May 2006 - Haskell-Cafe

Existentially-quantified constructors: Hugs is fine, GHC is not?
by Otakar Smrz 15 May '06

15 May '06

Dear all, while WinHugs (20051031) lets me match against an existentially quantified constructor data ... = ... | forall b . FMap (b -> a) (Mapper s b) ... where FMap qf qc = stripFMap f q the GHC compiler as well as GHCi (6.4.2 and earlier) issue an error My brain just exploded. I can't handle pattern bindings for existentially-quantified constructors. Let me give the whole (non-practical) code, which is well typed and compiles in Hugs, and then show the change I had to do to make it work in GHC, too. The question is why there is a difference. Am I misusing something? The point of the complexFun below is to explore the Mapper data structure, taking care of the :&: constructor and quickly (transitively) skipping the FMap constructors, only accumulating and composing the tranformation functions that these provide. ----------------------------------------------------------- module Problem where import Data.Map as Map hiding (map) type Labels a = [a] data Mapper s a = Labels a :&: Map.Map s (Mapper s a) | forall b . FMap (b -> a) (Mapper s b) stripFMap :: Ord s => (a -> c) -> Mapper s a -> Mapper s c stripFMap k (FMap f p) = stripFMap (k . f) p stripFMap k x = FMap k x complexFun :: Ord s => (b -> a) -> Mapper s b -> s -> [a] complexFun f c y = case c of FMap t q -> complexFun qf qc y where FMap qf qc = stripFMap (f . t) q -- !!! r :&: k -> case Map.lookup y k of Just q -> let FMap qf qc = stripFMap f q -- !!! in case qc of ([] :&: _) -> complexFun qf qc y (xs :&: _) -> map qf xs _ -> error "Never matching" Nothing -> error "Irrelevant" ----------------------------------------------------------- Even though GHC does not let me pattern-match on FMap, I can use a case expression in complexFun instead -- then it compiles alright: ------ complexFun f c y = case c of FMap t q -> case stripFMap (f . t) q of -- !!! FMap qf qc -> complexFun qf qc y -- !!! _ -> error "No option" -- !!! r :&: k -> case Map.lookup y k of Just q -> case stripFMap f q of -- !!! FMap qf qc -> case qc of -- !!! ([] :&: _) -> complexFun qf qc y (xs :&: _) -> map qf xs _ -> error "Never matching" _ -> error "No option" -- !!! Nothing -> error "Irrelevant" ------ If I wanted to make this auxiliary case on stripFMap local, there would be a problem for both Hugs and GHC: Hugs: Existentially quantified variable in inferred type! *** Variable : _48 *** From pattern : FMap xf xc *** Result type : (_48 -> _32,Mapper _30 _48) GHC: Inferred type is less polymorphic than expected Quantified type variable `b' is mentioned in the environment: qc :: Mapper s1 b (bound at Problem.hs:65:27) qf :: b -> a1 (bound at Problem.hs:65:23) When checking an existential match that binds xf :: b -> a xc :: Mapper s b The pattern(s) have type(s): Mapper s1 a1 The body has type: (b -> a1, Mapper s1 b) In a case alternative: FMap xf xc -> (xf, xc) ------ complexFun f c y = case c of FMap t q -> complexFun qf qc y where (qf, qc) = case stripFMap (f . t) q of -- !!! FMap xf xc -> (xf, xc) _ -> error "No option" r :&: k -> case Map.lookup y k of Just q -> let (qf, qc) = case stripFMap f q of -- !!! FMap xf xc -> (xf, xc) _ -> error "No option" in case qc of ([] :&: _) -> complexFun qf qc y (xs :&: _) -> map qf xs _ -> error "Never matching" Nothing -> error "Irrelevant" ------ Many thanks for your comments or advice! Best, Otakar Smrz

4 3

CDouble type coercion
by SevenThunders 14 May '06

14 May '06

I am new to Haskell and found myself in a bind concerning the use of the C types, CDouble in particular. I extract a CDouble via it's pointer from a StorableArray. Since the array must interface with C the elements of the array must be CDouble. Now I'd like to use Text.Printf to format print statements of elements of the array, but Text.Printf requires Doubles as inputs and so far I have not found an obvious way to coerce CDoubles into Doubles. Additionally, withStorableArray returns a monad containing a CDouble instead of Double as peek I suppose does. Because I cant coerce my CDouble into a Double, printf chokes. Thus the following code fails to compile on GHC import Foreign import Foreign.C import Foreign.C.Types (CInt, CDouble ) import Data.Array.Storable import Text.Printf foreign import ccall "matrix_c.h sumarr" sumarr :: Ptr CDouble -> IO (CDouble) main = do arr <- newListArray (1 , 3) [3,2,1]:: IO (StorableArray Int CDouble) -- extract the pointer to arr withStorableArray arr sumarr >>= (\x -> printf "15.7f\n" x) The error message is, test2.hs:13:44: No instance for (PrintfArg CDouble) arising from use of `printf' at test2.hs:13:44-49 Probable fix: add an instance declaration for (PrintfArg CDouble) In a lambda abstraction: \ x -> printf "15.7f\n" x In the second argument of `(>>=)', namely `(\ x -> printf "15.7f\n" x)' In the result of a 'do' expression: (withStorableArray arr sumarr) >>= (\ x -> printf "15.7f\n" x) -- View this message in context: http://www.nabble.com/CDouble-type-coercion-t1615450.html#a4378492 Sent from the Haskell - Haskell-Cafe forum at Nabble.com.

