- Beginners - Haskell.org

finding the cause of an error (here head: empty list)
by Nathan Hüsken 31 Mar '13

31 Mar '13

Hey, I have written a program which, when executed produces: Main: Prelude.head: empty list Now, I can go through my program and replace all "head" with (\a -> trace ("<line-nr>: length=" ++ (show (length a))) a) Actually, I already did that, and by this I found the error. But I wonder if there would have been an easier way? Has anyone any debug advice how I could find out which call to "head" causes this without so much typing work? Thanks! Nathan

4 3

cabal install local troubles (osx)
by Miro Karpis 31 Mar '13

31 Mar '13

please, can you help me with following. I have installed haskell on mac from the package. When I check cabal -V I can see 1.14.0 That is because the cabal is loading from /Library/Haskell/bin. After that I run 'cabal install cabal-install', I get following output: Installing executable(s) in /Users/miro/Library/Haskell/ghc-7.4.2/lib/cabal-install-1.16.0.2/bin Updating documentation index /Users/miro/Library/Haskell/doc/index.html So the new version of cabal is installed in another place. When I install new package via cabal, it uses the 1.14.0 version. I have tried to copy the new 1.16.0.2 version to bin, but that just didn't help because the new packages have been installed somewhere else. Please what is the proper way to fix this? many thanks, miro

4 6

Quickcheck2 - writing a modifier for test data
by Nathan Hüsken 31 Mar '13

31 Mar '13

Hey, Using quickcheck2 I want to generate test data, that is a tuple of integers where both elements are >=0 and < 8. So I am trying to write a modifier (in the sense of the modifiers found in Modifiers.hs): newtype ValidPos = ValidPos Pos deriving Show instance Arbitrary ValidPos where arbitrary = do x <- elements [0..7] y <- elements [0..7] return (ValidPos (x,y)) prop_dummy (ValidPos pos) = fst pos > 0 tests = [ testGroup "Tests" [ testProperty "dummy" prop_dummy ] ] main = defaultMain tests doing so gives me this error: No instance for (QuickCheck-2.4.2:Test.QuickCheck.Arbitrary.Arbitrary ValidPos) arising from a use of `testProperty' Possible fix: add an instance declaration for (QuickCheck-2.4.2:Test.QuickCheck.Arbitrary.Arbitrary ValidPos) In the expression: testProperty "dummy" prop_dummy In the second argument of `testGroup', namely `[testProperty "dummy" prop_dummy]' In the expression: testGroup "Tests" [testProperty "dummy" prop_dummy] I declared the instance, why canit not be found? Thanks! Nathan

2 2

point-free + IO Monad
by Ovidiu D 31 Mar '13

31 Mar '13

I have the following code: f1 :: String -> String f2 :: IO String -> IO () f2 a = do b <- a putStr $ f1 b How can I write the function f2 in a point-free style? I have tried this: f2 = return.f1 >>= putStr ...but it doesn't work. Thanks!

3 3

"downcasting"
by Emmanuel Touzery 29 Mar '13

29 Mar '13

Hello, using the Text.Json library (and I'm reading I should be using Aeson instead... I'll migrate to it later), I'm a bit frustrated. At one point I'm getting a JSValue object, which could be anything, including a JSObject. But in this case I know I'm looking at a JSObject. I don't mind if Haskell crashes if that's not the case (for now), either it's a JSObject or the JSON I'm parsing is invalid. I hoped to do that conversion using pattern matching by giving non-exhaustive pattern match, considering only cases where the JSValue is in fact a JSONObject but somehow I can't make it... getObject :: JSValue -> JSObject JSValue getObject x@(JSObject _) = x --> doesn't compile the function returns a JSValue, I somehow wanted to point out to the compiler that if it gets there it's in fact a JSObject. getObject :: JSValue -> JSObject JSValue getObject (JSObject x) = JSObject x -> same as with the other attempt.. the compiler is saying the function is returning a JSValue: Couldn't match expected type `JSObject JSValue' with actual type `JSValue' In the return type of a call of `JSObject' In the expression: JSObject x In an equation for `getObject': getObject (JSObject x) = JSObject x I think I'm missing an obvious pattern in Haskell here, probably something I've already used elsewhere too, but somehow I don't make it. Otherwise answers saying that if I need to make such a "cast" then I'm using that library the wrong way are welcome (together to maybe a link to a tutorial explaining it the right way), but by curiosity I would still be interested in how to actually do that "cast" in this case. Thank you! Emmanuel

7 13

stack overflow summing numbers read from a big file
by Axel Wegen 25 Mar '13

25 Mar '13

When I try to run the following code on a 50M file consisting of one number per line I receive a `Stack space overflow' error and the program seems to consume a lot of memory. Why does that happen? And how can I fix the problem without increasing the Stack with -Ksize hoping that it will be big enough? Any general advice on processing big files with Haskell? -- sumFile.hs -- adapted from Real World Haskell's SumFile.hs at the beginning of -- Chapter 8 main = interact sumFile where sumFile = show . sum . map read . words ghc -o sumFile sumFile.hs ./sumFile < ./numbers

