March 2009 - Haskell-Cafe

Binary Parsing
by Rick R 11 Mar '09

11 Mar '09

I have basic beginning to a parser for the BSON spec: http://www.mongodb.org/display/DOCS/BSON It is basically a binary compressed form of JSON. The usage model should be general, but I intend to read this data over TCP. Currently my system is quite inefficient, I convert leading bytes to Int then switch based on that type that that Int represtents. I was wondering if there is a tool set that exists for this purpose. For instance, should I use Data.Binary and make my Binary instance of get and put use the BSON protocol? Or is that not correct? I also looked at Parsec.ByteString, but that seems to only have a file input mechanism, and tcp buffers may be out of its scope. Are there any other tools that I should look at? I'm sure similar things have been done before. Can anyone point me to some open, successful implementations that I could mimick? -- We can't solve problems by using the same kind of thinking we used when we created them. - A. Einstein

2 1

generic sum for an UArr
by Manlio Perillo 11 Mar '09

11 Mar '09

Hi. I have an array (Data.Array.Vector.UArr) of Word8 elements. Elements are stored as Word8 to keep memory usage down. Now I need to compute the sum of all the array elements; what is the best method? The sum must be of type Int. How efficient is to simply convert to an UArr Int, using: mapU fromIntegral v :: UArr Int ? What about, instead, of something like: genericSumU :: (UA e, Integral e) => UArr e -> Int genericSumU = foldU add 0 where add x y = x + fromIntegral y ? Unfortunately the latter does not compile: Couldn't match expected type `Int' against inferred type `e' `e' is a rigid type variable bound by the type signature for `genericSumU' at bin/process-data-1.hs:64:19 In the expression: foldU add 0 In the definition of `genericSumU': genericSumU = foldU add 0 where add x y = x + fromIntegral y Moreover, this is not really a generic version. I would like to have: genericSumU :: (Num i, UA e, Num e) => UArr e -> i is this possible? Thanks Manlio

2 2

How to insert character key self in sourceView?
by Andy Stewart 11 Mar '09

11 Mar '09

Hi all, I use gtk2hs develop editor. I use below function handle key press event. keyPressHandler :: Event -> IO Bool keyPressHandler (Key {eventKeyName = keyName}) = do case keyName of "Escape" -> do mainQuit >> return True _ -> do -- How to insert character self? return True I need handle key press event, and insert itself if key is character (like: 'a' 'b' 'c'). How to insert character self in sourceView buffer? Below are complete source code: ------------------------------> source code start <------------------------------ import Graphics.UI.Gtk import Graphics.UI.Gtk.SourceView import Graphics.UI.Gtk.Abstract.Widget import Graphics.UI.Gtk.Gdk.Events import Text.Printf import Control.Monad main :: IO () main = do -- Init. initGUI -- Root frame. rootFrame <- windowNew onDestroy rootFrame mainQuit -- quit main loop when root frame close -- Root frame status. -- windowFullscreen rootFrame -- fullscreen windowSetPosition rootFrame WinPosCenter -- set init position windowSetDefaultSize rootFrame 400 300 -- set init size -- Main box. mainBox <- vBoxNew False 0 containerAdd rootFrame mainBox -- Source view box. sourceViewBox <- vBoxNew False 0 containerAdd mainBox sourceViewBox -- Source view. sourceView <- sourceViewNew boxPackStart sourceViewBox sourceView PackNatural 0 -- Echo area box. echoAreaBox <- vBoxNew False 0 echoAreaAlign <- alignmentNew 0 1 1 1 containerAdd echoAreaBox echoAreaAlign containerAdd mainBox echoAreaBox -- Echo area. echoArea <- statusbarNew boxPackStart echoAreaBox echoArea PackNatural 0 -- Display. widgetShowAll rootFrame -- Handle keystroke. onKeyPress rootFrame $ keyPressHandler -- Update echo area. updateEchoArea sourceView echoArea -- Loop mainGUI keyPressHandler :: Event -> IO Bool keyPressHandler (Key {eventKeyName = keyName}) = do case keyName of "Escape" -> do mainQuit >> return True _ -> do return True updateEchoArea :: SourceView -> Statusbar -> IO () updateEchoArea sv sb = do buf <- textViewGetBuffer sv mark <- textBufferGetInsert buf iter <- textBufferGetIterAtMark buf mark line <- textIterGetLine iter col <- textIterGetLineOffset iter statusbarPop sb 1 statusbarPush sb 1 $ printf "Line %4d, Column %3d" (line + 1) (col + 1) return () ------------------------------> source code end <------------------------------ Any help? Thanks! -- Andy

