- Beginners - Haskell.org

Profiling introduces a space leak where there was none before?
by Travis Erdman 13 Aug '10

13 Aug '10

In Ch 25 of Real World Haskell, the authors introduce some naive code for finding the average of a big list; it has a space leak, and they present several solutions. Below are two of the solutions that successfully eliminate the space leak (though, the first one -- the one that uses foldl'rnf -- is quite a bit faster). However, if compiled with profiling, the first one (using foldl'rnf) NOW has a leak. The second solution (foldl') does not have a leak even when profiling is enabled. I have used this foldl'rnf function in my own code, as it is the only solution I have found for a space leak in my own code. But, since it leaks when profiled, it is making analysis difficult. Is this a feature, bug, or user error? If a known issue, is there a workaround? The code and some documenting output follows. thanks, Travis ------------------------------------ {-# LANGUAGE BangPatterns #-} import System.Environment import Text.Printf import Control.Parallel.Strategies import Control.DeepSeq import Data.List (foldl') main = do [d] <- map read `fmap` getArgs printf "%f\n" (mean [1..d]) foldl'rnf :: NFData a => (a -> b -> a) -> a -> [b] -> a foldl'rnf f z xs = lgo z xs where lgo z [] = z lgo z (x:xs) = lgo z' xs where z' = f z x `using` rdeepseq -- first mean fn aka foldl'rnf mean :: [Double] -> Double mean xs = s / fromIntegral n where (n, s) = foldl'rnf k (0, 0) xs k (n, s) x = (n+1, s+x) :: (Int, Double) -- second mean fn aka foldl' -- mean :: [Double] -> Double -- mean xs = s / fromIntegral n -- where -- (n, s) = foldl' k (0, 0) xs -- k (!n, !s) x = (n+1, s+x) ------------------------------------------ [NO PROFILING, NO SPACE LEAK] C:\Documents and Settings\Travis\My Documents\Haskell Code>ghc --make temp5 -O2 -fasm [1 of 1] Compiling Main ( temp5.hs, temp5.o ) Linking temp5.exe ... C:\Documents and Settings\Travis\My Documents\Haskell Code>temp5 1e7 +RTS -sstderr temp5 1e7 +RTS -sstderr 5000000.5 1,170,230,652 bytes allocated in the heap 128,876 bytes copied during GC 3,372 bytes maximum residency (1 sample(s)) 13,012 bytes maximum slop 1 MB total memory in use (0 MB lost due to fragmentation) Generation 0: 2232 collections, 0 parallel, 0.05s, 0.05s elapsed Generation 1: 1 collections, 0 parallel, 0.00s, 0.00s elapsed INIT time 0.02s ( 0.03s elapsed) MUT time 1.52s ( 1.55s elapsed) GC time 0.05s ( 0.05s elapsed) EXIT time 0.00s ( 0.00s elapsed) Total time 1.58s ( 1.63s elapsed) %GC time 3.0% (2.9% elapsed) Alloc rate 764,232,262 bytes per MUT second Productivity 96.0% of total user, 93.3% of total elapsed C:\Documents and Settings\Travis\My Documents\Haskell Code>temp5 1e8 +RTS -sstderr temp5 1e8 +RTS -sstderr 50000000.5 11,702,079,228 bytes allocated in the heap 1,253,872 bytes copied during GC 3,372 bytes maximum residency (1 sample(s)) 13,012 bytes maximum slop 1 MB total memory in use (0 MB lost due to fragmentation) Generation 0: 22321 collections, 0 parallel, 0.38s, 0.39s elapsed Generation 1: 1 collections, 0 parallel, 0.00s, 0.00s elapsed INIT time 0.02s ( 0.00s elapsed) MUT time 15.47s ( 15.72s elapsed) GC time 0.38s ( 0.39s elapsed) EXIT time 0.00s ( 0.00s elapsed) Total time 15.86s ( 16.11s elapsed) %GC time 2.4% (2.4% elapsed) Alloc rate 755,734,682 bytes per MUT second Productivity 97.5% of total user, 96.0% of total elapsed [NOW TURN ON PROFILING, GET SPACE LEAK] C:\Documents and Settings\Travis\My Documents\Haskell Code>ghc --make temp5 -O2 -fasm -prof -auto-all [1 of 1] Compiling Main ( temp5.hs, temp5.o ) Linking temp5.exe ... C:\Documents and Settings\Travis\My Documents\Haskell Code>temp5 1e6 +RTS -sstderr -p -K128M temp5 1e6 +RTS -sstderr -p -K128M 500000.5 395,774,976 bytes allocated in the heap 238,684,620 bytes copied during GC 102,906,760 bytes maximum residency (7 sample(s)) 66,283,900 bytes maximum slop 179 MB total memory in use (4 MB lost due to fragmentation) Generation 0: 493 collections, 0 parallel, 4.83s, 4.84s elapsed Generation 1: 7 collections, 0 parallel, 0.23s, 0.30s elapsed INIT time 0.02s ( 0.03s elapsed) MUT time 0.81s ( 0.91s elapsed) GC time 5.06s ( 5.14s elapsed) RP time 0.00s ( 0.00s elapsed) PROF time 0.00s ( 0.00s elapsed) EXIT time 0.00s ( 0.00s elapsed) Total time 5.89s ( 6.08s elapsed) %GC time 85.9% (84.6% elapsed) Alloc rate 477,916,952 bytes per MUT second Productivity 13.8% of total user, 13.4% of total elapsed

