- Haskell-Cafe - Haskell.org

Using GADTs
by Matthew Pocock 29 Jul '07

29 Jul '07

Hi, I'm trying to get to grips with GADTs, and my first attempt was to convert a simple logic language into negative normal form, while attempting to push the knowledge about what consitutes negative normal form into the types. My code is below. I'm not entirely happy with it, and would appreciate any feedback. The rules are that in nnf, only named concepts, the concept Top and the concept Bottom can be negated. So, I've added three NNFNegation_* constructors capturing these legal cases. Is there a way to use one constructor, that is allowed to 'range over' these three cases and none of the others? I've ended up producing two data types, one for the general form and one for the nnf. Actually, the constraint on what constitutes nnf is fairly obvious - no complex terms are negated. Is there a way to use just the one data type but to describe two levels of constraints - one for the general case, and one for the nnf case? Or is the whole point that you capture each set of constraints in a different data type? Thanks, Matthe data Named data Equal data Conjunction data Disjunction data Negation data Existential data Universal data Top data Bottom data Concept t where CNamed :: String -> Concept Named CEqual :: Concept a -> Concept b -> Concept Equal CConjunction :: Concept a -> Concept b -> Concept Conjunction CDisjunction :: Concept a -> Concept b -> Concept Disjunction CNegation :: Concept a -> Concept Negation CExistential :: Role Named -> Concept Existential CUniversal :: Role Named -> Concept Universal CTop :: Concept Top CBottom :: Concept Bottom data NNFConcept t where NNFCNamed :: String -> NNFConcept Named NNFCEqual :: NNFConcept a -> NNFConcept b -> NNFConcept Equal NNFCConjunction :: NNFConcept a -> NNFConcept b -> NNFConcept Conjunction NNFCDisjunction :: NNFConcept a -> NNFConcept b -> NNFConcept Disjunction NNFCExistential :: Role Named -> NNFConcept Existential NNFCUniversal :: Role Named -> NNFConcept Universal NNFCTop :: NNFConcept Top NNFCBottom :: NNFConcept Bottom NNFCNegation_N :: NNFConcept Named -> Concept Negation NNFCNegation_T :: NNFConcept Top -> Concept Negation NNFCNegation_B :: NNFConcept Bottom -> Concept Negation data Role t where RNamed :: String -> RNamed Named -- terms not prefixed with a negation are already in nnf nnf :: Concept t -> NNFConcept u nnf CNamed name = NNFCNamed name nnf CEqual lhs rhs = NNFConcept (nnf lhs) (nnf rhs) nnf CConjunction lhs rhs = NNFCConjunction (nnf lhs) (nnf rhs) nnf CDijunction lhs rhs = NNFCDisjunction (nnf lhs) (nnf rhs) nnf CExistential r c = NNFCExistential r (nnf c) nnf CUniversal r c = NNFCUniversal r (nnf c) -- if negated, we must look at the term and then do The Right Thing(tm) nnf CNegation (CNamed name) = NNFCNegation_N NNFCNamed name nnf CNegation (CEqual lhs rhs) = NNFCEqual (nnf $ CNegation lhs) (nnf $ CNegation rhs) nnf CNegation (CConjunction lhs rhs) = NNFCDisjunction (nnf $ CNegation lhs) (nnf $ CNegation rhs) nnf CNegation (CDisjunction lhs rhs) = NNFCConjunction (nnf $ CNegation lhs) (nnf $ CNegation rhs) nnf CNegation (CNegation c) = nnf c nnf CNegation (CExistential r c) = NNFCUniversal r (nnf $ CNegation c) nnf CNegation (CUniveral r c) = NNFCExistential r (nnf $ CNegation c) nnf CNegation CTop = NNFCNegation_T NNFCTop nnf CNegation CBottom = NNFCNegation_B NNFCBottom

2 2

ANN: Finance.Quote.Yahoo 0.1 on hackage
by brad clawsie 29 Jul '07

29 Jul '07

http://hackage.haskell.org/cgi-bin/hackage-scripts/package/Finance-Quote-Ya… this is a simple module to get stock quote information from yahoo finance, considered alpha quality

4 6

Text.XHtml Documentation
by David F. Place 29 Jul '07

29 Jul '07

Hello: I was just looking into using Text.XHtml for a project. The Haddock documentation points to an introduction by Andy Gill, but the link is broken. Can anyone point me to the correct location of the introduction? Thanks. Cheers, David ___________________ (---o-------o-o-o---o-o-o----( David F. Place mailto:d@vidplace.com

2 1

problems installing Takusen-0.6
by Rahul Kapoor 29 Jul '07

29 Jul '07

I am having problems installing Takusen-0.6 (ghc 6.6.1 on FreeBSD) The configure and build works fine. running ./setup install fails with: Installing: /usr/local/lib/Takusen-0.6/ghc-6.6.1 & /usr/local/bin Takusen-0.6... setup: Error: Could not find module: Database.Oracle.Enumerator with any suffix: ["hi"] I only have sqlite3 installed (no Oracle or Postgres). I followed the usual steps for the install ghc --make -o setup Setup.hs ./setup configure ./setup build ./setup install Any Ideas? Rahul

