- Beginners - Haskell.org

Closing a channel
by Lyndon Maydwell 13 May '11

13 May '11

Hi Beginners. I'm trying to implement a program that accumulates a list of items through many IO actions, then use these in another part of the program as a list. I've found Control.Concurrent.Chan which seems to provide a nice abstraction with getChanContents :: Chan a -> IO [a], but there doesn't seem to be a way for the source to close the channel. Are channels the right abstraction for something like this? Should look in to (itter|enumer)at(or|ee)s instead... They seem to encompass this model, but look quite involved.

3 3

Compiling with HDBC
by Christopher Howard 13 May '11

13 May '11

Hi. I'm very new to Haskell, but I'm trying to compile a simple program that uses HDBC with ODBC. I am working on an amd64 Gentoo Linux system. I've installed hdbc and hdbc-odbc with the unixOBDC bindings. So far, I created a small hello world type program. It doesn't use any HDBC functionality yet, but I inserted the following two lines at the top to see if I could compile correctly: import Database.HDBC import Database.HDBC.ODBC However, when I try to compile I get this: $ ghc Main.hs Main.o: In function `s1cY_info': (.text+0x1ef): undefined reference to `__stginit_HDBCzmodbczm1zi1zi3zi0_DatabaseziHDBCziODBC_' Main.o: In function `s1cY_info': (.text+0x1fb): undefined reference to `__stginit_HDBCzm2zi2zi6zi1_DatabaseziHDBC_' collect2: ld returned 1 exit status Though, these errors do not show up when I run the .hs file with runhaskell. I'm guessing I just don't know how to link to the correct library or libraries. I tried adding -lodbc to the ghc line but that did not help. In /usr/lib I see libodbc.so, as well as the directories hdbc-1.1.3 and hdbc-odbc-1.1.3.0. I would appreciate any guidance. -- frigidcode.com theologia.indicium.us

4 4

Deserializing data with ProtocolBuffers
by Raphael Päbst 12 May '11

12 May '11

Hey! I am currently working on a project, where I need to serialize data with ProtocolBuffers. I have managed to compile a .proto file and serialize messages with messagePut. However, when I try to deserialize a ByteString again, I get the following error: ambiguous type variable `msg0' in the constraints: (Text.ProtocolBuffers.WireMessage.Wire msg0) arising from a use of "messageGet" at... (Text.ProtocolBuffers.Reflections.ReflectDescriptor msg0) arising from a use of "messageGet" at ... Probable fix: add a type signature that fixes these type variables in the expression "messageGet bs" in an equation for `IT': "IT = messageGet bs" I'm trying to call messageGet from the prompt in ghc. What exactly am I doing wrong? And how can I fix this? Many Thanks Raphael

2 3

Are there summary sheets of operations on lists etc. i.e. shorter than Data.List
by KC 12 May '11

12 May '11

-- -- Regards, KC

1 0

Problem with Data.Map and IO...
by Manfred Lotz 12 May '11

12 May '11

Hi there, I've got a problem creating a map. I have a list of dataset names and adding those dataset names into a map works fine like this: <----------------------------snip--------------------------------> module Main where import Data.List import qualified Data.Map as M insertRec :: M.Map String String -> String -> M.Map String String insertRec m e = let f = M.lookup e m in if f == Nothing then M.insert e "bla" m else m createMap :: [String] -> M.Map String String createMap lst = let m = M.empty in foldl' insertRec m lst main :: IO () main = do let xmlfiles = [ "a.xml","f.xml"] let ht = createMap xmlfiles print ht <----------------------------snap--------------------------------> However, actually I want to parse the xml files and adding content into the map. For this I have a function getXmlContent :: FilePath -> IO (String,String) and would like to do something like this: <----------------------------snip--------------------------------> insertRec m e = do c <- getXmlContent e let k = fst c let v = snd c M.insert k v m <----------------------------snap--------------------------------> Unfortunately, this doesn't work because of IO. I have no glue how to get this done. Any help appreciated. -- Thanks, Manfred

3 3

List Comprehensions
by Hamster 11 May '11

11 May '11

Hello, I am trying to understand list comprehensions (I am very new to Haskell) and there's one thing I simply don't understand. If I have code that looks like: combinations = [ (x,y) | x <-[1,2,3], y<-[1,2,3]] Then it happily gives me all the possible combinations of x and y. [(1,1),(1,2),(1,3),(2,1),(2,2),(2,3),(3,1),(3,2),(3,3)] However... If I change the function to look like: combinations xs ys = [ (x,y) | x <-[xs], y<-[ys]] The output changes to: [([1,2,3],[1,2,3])] I don't understand what's going on there. Is there any way I can get it to produce all the combinations of a user supplied list? Thanks H.