3 2

ghci access to .hs functions
by prad 12 Aug '10

12 Aug '10

in an effort to figure out type declarations i thought i'd let ghci do the work. so i wrote a program: import ... main = do ... func1 func2 now when i load this into ghci i can't to :t func1 etc and get a not in scope error however, if i put a return () onto the end of main: import ... main = do ... return () func1 func2 the functions are in scope of ghci and i can find out the types. so what is happening here? i understand that return is different in haskell than in other languages, but i don't understand just what it is doing. :( -- In friendship, prad ... with you on your journey Towards Freedom http://www.towardsfreedom.com (website) Information, Inspiration, Imagination - truly a site for soaring I's

4 7

FastCGI error
by Wayne R 12 Aug '10

12 Aug '10

I'm trying to use Network.FastCGI with Apache2, but running into an error. I'd like to believe that the web server is configured correctly, because I can successfully view the following, compiled to <web root>/fastcgi/test.fcgi: #include <fcgi_stdio.h> int main(void){ int count = 0; while(FCGI_Accept()>=0){ printf("Content-Type: text/html\r\n"); printf("\r\n"); printf("Hello, World: %d", count++); } return 0; } Now I compile Fcgi.hs to <web root>/fastcgi/test2.fcgi: import Control.Concurrent import Network.FastCGI action :: CGI CGIResult action = do setHeader "Content-type" "text/plain" tid <- liftIO myThreadId output $ unlines [ "I am a FastCGI process!" , "Hear me roar!" , "" , show tid ] main = runFastCGIConcurrent' forkIO 10 action When I view localhost/fastcgi/test.fcgi, I get the hello world page, but attempting to view test2.fcgi shows an 'Internal Server Error" in firefox. The apache error log says: [Sun Aug 08 07:44:17 2010] [notice] Apache/2.2.14 (Ubuntu) mod_fastcgi/2.4.6 mod_lisp2/1.3.1 PHP/5.3.2-1ubuntu4.2 with Suhosin-Patch mod_ssl/2.2.14 OpenSSL/0.9.8k configured -- resuming normal operations [Mon Aug 09 11:11:23 2010] [warn] FastCGI: (dynamic) server "/var/www/fastcgi/test.fcgi" started (pid 20354) [Mon Aug 09 11:11:27 2010] [warn] FastCGI: (dynamic) server "/var/www/fastcgi/test2.fcgi" started (pid 20360) [Mon Aug 09 11:11:31 2010] [warn] FastCGI: (dynamic) server "/var/www/fastcgi/test2.fcgi" started (pid 20365) [Mon Aug 09 11:11:34 2010] [warn] FastCGI: (dynamic) server "/var/www/fastcgi/test2.fcgi" started (pid 20371) [Mon Aug 09 11:11:37 2010] [warn] FastCGI: (dynamic) server "/var/www/fastcgi/test2.fcgi" started (pid 20375) [Mon Aug 09 11:11:40 2010] [warn] FastCGI: (dynamic) server "/var/www/fastcgi/test2.fcgi" started (pid 20379) [Mon Aug 09 11:11:43 2010] [warn] FastCGI: (dynamic) server "/var/www/fastcgi/test2.fcgi" started (pid 20383) [Mon Aug 09 11:11:46 2010] [warn] FastCGI: (dynamic) server "/var/www/fastcgi/test2.fcgi" started (pid 20387) [Mon Aug 09 11:11:49 2010] [warn] FastCGI: (dynamic) server "/var/www/fastcgi/test2.fcgi" started (pid 20391) [Mon Aug 09 11:11:52 2010] [warn] FastCGI: (dynamic) server "/var/www/fastcgi/test2.fcgi" started (pid 20395) [Mon Aug 09 11:11:55 2010] [warn] FastCGI: scheduled the start of the last (dynamic) server "/var/www/fastcgi/test2.fcgi" process: reached dynamicMaxClassProcs (10) [Mon Aug 09 11:11:55 2010] [warn] FastCGI: (dynamic) server "/var/www/fastcgi/test2.fcgi" started (pid 20399) [Mon Aug 09 11:12:01 2010] [error] [client 127.0.0.1] FastCGI: comm with (dynamic) server "/var/www/fastcgi/test2.fcgi" aborted: (first read) idle timeout (30 sec) [Mon Aug 09 11:12:01 2010] [error] [client 127.0.0.1] FastCGI: incomplete headers (0 bytes) received from server "/var/www/fastcgi/test2.fcgi" I've tried Fcgi.hs with two different sample FastCGI programs with the same results. Any suggestions? Wayne

2 1

Trying to statically link Gtk2Hs application (Windows XP, network drive)
by Peter Schmitz 11 Aug '10