4 5

[newbie] processing large logs
by Eugene Crosser 14 May '06

14 May '06

Hello gentlemen, I am exposed to functional programming for less than a month, and just trying to understand the concepts, so please bear with me. I tried to use Haskell for a simple task on my dayjob, that involved parsing mail system logs and counting distinct addresses (I work for a relatively big ISP). Reducing unnecessary details, let's say that we have a file of N lines, some of them repeating, so that there are only M distinct lines, where M << N. The task is to count the number of times each distinct line appears in the file, and print the most frequent one with its count. This is my program: ======== module Main where import Data.Map main = printMax . (foldr processLine empty) . lines =<< getContents processLine line map = insertWith (\new old -> new + old) line 1 map printMax map = putStrLn $ show $ foldWithKey (\key val accum -> if val > (snd accum) then (key,val) else accum) ("",0) map ======== The thing kinda works on small data sets, but if you feed it with 250,000 lines (1000 distinct), the process size grows to 200 Mb, and on 500,000 lines I get "*** Exception: stack overflow" (using runhaskell from ghc 6.2.4). (For comparison, a perl script does the same job (using global hash) an order of magnitude faster, consumes 3 Mb RAM and can process billion lines without a problem). The question is: what I am doing wrong? Thanks Eugene crosser@pccross:~/src$ perl genlist.pl 500000 1000|time runhaskell distinct.hs *** Exception: stack overflow Command exited with non-zero status 1 27.46user 0.49system 0:29.75elapsed 93%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+34676minor)pagefaults 0swaps crosser@pccross:~/src$ perl genlist.pl 250000 1000|time runhaskell distinct.hs ("a0000000531",300) 36.66user 0.72system 0:54.82elapsed 68%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (522major+50361minor)pagefaults 0swaps crosser@pccross:~/src$ perl genlist.pl 250000 1000|time perl distinct.pl a0000000531: 300 0.22user 0.00system 0:01.14elapsed 20%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (15major+472minor)pagefaults 0swaps

5 10

foldl to foldl' , ok ; but with foldM ?
by minh thu 14 May '06

14 May '06

Hi, I know the question is raised often but i cannot find the answer. In a previous mail, I was said to use foldl' instead of foldl and it worked really well : acc1 values = foldl1' (+) values How is foldl' defined ? I have two other related questions : What if I want to write something like this : foldl' f emptyArray values where f accumulates a value into an array. What's the best representation for an array here ? Also, given the fact that I though to represent my array with a c-like Ptr Word8 array, i tried this code, where the accumulator is referenced by an IORef : acc2 ints = do accRef <- newIORef 0 foldM add accRef ints acc <- readIORef accRef putStrLn $ show acc return acc where add accRef i = do modifyIORef accRef (+ i) return accRef but it's far worse than the code with foldl'. I guess someone will tell me to try to strictify something... but I don't know what and how. I think trying to strictify the accRef is worthless : it's the referenced acc that needs to be evaluated. Is-it correct ? Thanks !! VO Minh Thu

1 0

develop new Haskell shell?
by Marc Weber 13 May '06

13 May '06

Hi. Who wants to try devloping a new shell with me? The main goals: try adding some haskell scriting instead of bash/zsh, history dependend on a) executing program b) current dir c) last commands d) workspaces which should mean that the shell should save at least the last 10 commands of a,b,c,d. So you can do emerge (lookup parameters in history) even if you haven't used emerge for ages.. :) Nice, isn't it? d) Workspaces should mean: You can define some kind of workspace like workspace=haskellproject, wash, apache to add these tags together with the commands to the history.. So when working only in the "wash" workspace you can easily find those commands.. Perhpas it's even useful to attach commands or even scripts to those workspaces? eg the startApache script may be attached to admin, apache, ..., the cd /etc/init.d command only to admin.. I also would like to have some advanced kind of directory matching, defining aliases for directories. eg just type cd /usl to get a list of diretories looking like this: /UserShupportLocales /usr/src/linux /usl ? Using tab and bash is nice but it might be done better? Any suggestions? One would have to think about how to run processes in background and so on ... adding files as parameters the way it's possible in mc ( select them and add them to the command line ) perhaps even implement cp/mv/ ... for virtual file systems like zip files/ ftp/ ... ? and last but not least: on windows add all Programs beeing found in Start-> Programs to the path list... I wish I could just do word/ Enterprise Manager at a shell and not searching for the menu entries over and over again.. ;) I know I can add them the to the path.. but that would be some work, too.. and not desirable in any case. I could imagine adding a small prefix to each cmd eg. eb (execute bash cmd) ez (execute zsh cmd) r (remove file list) efs (execute from windows start menu) bg <cmd> run in background like bashs & feature. Perhaps even introduce some new syntax ? or use ghci or hugs with a preprocessor to translate these commands to haskell commands? What do you think? Marc Weber

