of the typing problem. Maybe it triggers a constraint with something above? > -- Loops > -- absurd :: False > -- absurd = r_eqv_not_R_1 rir rir [1] http://thread.gmane.org/gmane.science.mathematics.logic.coq.club/4322/focus… -- Matthieu

4 4