11 Aug '10

I am trying to statically link Gtk2Hs applications. E.g., the Gtk2Hs demo ...\demo\hello\World.hs from the SourceForge tarball: http://sourceforge.net/projects/gtk2hs/files/gtk2hs/0.10.1/gtk2hs-0.10.1.ta… My failed link attempts are listed at the end of this mail, but first some notes about my configuration. I need to develop in a Windows XP environment, where the toolset (Haskell Platform, Cabal, gtk, Gtk2hs, etc.) is located on a network drive (H:) instead of C:. (I know this may be unusual, but I need to be able to use any one of several PCs for development, with one common tools tree.) It is okay for temp files to reside on C:, but otherwise I have: H:\proc\tools\Haskell Platform\2010.1.0.0\ H:\proc\tools\gtk\ H:\proc\tools\cabal\ etc. (Fyi, I set up Cabal for this environment with some kind help: http://www.haskell.org/pipermail/haskell-cafe/2010-July/thread.html#80930 ). The _dynamically_ linked World.exe runs okay: PATH is: PATH= H:\proc\tools\Haskell Platform\2010.1.0.0\bin; H:\proc\tools\cabal\bin; H:\proc\tools\gtk\bin; H:\proc\tools\Haskell Platform\2010.1.0.0\lib\extralibs\bin; H:\proc\tools\Haskell Platform\2010.1.0.0\mingw\bin H:\proc\dev\Gtk2Hs\demo\hello>ghc --make World.hs [1 of 1] Compiling Main ( World.hs, World.o ) Linking World.exe ... Process completed, Exit Code 0. I had to add the mingw\bin dir to PATH to get cpp (needed during "cabal install gtk"), and the extralibs dir to get alex (for "cabal install gtk2hs-buildtools"). I would like users to be able to run my statically linked apps without having to worry about .dll issues at runtime, and just by double clicking on the app in a Windows explorer (file system display) window. When I try that for this dynamically linked World.exe, I get the error dialog: World.exe - Unable To Locate Component The application has failed to start because libcairo-2.dll was not found. Reinstalling the application may fix this problem. I consulted the ghc user guide, and did some Googling. I am not sure if the static linking difficulty is due to needing to link with the gtk (non Haskell) code (i.e., http://www.gtk.org/download-windows.html and http://ftp.gnome.org/pub/gnome/binaries/win32/gtk+/2.20/gtk+_2.20.0-1_win32… ), as perhaps alluded to in section "3. Static linking vs dynamic linking" in http://www.cs.virginia.edu/~wh5a/blog/Some%20GHC%20notes%20for%20myself.html (... The C libraries are not statically linked by default. To produce a completely static binary ...) or if it is just due in some way to my non-standard network drive configuration (ghc not finding some ".a" files library dirs to complete the static link, and so it defaults to dynamic linking?). Fyi, here are some of my dirs that have various *.a files: H:\proc\tools\cabal\cairo-0.11.0\ghc-6.12.1 