2 2

advice on a parsing function
by Manlio Perillo 11 Mar '09

11 Mar '09

Hi. I'm still working on the Netflix Prize; now I have managed to implement a parsing function for the qualifying data set (or quiz set). The quiz set is in the format: 1: 10 20 30 2: 100 200 3: 1000 2000 3000 4000 5000 Where 1, 2, 3 are movie ids, and the others are user ids. The parsing program is at: http://hpaste.org/fastcgi/hpaste.fcgi/view?id=2300 The program reads the file using lazy IO. One of the feature I want is, for the quiz function, to be a *good producer*. I'm quite satisfied with the result (this is the first "complex" parsing function I have written in Haskell), and I have also managed to avoid the use of an accumulator. However I'm interested to know it there is a more efficient solution. The qualifying.txt file is 51MB, 2834601 lines. On my laptop, the function performance are: real 1m14.117s user 0m2.496s sys 0m0.136s CPU usage is about 3%, system load is about 0.20, memory usage is 4956 KB. What I'm worried about is: quiz' ((id, ":") : l) = (id, quiz'' l) : quiz' l quiz' ((id, _) : l) = quiz' l the problem here is that the same elements in the list are processed multiple times. I have written alternate versions. The first using foldl http://hpaste.org/fastcgi/hpaste.fcgi/view?id=2303 (that, however, builds the entire data structure in memory, and in reverse order) The latter using foldr http://hpaste.org/fastcgi/hpaste.fcgi/view?id=2304 (that, however, is incorrect and I'm unable to fix). The performances of the foldr version are very similar to the performances of the first implementation (it make use, however, of 3704 KB, and it is about 3 seconds faster). P.S: the expected result for the sample quiz set I have posted is: [(1,[10,20,30]),(2,[100,200]),(3,[1000,2000,3000,4000,5000])] The foldl version produces: [(3,[5000,4000,3000,2000,1000]),(2,[200,100]),(1,[30,20,10])] The foldr version produces: [(1,[]),(2,[10,20,30]),(3,[100,200]),(5000,[1000,2000,3000,4000])] Thanks Manlio Perillo

2 7

ByteString in patterns
by Manlio Perillo 11 Mar '09

11 Mar '09

Hi. Using normal String type I can define a pattern like: > let foo "baz" = 777 > foo "baz" 777 But if I want to use ByteString, what should I do? This seems impossible, since ByteString data constructor is not available. Thanks Manlio Perillo

5 12

View patterns and warnings about overlapping or non-exhaustive patterns
by Stephan Friedrichs 11 Mar '09

11 Mar '09

Hi, I'm working on a data structure that uses Data.Sequence a lot, so views are important and I tried to simplify my code using view patterns. The problem is, that I keep getting warnings about both overlapping and non-exhaustive pattern matches. A simple test case: ===============================T.hs=============================== {-# LANGUAGE ViewPatterns #-} import Data.Sequence test :: Seq a -> Seq b -> String test (viewl -> EmptyL) (viewl -> EmptyL) = "empty, empty" test (viewl -> EmptyL) (viewl -> _ :< _) = "empty, non-empty" test (viewl -> _ :< _) (viewl -> EmptyL) = "non-empty, empty" test _ _ = "non-empty, non-empty" ================================================================== > ghci -Wall T.hs GHCi, version 6.10.1: http://www.haskell.org/ghc/ :? for help Loading package ghc-prim ... linking ... done. Loading package integer ... linking ... done. Loading package base ... linking ... done. [1 of 1] Compiling Main ( T.hs, interpreted ) T.hs:6:0: Warning: Pattern match(es) are overlapped In the definition of `test': test ((viewl -> EmptyL)) ((viewl -> EmptyL)) = ... test ((viewl -> EmptyL)) ((viewl -> _ :< _)) = ... test ((viewl -> _ :< _)) ((viewl -> EmptyL)) = ... test _ _ = ... T.hs:6:0: Warning: Pattern match(es) are non-exhaustive In the definition of `test': Patterns not matched: Ok, modules loaded: Main. *Main> test empty (singleton 'a') "empty, non-empty" *Main> test (singleton 'b') (singleton 'a') "non-empty, non-empty" *Main> test (singleton 'b') empty "non-empty, empty" *Main> test empty empty "empty, empty" There are warnings about non-exhaustive and overlapping pattern matches, but the tests show that this isn't the case. So what's the problem? I don't want to turn off or ignore warnings. //Stephan -- Früher hieß es ja: Ich denke, also bin ich. Heute weiß man: Es geht auch so. - Dieter Nuhr