2 3

RE: transactional cache
by Alberto G. Corona 13 May '06

13 May '06

Albert Lai: Thanks. I got the poit more or less; Each invocation creates a new IORef instance. UnsafePerformIO appears to generate a unique IORef that can be shared (sorry for my imperative vocabulary, I´m sill contaminated by al these evil languages ;). I tried with "usafePerformIO NewTVar v" but the program fails miserably in a memory fault. I finally did it well usign a IORef than contains the TVar: refcache =unsafePerformIO $ (do c <- atomically $ newTVar emptyFM newIORef c) and then dereferencing refcache in the IO Monad I get ever the same context no matter where i do it: do tvcache <- readIORef refcache atomically $ do finiteMap <- readTVar tvcache (useful code here at last)... --------------------- Who said that Haskell is difficult?. Jokes apart, STM is powerful. I will share the transactional cache when I have it tested. -----------referred message:------- Message: 1 Date: 12 May 2006 00:19:28 -0400 From: Albert Lai <trebla(a)vex.net> Subject: Re: [Haskell-cafe] Re: Haskell-Cafe Digest, Vol 33, Issue 9 To: haskell-cafe(a)haskell.org Message-ID: <4u3bffq3hr.fsf(a)shell.vex.net> Content-Type: text/plain; charset=us-ascii "Alberto G. Corona " <agocorona(a)gmail.com> writes: > stmcache= newTVar 0 I will explain what this doesn't with an analogy. import Data.IORef notglobal = newIORef True main = do a <- notglobal b <- notglobal writeIORef a False x <- readIORef b print x To better show what's going on, I also provide this for contrast: import Data.IORef import System.IO.Unsafe global = unsafePerformIO (newIORef True) main = do x <- readIORef global print x writeIORef global False x <- readIORef global print x

2 1

(no subject)
by rami makram 12 May '06

12 May '06

Hi, Thx for replying and answering my question.. But, The idea is that i have to implement it using this kind of type generateTruthTable->[[(Char,Bool)]] and we didn't take the Shell and libraries in common, so i have to implement in a simpler way.. Can you give me another way or even some hints.. I have already implemented the And,Or,Impl,Neg,and other functions.. Thx.. _________________________________________________________________ Don't just search. Find. Check out the new MSN Search! http://search.msn.click-url.com/go/onm00200636ave/direct/01/

1 0

Haskell + Windows API = wuh?
by ihope 12 May '06

12 May '06

Ello. I'm looking for a way to interface the Windows API with Haskell in order to write a Win32 console handler thingy. I don't know anything about the Haskell Foreign Function Interface beyond what its name implies and that it's an extension to Haskell 98 (or is it?). So just where do I start? --ihope

3 2

Re: Haskell-Cafe Digest, Vol 33, Issue 9
by Alberto G. Corona 12 May '06

12 May '06

Hi, I´m trying to make a searchable transactional cache using STM. The whole idea is to use indexed TVar variables using a FiniteMap. another TVar holds the finitemap . This last TVar has to be a global variable. I found that when handled as global, a TVar does not keep the state. For example: stmcache= newTVar 0 main= do v<- val v<- val print v val = atomically $! do cache <- stmcache x <- readTVar cache writeTVar cache (x+1) return x this code print 0 instead of 1 as the result of incrementing cache and retrieving it the second time The same happens with: iocache= atomically $ newTVar 0 `debug` "newTVar" main= do v<- val v<- val print v getChar val = do cache<- iocache atomically $! do x <- readTVar cache writeTVar cache (x+1) return x here i used a IO TVar global instead of a STM TVar This other version , that passes TVar as parameter, does work: iocache= atomically $ newTVar 0 `debug` "newTVar" main= do cache<-iocache v<- val cache v<- val cache print v val cache= atomically $! do x <- readTVar cache writeTVar cache (x+1) I need the global version since the cache acces may be in any location on the program.... Anyone know a solution using STM? The MVar version works OK, but blocking the entire index is not the best solution....

2 1

RE: Template Haskell Person/Paper entry
by Simon Peyton-Jones 11 May '06

11 May '06

Can anyone help Christoph with uploading a Postscript file to a haskellwiki page? Thanks | -----Original Message----- | From: Ch. A. Herrmann [mailto:herrmann@uni-passau.de] | Sent: 11 May 2006 14:29 | To: Ian Lynagh | Cc: Simon Peyton-Jones | Subject: Re: Template Haskell Person/Paper entry | | Hi, | | I was able to edit | | http://haskell.org/haskellwiki/Template_Haskell | | and add a paragraph to it, but there are still two problems. | | (1) When I try to upload a Postscript file (597453 bytes) I get | the error message "file contains no data" |

1 0