H:\proc\tools\cabal\gio-0.11.0\ghc-6.12.1 H:\proc\tools\gtk\lib H:\proc\tools\Haskell Platform\2010.1.0.0\lib H:\proc\tools\Haskell Platform\2010.1.0.0\mingw\lib Other than PATH, I don't have any environment shell variable defined for searching a library path. Finally, here are my attempts to statically link: PATH is: PATH=H:\proc\tools\Haskell Platform\2010.1.0.0\bin;H:\proc\tools\cabal\bin;H:\proc\tools\gtk\bin;H:\proc\tools\Haskell Platform\2010.1.0.0\lib\extralibs\bin;H:\proc\tools\Haskell Platform\2010.1.0.0\mingw\bin H:\proc\dev\Gtk2Hs\demo\hello>ghc --make -static -optl-static World.hs [1 of 1] Compiling Main ( World.hs, World.o ) Linking World.exe ... H:\proc\tools\Haskell Platform\2010.1.0.0\mingw\bin\ld.exe: cannot find -lz collect2: ld returned 1 exit status Process completed, Exit Code 1. PATH is: PATH=H:\proc\tools\Haskell Platform\2010.1.0.0\bin;H:\proc\tools\cabal\bin;H:\proc\tools\gtk\bin;H:\proc\tools\Haskell Platform\2010.1.0.0\lib\extralibs\bin;H:\proc\tools\Haskell Platform\2010.1.0.0\mingw\bin H:\proc\dev\Gtk2Hs\demo\hello>ghc --make -optl-static -optl-pthread World.hs [1 of 1] Compiling Main ( World.hs, World.o ) Linking World.exe ... H:\proc\tools\Haskell Platform\2010.1.0.0\mingw\bin\ld.exe: cannot find -lz collect2: ld returned 1 exit status realgcc.exe: unrecognized option `-pthread' Process completed, Exit Code 1. Any advice, pointers, etc. would be appreciated. Thanks very much, -- Peter

2 3

Monadic fixed-point combinator.
by Lyndon Maydwell 11 Aug '10

11 Aug '10

Hi beginners. I've been looking at fixed-point combinators for the purpose of memoization for a few days recently and was wondering if there is a nice way to add monadic actions to an implicitly recursive[1] function. For example: If I have fib defined like so: > fib :: (Integer -> Integer) -> Integer -> Integer > fib f 0 = 0 > fib f 1 = 1 > fib f n = f (n-1) + f (n-2) I can create an inefficient fib' using fix: > fix :: (a -> a) -> a > fix f = f (fix f) > > fib' :: Integer -> Integer > fib' = fix fib Or a memoized fib'' using memoFix and a memoization scheme: > type MFB a b = (a -> b) -> a -> b > memoFix :: MFB a b -> MFB a b -> a -> b > memoFix mem f = let mf = mem (f mf) in mf > > integer :: (Int -> b) -> Int -> b > integer f = (A.listArray (0,200) (L.map f [0..200]) A.!) > > fib'' = memoFix integer fib But would there be a way to add, say, execution tracing using an IO combinator? > ioFix :: (a -> IO a) -> IO a > fix f = print "step" >> f (fix f) I've had a play around with these functions, attempting to lift them into monadic actions, but I really have no idea how to approach this. Thanks for your help! [1] - What do you call the form of a function where recursion is implemented by a fixed point combinator?

