May 2008 - Haskell-Cafe

Data.Dynamic over the wire
by Jules Bean 14 May '08

14 May '08

> {-# LANGUAGE ScopedTypeVariables #-} Data.Dynamic gives a passable impression of adding support for dynamically typed code and runtime typing to GHC, without changing the basic statically typed, all types known at runtime nature of the language. Note that Data.Dynamic relies upon two things: it relies upon a concrete representation of types, given by TypeRep, and a primitive which has to be provided by the compiler to actually implement fromDynamic. (In GHC it uses unsafeCoerce# which is already available, but you could imagine providing other primitives). In principle TypeReps could be derived by hand, although if you do so you can break everything by providing invalid instances. In practice we'd rather the compiler did it for us and guaranteed safety. You can do all sorts of things with Dynamic, but the general pattern is that data which has some fixed, known type, can be passed through a chunk of code which doesn't know its type (wrapped in Dynamic) and then eventually consumed by another piece of code which *does* know the type, and can unwrap it. The consuming code has to know the type to unwrap it, although it can 'guess' various alternatives if it wants, and thus type safety is preserved. One thing which you can't obviously do is write Read or Show instances for Dynamic. So can we pass Dynamic data over the wire? If not, Dynamic is limited to the context of "within a single program", and can't be used over the network between cooperating programs, or in file formats, etc. You can try this: > import Data.Typeable > data SerialisedDynamic = SD TypeRep String deriving (Show) > freeze :: (Show a, Typeable a) => a -> SerialisedDynamic > freeze x = SD (typeOf x) (show x) > thaw :: forall a . (Read a, Typeable a) => SerialisedDynamic -> Maybe a > thaw (SD t s) = if typeOf (undefined :: a) == t then > Just (read s) > else Nothing This is close, and works as far as it goes. It is a limited reimplementation of Dynamic which uses show/read instead of unsafeCoerce#. As such it is pure haskell (but relies on Typeable instances). You can't finish it off because you can't derive a 'Read' instance for SD, because there is no read instance for TypeRep. Off-hand I can't think of any reason why there can't be a Read instance for TypeRep, but it would be a bit tricky with the current TypeRep because of the way its implemented, I think. You need to take care about globally qualified types and might want to use package names like ghc does in its linking phase, but those are definitely surmountable problems. Having said all that, I'm not sure how useful this really is. Most of the time you could use this, you could equally just pass around the String and 'read' it once you get to the place where you want to use the value. Practical over-the-wire protocols necessarily have some kind of tagging mechanism, and all this adds is a global "tag table" for Typeable types via TypeRep. Jules

5 4

Using Template Haskell to make type-safe database access
by Mads Lindstrøm 14 May '08

14 May '08

Hi, I was wondering if anybody had experimented with using Template Haskell (TH) and ordinary SQL to make type-safe database access? To clarify what I am thinking about I will sketch how it could be done. The TH function should take two inputs. SQL (as a string) and a database source name (DSN). It should return an IO action as output. The TH-function should: 1. Connect to the database using the DSN 2. Ask the database which types will be returned from the expression 3. Build an IO action which can be used to execute the SQL at run-time. The action could return the result as a (lazy) list. Due to step two we can make the returned values type-safe. Greetings, Mads Lindstrøm

4 9

Re: [Haskell-cafe] Richer (than ascii) notation for haskell source?
by Andrew Coppin 14 May '08

14 May '08

Reinier Lamers wrote: > > Op 14-mei-2008, om 20:32 heeft Andrew Coppin het volgende geschreven: >> Personally, I'd just like to be able to get rid of "->", "\" and >> other such hacks. Would it be possible to amend GHC so that it >> accepts "->" and [whatever the Unicode codepoint for "left arrow" is] >> and treats both the same? > I believe GHC already accepts "∀" instead of "forall", and some other > Unicode neatness. However, you can't use "λ" for "\" because "λ" is a > lowercase letter and thus valid in identifiers. So getting rid of "\" > might be impossible even with Unicode. Ooo... well spotted on that subtly edge case with lambda... Damn, that's just too bad. Using a backslash makes it look like a character escape or something. It's really quite ugly...