2 1

Videos of Haskell talks
by dons＠cse.unsw.edu.au 29 Jul '07

29 Jul '07

A new wiki page collecting tutorials and presentations about Haskell, is on the wiki: http://haskell.org/haskellwiki/Video_presentations Highlights include: Transactional Memory for Concurrent Programming Parametric Polymorphism and the Girard-Reynolds Isomorphism Programming in the Age of Concurrency Nested Data Parallelism in Haskell If you know of other online video material, please add it! -- Don

1 0

Using Gtk2Hs version 0.9.12 on a PPC Mac
by Thorkil Naur 28 Jul '07

28 Jul '07

Hello, (From the archives:) > [Haskell] ANNOUNCE: Gtk2Hs version 0.9.12 released > Duncan Coutts duncan.coutts at worc.ox.ac.uk > Fri Jul 27 15:20:57 EDT 2007 > Gtk2Hs - A GUI Library for Haskell based on Gtk+ > Version 0.9.12 is now available from: > ... > Duncan > (on behalf of the Gtk2Hs team) Following the advice of nominolo_ from #haskell yesterday, I added -L/opt/local/lib to my ghc command and then the helloworld demo of Gtk2Hs worked. The problem seems to be the order in which the two library directories /usr/X11R6/lib and /opt/local/lib appear on the link command: For the GTK+ example that worked, /opt/local/lib comes first. For the Gtk2Hs demo that failed, /usr/X11R6/lib comes first. I compared these two directories and found these common files: X11 libXft.2.dylib libXft.a libXft.dylib libXrender.1.dylib libXrender.a libXrender.dylib libexpat.0.dylib libexpat.a libexpat.dylib libfontconfig.1.dylib libfontconfig.a libfontconfig.dylib libfreetype.6.dylib libfreetype.a libfreetype.dylib pkgconfig Since some of these (Xrender, fontconfig, and freetype) are actually used in GTK+ (and therefore also Gtk2Hs) applications, there is clearly a potential for conflict here. In any case, adding -L/opt/local/lib to the ghc command makes this library appear before /usr/X11R6/lib on the link command and this seems to solve the problem. Subsequently, I have tried a handful of the other demos, and as far as they didn't require something that wasn't available (glade, for example), they seemed to work. Thanks and best regards Thorkil

2 3

Frag
by Jon Harrop 28 Jul '07

28 Jul '07

I just stumbled upon this fast action 3D shooter written entirely in Haskell: http://haskell.org/haskellwiki/Frag After 15 minutes trying to build it I find that it segfaults. Can anyone else get this to work? -- Dr Jon D Harrop, Flying Frog Consultancy Ltd. The OCaml Journal http://www.ffconsultancy.com/products/ocaml_journal/?e

5 7

Handling custom types in Takusen
by Salvatore Insalaco 28 Jul '07

28 Jul '07

Disclaimer: I don't know Takusen very well, and I'm talking of Sqlite backend. I noticed that in Takusen there're just two instances to implement to make any Haskell type db-serializable: DBBind / SqliteBind for serialization and DBType for deserialization. I wanted to implement blob serialization for PackedStrings, but I noticed that both DBBind and DBType classes are not public. Defining custom serialization (not just Show / Read string serialization) is very useful in applicative code, and not just in Takusen backend code. There's any way to do it without modifying Takusen Sqlite backend? Salvatore Insalaco

4 8

Memory profiler
by Jon Harrop 27 Jul '07

27 Jul '07

Is there a memory profiler for Haskell? -- Dr Jon D Harrop, Flying Frog Consultancy Ltd. OCaml for Scientists http://www.ffconsultancy.com/products/ocaml_for_scientists/?e

2 1

Avoiding boilerplate retrieving GetOpt cmd line args
by Dave Bayer 27 Jul '07

27 Jul '07

Ok, I'm writing a command line tool, using System.Console.GetOpt to handle command line arguments. My Flags structure so far is > data Flag > = Filter String > | DateFormat String > | DocStart String > | DocEnd String ... and I want to write accessor functions that return the strings if specified, otherwise returning a default. The best I've been able to do is this: > getFilter = getString f "Markdown.pl" > where f (Filter s) = Just s > f _ = Nothing > > getDateFormat = getString f "%B %e, %Y" > where f (DateFormat s) = Just s > f _ = Nothing > > getDocStart = getString f "^{-$" > where f (DocStart s) = Just s > f _ = Nothing > > getDocEnd = getString f "^-}$" > where f (DocEnd s) = Just s > f _ = Nothing using a generic accessor function `getString`. There are eight (and growing) needless lines here, where what I really want to do is to pass the constructors `Filter`, `DateFormat`, `DocStart`, or `DocEnd` to the function `getString`. ghci types each of these as `String -> Flag`, so one at least knows how to type such a `getString`, but using a constructor-passed-as-an-argument in a pattern match is of course a "Parse error in pattern". (I expected as much, but I had to try... `String -> Flag` is not enough information to make it clear we're passing a constructor, rather than some hairy arbitrary function, so such a pattern match would be undecidable in general.) So what's the right idiom for avoiding this boilerplate? Thanks, Dave

8 8