2 1

mayBe stuck
by prad 11 Aug '10

11 Aug '10

i'm trying to create my own split function with % as delimiter. so i have eqD = (=='%') and send let s = "zaoeu%aeuasnt%staashaeu%nthdanoe%nthd" putStrLn $ show $ brS (findIndex eqD s) s to a function brS: brS i ss | isNothing i = ss | otherwise = (take i ss) : (brS newIndex newStr) where newIndex = findIndex eqD newStr newStr = drop (i+1) ss but get the following error: Couldn't match expected type `Maybe a' against inferred type `Int' In the first argument of `isNothing', namely `i' In the expression: isNothing i :: mayBe a In a stmt of a pattern guard for the definition of `brS': isNothing i :: mayBe a my understanding is that i need the isNothing because findIndex will return Just Int or Nothing. however, i'm not sure how to resolve what seems to me to be an issue between a Maybe and an Int. -- In friendship, prad ... with you on your journey Towards Freedom http://www.towardsfreedom.com (website) Information, Inspiration, Imagination - truly a site for soaring I's

7 26

Applicative Functors: Rose Trees with an alternate behavior
by Travis Erdman 11 Aug '10

11 Aug '10

> The correct implementation of pure looks like this: > > pure x = Node x (repeat (pure x)) > > (Hooray for lazy infinite data structures!) With this definition the > standard sequenceA works just fine. Ha, I would never have arrived at that solution in a million years on my own, but now that you show me, it makes perfect sense! Following the discussion of ZipLists from http://learnyouahaskell.com/functors-applicative-functors-and-monoids, my initial solution attempt was pure x = Node x (repeat x) but that doesn't even compile. I had suspected that I needed to make the pure tree "go down", but I couldn't see how to do it. OK, some follow-on questions if I may ... sequenceA [Tree a] now returns Tree [a], as indicated. Now, I'd also like sequenceA Tree [a] to return [Tree a]. To do that, I need to make Tree an instance of Traversable and, hence, Foldable. Here's my stab at doing that ... instance Foldable Tree where foldMap f (Node cargo subtrees) = f cargo `mappend` foldMap (foldMap f) subtrees instance Traversable Tree where traverse f (Node cargo subtrees) = Node <$> f cargo <*> traverse (traverse f) subtrees The Foldable code appears to be correct; at least, I can the fold the trees. "Foldable" also allows the toList Tree a to work as expected. Given that, I don't really "get" Traversable, and the code here is just my best guess as to what it should be. But, I think it must not be correct, because sequenceA Tree [a] is not returning what I think it should be returning (ie [Tree a]). Aside from this, what other things can I do with a Traversable Tree? My intuition suggests I might be able to do Scan's on a tree, say calculate a cumulative sums Tree from root to leaves (or vice versa). But I've no idea how to implement that using it's "Traversability". As of now, I have implemented this up-and-down scanning thusly ... treeScanDown :: (a -> b -> a) -> a -> Tree b -> Tree a treeScanDown f x (Node y subtrees) = Node g (fmap (treeScanDown f g) subtrees) where g = f x y treeScanUp :: (a -> [a] -> a) -> Tree a -> Tree a treeScanUp f (Node x []) = Node x [] treeScanUp f (Node x subtrees) = Node (f x (fmap fromNode g)) g where g = fmap (treeScanUp f) subtrees thanks again, travis

2 1

Why does this work in a program, but cause a seg fault in ghci?
by Kurt Häusler 11 Aug '10