1 0

printing a list of directories which don't exist
by Mike Jarmy 14 May '08

14 May '08

Newbie question: Given a list of type '[FilePath]', how do I create a list of all those directories which do not actually exist, and then print the list? I've figured out how to extract the ones which *do* exist, like so: module Main where import Control.Monad (filterM) import System.Directory (doesDirectoryExist) import System.Environment (getArgs) main :: IO () main = do dirs <- getArgs let existing = filterM doesDirectoryExist dirs ...... which gives me a list of type 'IO [FilePath]'. However, because of the 'IO' tag, I cannot figure out how to do any of the following 3 things (noted in comments): -- filter via composition let bogusDirs = filterM (not . doesDirectoryExist) dirs -- test for emptiness if bogusDirs /= [] -- print the list then putStrLn $ "bogus: " ++ show bogusDirs else putStrLn "OK" Can anyone set me straight? How do I make the IO tag go away, or am I going about this all wrong? E.g. the 'filterM (not . doesDirectoryExist) dirs' expression gives the following compilation error: ~/code/haskell$ ghc -o newbie newbie.hs newbie.hs:16:35: Couldn't match expected type `Bool' against inferred type `IO Bool' Expected type: FilePath -> Bool Inferred type: FilePath -> IO Bool In the second argument of `(.)', namely `doesDirectoryExist' In the first argument of `filterM', namely `(not . doesDirectoryExist)'

3 2

Maybe a, The Rationale
by PR Stanley 14 May '08

14 May '08

Hi folks data Maybe a = Nothing | Just a What is the underlying rationale for the Maybe data type? is it the safe style of programming it encourages/ Something tells me this is going to start a lengthy discussion. :-) Cheers, Paul