3 2

Complexity of Arrays in Haskell
by Øystein Kolsrud 10 May '11

10 May '11

Hi! What's the complexity of indexing arrays in Haskell? The Data.Array documentation doesn't seem to mention complexity of any of its functions, and it's not obvious to me how to implement an array data structure with indexing being O(1). -- Mvh Øystein Kolsrud

3 3

Is there such a thing as a lazy monadic list?
by Arlen Cuss 10 May '11

10 May '11

Hi all, I'm currently trying to write a very simple VM in order to learn and stretch my Haskell. I'm currently up against a problem. I'm using an (IOUArray Int Word32) to represent the memory of a VM instance. This is all fine so far, but now I'm trying to write a 'disassemble' function that will disassemble an instruction starting from a given location in memory. Ideally I could write a function that would lazily yield successive elements from memory, sort of like this: > readVMFrom :: VM -> Address -> IO [Word32] > readVMFrom v f = do > i <- readArray (memory v) f > rest <- readVMFrom v (succ f) > return $ i : readVMFrom v (succ f) Of course, this isn't lazy, and so happily runs off the end of memory each time. (I could stop that, but that's not the point!) I'm not sure how to lazily construct a list in IO. My next thought is to instead make a function to return an action that yields successive values of memory, given some start, i.e. something like: > readVMFrom :: VM -> Address -> IO Word32 .. but I'm thinking I might have to do some StateT nonsense in order to actually keep track of which index we want to read next. (subtext: I don't know how to do that!) So I then think that perhaps I can do this by having readVMFrom return an action which yields a) the next value, b) the address of the next value (so we know how much we ate), and c) an action to yield the next value and action and so on, so on. I managed that this way (WARNING, BEGINNER HASKELL CODE FOLLOWS): > newtype RVFResult = RVFResult { getRVFResult :: (Word32, Address, IO RVFResult) } > > readVMFrom :: VM -> Address -> IO RVFResult > readVMFrom v f = do > i <- readArray (memory v) f > return $ RVFResult (i, succ f, readVMFrom v $ succ f) A function can then take an argument of type RVFResult as an indication it wants 1..n sequential values from memory: > disassemble :: RVFResult -> IO (Instruction, Address) > disassemble r = do > let (i, p, a) = getRVFResult r > case i of > 0x01 -> return (Push, p) > 0x02 -> return (Pop, p) > 0x03 -> return (Call, p) > 0x04 -> do > r <- a > let (v, p, a) = getRVFResult r > return (Out v, p) Here you can see how we consume an additional value by executing the returned action 'a'. It's damn ugly, and this almost looks like the place for a monad (possibly something involving StateT, though, I don't know!). Ideally I'd manage a lazily constructed list -- somehow I feel that's most "Haskelly" -- but does anyone else have any ideas? I kinda fleshed this last solution out as I wrote this email, so it's kinda cruddy. Cheers, Arlen

3 6

help understanding JSON
by Sean Perry 10 May '11

10 May '11

I have some JSON which looks like this: {"count":2, "result":[{"LastTrade":"31.24", "Symbol":"FOO"}, {"LastTrade":"345.12", "Symbol":"BAR"}]} (named testQuoteResult below) When I try to parse this into types defined below it failed because the Double is encoded as a String. If I change the type of LastTrade in my code to String everything is ok but I am left read'ing the String. I have defined the following using Language.JsonGrammar: import Data.Iso import Language.JsonGrammar import Prelude hiding (id, (.), head, either) import Control.Category ((.)) import qualified Data.Aeson as JS import qualified Data.ByteString.Char8 as BS import qualified Data.Attoparsec as P data MyQuote = MyQuote { price :: Double, symbol :: String } deriving(Show) myQuote = $(deriveIsos ''MyQuote) instance Json MyQuote where grammar = myQuote . object ( prop "LastTrade" . prop "Symbol" ) data MyQuoteResult = MyQuoteResult { count :: Int, result :: [MyQuote] } deriving(Show) myQuoteResult = $(deriveIsos ''MyQuoteResult) instance Json MyQuoteResult where grammar = myQuoteResult . object ( prop "count" . prop "result" ) tryJSON input = case P.parse JS.json (BS.pack input) of P.Done s v -> Right v other -> Left ("failed parse " ++ show other) Using it like this: > let Right qr = tryJSON testQuoteResult > fromJson qr :: Maybe MyQuoteResult Nothing Is there some way to convince the JSON text parser to turn "12.34" into (Double 12.34)?? An example using pure aeson would be fine if required. I just decided to play with JsonGrammar since it was announced today and it sounded interesting.

4 4

Typeclasses and generic types
by Christopher Howard 10 May '11

10 May '11

I'm learning how to use typeclasses for the first time, and I tried something like so: [CODE] data Point a = Point a a getX (Point a _) = a getY (Point _ a) = a instance Eq Point where (==) a b = getX a == getX b && getY a == getY b [/CODE] But when I try to load this in GHCI I get: [CODE] `Point' is not applied to enough type arguments The first argument of `Eq' should have kind `*', but `Point' has kind `* -> *' In the instance declaration for `Eq Point' [/CODE] This seems to be saying that Eq requires a concrete argument, which I suppose I can understand. I can fix this by changing "data Point a = Point a a" into "data Point = Point Int Int" or something like that, but that seems kind of lame: what if I want to leave my Point type more generic? (Maybe sometimes I'll need a Point Int, and other times I'll need a Point Float, or some other weird combination I haven't thought of yet.) Instead of changing the "data" declaration, can I change the "instance" declaration somehow so that it just defines an instance for a particular Point sub-type? I tried some variations like "instance Eq (Point Float)" but I haven't been able to come up with anything yet that satisfies the compiler. -- frigidcode.com theologia.indicium.us

3 3