4 7

How to design functions so they can be part of other larger systems?
by Costello, Roger L. 24 Mar '13

24 Mar '13

Hi Folks, I read this statement by Tim Berners-Lee [1]: It is not only necessary to make sure your own system is designed to be made of modular parts. It is also necessary to realize that your own system, no matter how big and wonderful it seems now, should always be designed to be a part of another larger system. Recently I have been working hard to learn how to better modularize. But now TBL says that I must do more - I must not only modularize well, but I must also build the modules so that they can be part of other larger systems. How do I design modules so that they may be part of other larger systems? Are there any articles that give guidelines on how to do this? What are your thoughts on how to do this? /Roger [1] http://www.w3.org/DesignIssues/Principles.html

4 3

hiding members of a data, separate accessors instead
by Emmanuel Touzery 24 Mar '13

24 Mar '13

hi, i was looking at the response type in http-streams: http://hackage.haskell.org/packages/archive/http-streams/0.4.0.0/doc/html/N… I'm used that simply the data type and all its "members" are visible -- the functions to access its contents. But in this case on the HTML documentation the response type looks like it has no members. And the author has defined like "public accessors" later in the code: getStatusCode :: Response -> StatusCode getStatusCode = pStatusCode So I'm not even sure how he achieved that the members are not visible, the data are exported with (..) as is usually done... And the other thing is why would you do that.. You could name the member getStatusCode in the first place, but then it might increase encapsulation to hide it (depending on how he managed to hide the members).. But did you then make it impossible to deconstruct a Response through pattern matching? That sounds like a minus... Although pattern matching on a data with 6 fields is always going to be a pain and decreasing the chances for modifying the data type without breaking compatibility. These "members" are also causing me problems in other situations, for instance I have some cases when I use a data type only a few times and with -Wall the compiler tells me I don't use the accessor; in fact I read that value from the data, but through pattern matching/deconstruction only, not through that particular function. I'm thinking to try to hide the warning as I think my code is correct. Anyway I'm curious on the mechanism used by that library... I've already noticed a few nice tricks in this library, like a small state monad to take optional parameters, much more elegant than any other mechanism i've seen so far to achieve the same effect. Thank you! Emmanuel

3 5

binary parsing problem
by Alexander Polakov 23 Mar '13

23 Mar '13

I'm trying to parse a binary format. One of the components is defined like this: "BITDOUBLE: 1st 2 bits : what it is 00 : A double follows 01 : 1.0 10 : 0.0 11 : not used Doubles are eight byte IEEE standard floating point values." So I wrote this code which looks almost like verbatim translation of the spec. import Data.Binary import Data.Binary.Get import Control.Applicative import qualified Data.Binary.Bits.Get as Bits newtype DWG_BD = DWG_BD Double deriving (Show) instance Binary DWG_BD where put = undefined get = do d <- Bits.runBitGet $ Bits.getWord8 2 >>= \i -> case i of 0 -> decode <$> Bits.getLazyByteString 8 1 -> return 1.0 2 -> return 0.0 _ -> fail "bad DWG_BD" return (DWG_BD d) But then when I try to parse anything starting with 00 bits, I get this: *Main> decodeFile "/tmp/foo" :: IO DWG_BD DWG_BD *** Exception: Data.Binary.Get.runGet at position 2: demandInput: not enough bytes I'm 100% sure there're enough bytes in the file. I recall seeing this problem when my record's fields were not strict, but this is not the case here. I suppose the problem lies in getLazyByteString, but have no idea how to solve it. -- Alexander Polakov | plhk.ru

1 0

Programming with Arrows
by Michael Baker 23 Mar '13

23 Mar '13

I'm reading through http://www.cse.chalmers.se/~rjmh/afp-arrows.pdf I was following it up until this example in the section about flip-flops: class Arrow arr => ArrowLoop arr where loop :: arr (a,c) (b,c) -> arr a b instance ArrowLoop (->) where loop f a = b where (b,c) = f (a,c) Although I've never seen a function declaration like this, but I think I get it. Because f :: (a, c) -> (b, c) then then loop f :: ((a, c) -> (b, c)) -> (a -> b) which is the same as ((a, c) -> (b, c)) -> a -> b. However, I don't see where the c comes from in f (a,c). Is this a mistake or am I missing something? A friend of mine realized that this is just a recursive definition so f (a, c) == f (a, snd $ f (a, snd $ f (a, ...))). I don't really understand this definition. I can see how it compiles, but I don't see how it would ever produce a legitimate value. Do I have to assume that f never evaluates the second element in the pair and just passes it through?

2 1