11 18

Re: Trying to avoid duplicate instances
by oleg＠okmij.org 14 May '08

14 May '08

Eric Stansifer wrote: > I am using a bunch of empty type classes to categorize some objects: > > > class FiniteSolidObject o > > class FinitePatchObject o > > class InfiniteSolidObject o > > Since "solid objects" are exactly "finite solid objects" plus > "infinite solid objects", there is an obvious way to code up this > logical relationship. So I try to write: > > > class SolidObject o > > instance FiniteSolidObject o => SolidObject o > > instance InfiniteSolidObject o => SolidObject o There is an easy way to accomplish your goal in GHC, with a couple of simple tricks. First of all, it pays in this and many related cases do define the full predicate for the classification of objects. That is, rather than defining the constraint (FiniteSolidObject o) that succeeds when 'o' is really a FiniteSolidObject and fails otherwise, define a constraint (FiniteSolidObjectP o f) which always succeeds; it unifies the second argument with HTrue if o is indeed FiniteSolidObject, and unifies f with HFalse otherwise. We can easily define the semi-predicate (FiniteSolidObject o) if so needed. That is the first trick. One can then introduce typeclasses HAnd, HOr, etc. and use them to define more complex predicates like SolidObject (which is the logical OR of FiniteSolidObjectP and InfiniteSolidObjectP). Defining the primitive predicates such as FiniteSolidObject does including the trick, the TypeCast. It may appear bizarre; well, if the point is to use it rather than contemplate it, one can just take the pattern for granted. It works. http://okmij.org/ftp/Haskell/typecast.html But there is even a more uniform way, explained in http://okmij.org/ftp/Haskell/poly2.hs http://okmij.org/ftp/Haskell/poly2.txt Using that file, we can write your code as follows: {-# OPTIONS -fglasgow-exts #-} {-# OPTIONS -fallow-undecidable-instances #-} {-# OPTIONS -fallow-overlapping-instances #-} module Mem where -- primitive types data Box = Box data Sphere = Sphere data Mesh = Mesh data Plane = Plane -- classes of types type FiniteSolidObjects = Box :*: Sphere :*: HNil type FinitePatchObjects = Mesh :*: HNil type InfiniteSolidObjects = Plane :*: HNil -- All of finite and infinite solid objects are solid objects type SolidObjects = AllOf FiniteSolidObjects :*: AllOf InfiniteSolidObjects :*: HNil -- membership predicate -- Statically tests if an object of the type x is a member of the class c is_of_class :: forall c x r. Apply (Member c) x r => x -> c -> r is_of_class x t = apply (undefined::Member c) x test1 = is_of_class Box (undefined::FiniteSolidObjects) -- type HTrue test2 = is_of_class Box (undefined::SolidObjects) -- type HTrue test3 = is_of_class Box (undefined::InfiniteSolidObjects) -- type HFalse test4 = is_of_class Plane (undefined::SolidObjects) -- type HTrue -- make a semi-predicate SolidObject class SolidObject c instance Apply (Member SolidObjects) x HTrue => SolidObject x test_solid :: SolidObject x => x -> () test_solid = undefined ts1 = test_solid Plane -- ts3 = test_solid Mesh -- causes the type error -- The following is borrowed verbatim from poly2.hs type Fractionals = Float :*: Double :*: HNil type Nums = Int :*: Integer :*: AllOf Fractionals :*: HNil type Ords = Bool :*: Char :*: AllOf Nums :*: HNil type Eqs = AllOf (TypeCl OpenEqs) :*: AllOfBut Ords Fractionals :*: HNil -- The Fractionals, Nums and Ords above are closed. But Eqs is open -- (i.e., extensible), due to the following: data OpenEqs instance TypeCls OpenEqs () HTrue -- others can be added in the future -- Type class membership testing data AllOf x data AllOfBut x y data TypeCl x -- Classifies if the type x belongs to the open class labeled l -- The result r is either HTrue or HFalse class TypeCls l x r | l x -> r -- the default instance: x does not belong instance TypeCast r HFalse => TypeCls l x r -- Deciding the membership in a closed class, specified -- by enumeration, union and difference data Member tl instance Apply (Member HNil) x HFalse instance TypeCls l x r => Apply (Member (TypeCl l)) x r instance (TypeEq h x bf, MemApp bf t x r) => Apply (Member (h :*: t)) x r instance (Apply (Member h) x bf, MemApp bf t x r) => Apply (Member ((AllOf h) :*: t)) x r instance (Apply (Member exc) x bf, Apply (MemCase2 h t x) bf r) => Apply (Member ((AllOfBut h exc) :*: t)) x r class MemApp bf t x r | bf t x -> r instance MemApp HTrue t x HTrue instance Apply (Member t) x r => MemApp HFalse t x r -- we avoid defining a new class like MemApp above. -- I guess, after Apply, we don't need a single class ever? data MemCase2 h t x instance Apply (Member t) x r => Apply (MemCase2 h t x) HTrue r instance Apply (Member ((AllOf h) :*: t)) x r => Apply (MemCase2 h t x) HFalse r testm1 = apply (undefined::Member Fractionals) (1::Float) testm2 = apply (undefined::Member Fractionals) (1::Int) testm3 = apply (undefined::Member Fractionals) () -- The standard HList stuff, extracted from HList library data HNil = HNil data a :*: b = a :*: b infixr 5 :*: data HTrue data HFalse data Z = Z newtype S n = S n class TypeCast a b | a -> b, b->a where typeCast :: a -> b class TypeCast' t a b | t a -> b, t b -> a where typeCast' :: t->a->b class TypeCast'' t a b | t a -> b, t b -> a where typeCast'' :: t->a->b instance TypeCast' () a b => TypeCast a b where typeCast x = typeCast' () x instance TypeCast'' t a b => TypeCast' t a b where typeCast' = typeCast'' instance TypeCast'' () a a where typeCast'' _ x = x class TypeEq x y b | x y -> b instance TypeEq x x HTrue instance TypeCast HFalse b => TypeEq x y b -- A heterogeneous apply operator class Apply f a r | f a -> r where apply :: f -> a -> r apply = undefined -- Normal function application instance Apply (x -> y) x y where apply f x = f x