11 Aug '10

Hi, I am going through the OpenGL tutorial 1 (http://www.haskell.org/haskellwiki/OpenGLTutorial1) The first example program in chapter 2 works if I compile it and run it, but I don't understand all of it so I am playing with bits of it in ghci. Look what happens: GHCi, version 6.10.4: http://www.haskell.org/ghc/ :? for help Loading package ghc-prim ... linking ... done. Loading package integer ... linking ... done. Loading package base ... linking ... done. Prelude> let myPoints = map (\k -> (sin(2*pi*k/12),cos(2*pi*k/12),0.0)) [1..12] Prelude> myPoints [(0.49999999999999994,0.8660254037844387,0.0),(0.8660254037844386,0.5000000000000001,0.0),(1.0,6.123233995736766e-17,0.0),(0.8660254037844388,-Segmentation fault Any ideas? Thanks

4 4

Using Debug.Trace
by Patrick LeBoutillier 10 Aug '10

10 Aug '10

Hi all, I'm writing a parser for a binary format, and I'm trying to debug it since I have a bug and the code is getting lost in the bits and bytes. Basically my main is like this: import Data.Binary.Get import Debug.Trace import qualified Data.ByteString.Lazy as L main = do bytes <- L.readFile "song.gp4" let (version, bytes') = getVersion bytes putStrLn version let stuff = runGet (getDocument version) bytes' putStrLn $ show stuff return () The getVersion and getDocument functions use the Data.Binary.Get monad to decode the byte string into various objects. I tried sprinkling "trace" calls a bit everywhere and I realize they don't always get called at the expected time. Most of them are only called when I reach the "putStrLn $ show stuff" statement. Unfortunately that doesn't help me because I get a *** Exception: too few bytes. Failed reading at byte position 35939 before that. Is this happening because I'm using lazy IO to read the file? Is there a way to force the evaluation of these "trace" calls? Are there any other ways to debug this kind of stuff in Haskell? Thanks a lot, Patrick -- ===================== Patrick LeBoutillier Rosemère, Québec, Canada

3 3

Applicative Functors: Rose Trees with an alternate behavior
by Travis Erdman 10 Aug '10

10 Aug '10

The rose trees implemented in Data.Tree behave in the same way as the applicative style on lists (ie all possible combinations/permutations). I need a different behavior: one-to-one application. So, using the Data.Tree code, if I start with this tree: *Main Data.Traversable> putStrLn $ drawTree (fmap show temp2) 1 | +- 2 | `- 3 and then try this: *Main Data.Traversable> putStrLn $ drawTree (fmap show $ sequenceA [temp2,temp2]) [1,1] | +- [1,2] | +- [1,3] | +- [2,1] | | | +- [2,2] | | | `- [2,3] | `- [3,1] | +- [3,2] | `- [3,3] It gives all the combinations/permutations. I need this: *Main Data.Traversable> putStrLn $ drawTree (fmap show $ seqA [temp2,temp2]) [1,1] | +- [2,2] | `- [3,3] But then only way I could get that result was to (re-)write my own sequenceA function (called it seqA), which I suspect shouldn't be necessary. If I use the built-in sequenceA, I get (only) this: *Main Data.Traversable> putStrLn $ drawTree (fmap show $ sequenceA [temp2,temp2]) [1,1] Here's my code: data Tree a = Node {fromNode :: !a, fromNodeForest :: !(Forest a)} deriving (Show, Eq) type Forest a = [Tree a] zipWithForest = zipWith emptyForest = [] instance Functor Tree where fmap f (Node x subtrees) = Node (f x) (fmap (fmap f) subtrees) instance Applicative Tree where pure x = Node x emptyForest (<*>) (Node f tfs) (Node cargo subtrees) = Node (f cargo) (zipWithForest (<*>) tfs subtrees) seqA :: (Applicative f) => [f a] -> f [a] seqA (x:xs) = foldr (liftA2 (:)) (fmap (const []) x) (x:xs) I think I've got something wrong in how I've defined the Applicative Functor methods, and "cheating" by writing a custom sequenceA function to work around the deficiency. What is the proper fix? thanks again, travis

2 1