3 3

QC 2.0 missing some stuff I need
by John Goerzen 11 Mar '09

11 Mar '09

Hi, QuickCheck 1.x had this function: evaluate :: Testable a => a -> Gen Result which I used in TestPack to help wrap a QuickCheck test as a HUnit test case. QuickCheck 2.x seems to have no pure evaluate-like function at all; all of its functions are in the IO monad and also write their result to stdout, according to the docs. Am I missing something? Thanks, -- John

2 1

Using FosSyDe to translate haskell to VHDL.
by wanghanyi 11 Mar '09

11 Mar '09

Dear all, I have used ForSyDe to translate a Haskell script to VHDL for a while, and I have one problem by applying where-clause during the translation. I wrote a small function Plus2.hs which is implied as following addTwof :: ProcFun (Int32 -> Int32) addTwof = $(newProcFun [d|addTwof :: Int32 -> Int32 addTwof n = addOnef+1 where addOnef = n +1 |]) *Plus2> writeVHDL plus2SysDef *** Exception: VHDL Compilation Error: Untranslatable function: where constructs are not supported in functions: where addOnef_0 = n_1 GHC.Num.+ 1 in process function `addTwof' (created in Plus2) used by process `plus2Proc' belonging to system definition `plus2' (created in <unkown>) The compiling process was fine, but the script could not be translated into VHDL, like the error i mentioned above. Is there any way to solve this problem if a where-clause is a must. Thanks all. Hanyi _________________________________________________________________ Live Search视频搜索，快速检索视频的利器！ http://www.live.com/?scope=video

2 2

Re: [Haskell-cafe] Design Patterns by Gamma or equivalent
by Sean Leather 11 Mar '09

11 Mar '09

Hi Mark, Because Haskell is not OO, it is functional, I was wondering if there is > some kind of analogous “design pattern”/”template” type concept that > describe commonly used functions that can be “factored out” in a general > sense to provide the same kind of usefulness that Design Patterns do for > OOP. Basically I’m asking if there are any kinds of “common denominator” > function compositions that are used again and again to solve problems. If > so, what are they called? > Look at Jeremy Gibbons' publications. All of the ones that have "pattern" or "origami" in the title describe design patterns in Haskell. Some of them relate to the translation of Gang-of-Four patterns. http://www.comlab.ox.ac.uk/people/publications/date/Jeremy.Gibbons.html Regards, Sean

1 0

Re: Design Patterns by Gamma or equivalent
by Achim Schneider 11 Mar '09

11 Mar '09

"Mark Spezzano" <mark.spezzano(a)chariot.net.au> wrote: > Basically I___m asking if there are any kinds of ___common denominator___ > function compositions that are used again and again to solve > problems. If so, what are they called? > Haskellers tend to cast their design patterns into functions and libraries. What they're supposed to be called is a zealously heated discussion, there's, for example, return == pure == point and fmap == map == (.) == liftM [1] OTOH, we're said to not care at all: | Don't underrate Haskell programmers. The ones I know have the kind of | minds where if you switched the names to pig latin, they wouldn't skip | a beat. -- [2] If you're looking for a text that could be named "The N most (use|success)ful Haskell Design Patterns", I recommend having a look at the Typeclassopedia[3]. [1]well, at least there _could_ be. [2]http://www.mail-archive.com/boost@lists.boost.org/msg08898.html [3]http://byorgey.wordpress.com/2009/02/16/the-typeclassopedia-request-for-f… -- (c) this sig last receiving data processing entity. Inspect headers for copyright history. All rights reserved. Copying, hiring, renting, performance and/or quoting of this signature prohibited.

1 0