2 1

HSFFIG compilation problem
by Harri Kiiskinen 14 May '08

14 May '08

Hello, I've tried to compile hsffig-1.0pl2, unsuccesfully, also from the unstable darcs repo. I'm using ghc-6.8 (ghc6-6.8.2 from Debian unstable). HSFFIG comes with its own version of Cabal, but I cannot get past the configuration phase (./cabal-setup configure), when I get the following output: Configuring HSFFIG-1.0... configure: searching for ghc in path. configure: found ghc at /usr/bin/ghc configure: looking for package tool: ghc-pkg near compiler in /usr/bin/ghc configure: found package tool in /usr/bin/ghc-pkg configure: Using install prefix: /usr/local configure: Using compiler: /usr/bin/ghc configure: Compiler flavor: GHC configure: Compiler version: 6.8.2 configure: Using package tool: /usr/bin/ghc-pkg configure: Using haddock: /usr/bin/haddock configure: Using happy: /usr/bin/happy configure: Using alex: /usr/bin/alex configure: Using hsc2hs: /usr/bin/hsc2hs configure: Using cpphs: /usr/bin/cpphs configure: Reading installed packages... configure: Dependency base-any: using base-3.0.1.0 cannot satisfy dependency text-any Does anyone have an idea, how the dependency 'text-any' could be satisfied? Harri K.

2 1

Analog data acquisition
by Tom Nielsen 13 May '08

13 May '08

Hello, I would like to use a lazy, purely functional language to create an experiement description (and execution!) language for cellular neuroscience, i.e. electrical recordings and stimulation. Unfortunately, this means I have to talk to a Analog-to-Digital/Digital-to-Analog converter board, with a precision of about 16 bit at a rate of 50 kHz. I currently have a National Instruments M-series board, but would be happy to try another board. I'm not looking for real-time control right now, but that might be of interest in the future. Has anyone looked at creating bindings to the NI-DAQmx drivers or the open-source COMEDI project, or similar hardware? Would anyone be interested in helping out with this driver binding? Regards Tom

3 3

Trying to avoid duplicate instances
by Eric Stansifer 13 May '08

13 May '08

I am using a bunch of empty type classes to categorize some objects: > class FiniteSolidObject o > class FinitePatchObject o > class InfiniteSolidObject o ... > instance FiniteSolidObject Box > instance FiniteSolidObject Sphere > instance FinitePatchObject Mesh > instance InfiniteSolidObject Plane ... and would like to write a function elsewhere with a type signature like: > intersection :: (SolidObject o1, SolidObject o2) => o1 -> o2 -> Intersection o1 o2 These classes are not exposed to the user (which will be me, but that's besides the point) so the the user (me) does not accidentally try to intersect two meshes, etc. So long as I define my instances correctly the compiler will verify this at compile time. Since "solid objects" are exactly "finite solid objects" plus "infinite solid objects", there is an obvious way to code up this logical relationship. So I try to write: > class SolidObject o > instance FiniteSolidObject o => SolidObject o > instance InfiniteSolidObject o => SolidObject o but inevitably GHC complains that I have duplicate instance declarations. I had mistakenly thought that GHC would only give such an error if it could exhibit a specific type 'o' which satisfied both the context 'FiniteSoildObject o' and the context 'InfiniteSolidObject o' (and I know that there does not exist any such 'o' because I define my instances such that 'FiniteSolidObject' and 'InfiniteSolidObject' will never coincide, and I do not export any of these type classes), but I see now that this is not the case. Adding flags like -fallow-overlapping-instances doesn't convince GHC. Is there a way around this other than manually writing out all the instances I want? That is, > instance FiniteSolidObject Box > instance FiniteObject Box > instance SolidObject Box > instance UnionableObject Box > instance Object Box > > instance FiniteSolidObject Sphere ... Thanks, Eric

2 1

Instance of Eq on cyclic data structure?
by Leonard Siebeneicher 13 May '08

13 May '08

Hi. I am testing a winged edge data structure. > module Main where > > type Point3 = (Float, Float, Float) > > data Body= Body [Vertex] [Edge] [Face] deriving(Eq) > data Face= Face Edge deriving(Eq) > data Edge= Edge (Vertex, Vertex) > (Face, Face) > (Edge, Edge) > (Edge, Edge) deriving(Eq) > data Vertex= Vertex Point3 Edge deriving(Eq) ... > --- Face, Edge, Vertex: Conversion > > getVertsFromFace :: Face -> [Vertex] > getVertsFromFace (Face startEdge) = process_wing_clockwise startEdge > where > process_wing_clockwise (Edge > (vp, _) > _ > (_, eWing1) > _) > | startEdge == eWing1 = > [vp] > | otherwise = > vp : (process_wing_clockwise eWing1) > > > --- Misc procedures > printPointList :: [Vertex] -> IO () > printPointList [] = return () > printPointList ((Vertex p _ ):vs) = putStr (show p) >> > printPointList vs > > --- main > main = > do > let q = createQuad (0,0,0) (0,1,0) (1,1,0) (1,0,0) > let f = (\(Body _ _ (face:faces)) -> face) q > let verts = getVertsFromFace f > printPointList verts > > Here I get a error message. > (0.0,0.0,0.0)(0.0,1.0,0.0)(1.0,1.0,0.0)runhugs: Error occurred > > ERROR - Control stack overflow > I think this come because the derived Eq-Instance "==" get into an infinite loop Alternatively I tried following > --- We want to identify each Body, Face, Edge or Vertex by a number ID > type ID = Int --- A special number for identifying, used for Eq > > data Body = Body ID [Vertex] [Edge] [Face] > data Face = Face ID (Edge) > data Edge = Edge ID > (Vertex , Vertex) > (Face , Face) > (Edge , Edge) > (Edge , Edge) > data Vertex = Vertex ID Edge > > --- Eq Instances > instance Eq Vertex) where > (Vertex i1 _) == (Vertex i2 _) = i1 == i2 > > instance Eq Face where > (Face i1 _ ) == (Face i2 _) = i1 == i2 > > instance Eq Edge where > (Edge i1 _ _ _ _) == (Edge i2 _ _ _ _) = i1 == i2 > > > instance Eq (Body) where > (Body i1 _ _ _) == (Body i2 _ _ _) = i1 == i2 ... This way my code does not hang up. But I have to manage those ugly ID's. Is there a better way to create instances of class Eq, without something like ID? Thanks for reading. Best regards, Leonard > > > -------------------- > -------------------- > ---- Begin ---- > -------------------- > module Main where > > type Point3 = (Float, Float, Float) > > data Body= Body [Vertex] [Edge] [Face] deriving(Eq) > data Face= Face Edge deriving(Eq) > data Edge= Edge (Vertex, Vertex) > (Face, Face) > (Edge, Edge) > (Edge, Edge) deriving(Eq) > data Vertex= Vertex Point3 Edge deriving(Eq) > > {- > implementing simple generative modelling > -} > > --- elementar object generation > createQuad :: Point3 -> > Point3 -> > Point3 -> > Point3 -> > Body > > createQuad p0 p1 p2 p3 = > let > {faceFront = Face edge0 > ;faceBack = Face edge2 > ;vert0 = Vertex p0 edge0 > ;edge0 = Edge > (vert0, vert1) > (faceFront, faceBack) > (edge3, edge1) > (edge1, edge3) > ;vert1 = Vertex p1 edge1 > ;edge1 = Edge > (vert1, vert2) > (faceFront, faceBack) > (edge0, edge2) > (edge2, edge0) > ;vert2 = Vertex p2 edge2 > ;edge2 = Edge > (vert2, vert3) > (faceFront, faceBack) > (edge1, edge3) > (edge3, edge1) > ;vert3 = Vertex p3 edge3 > ;edge3 = Edge > (vert3, vert0) > (faceFront, faceBack) > (edge2, edge0) > (edge0, edge2) > } > in > Body [vert0, vert1, vert2, vert3] [edge0, edge1, edge2, edge3] > [faceFront, faceBack] > > --- Face, Edge, Vertex: Conversion > > getVertsFromFace :: Face -> [Vertex] > getVertsFromFace (Face startEdge) = process_wing_clockwise startEdge > where > process_wing_clockwise (Edge > (vp, _) > _ > (_, eWing1) > _) > | startEdge == eWing1 = > [vp] > | otherwise = > vp : (process_wing_clockwise eWing1) > > > --- Misc procedures > printPointList :: [Vertex] -> IO () > printPointList [] = return () > printPointList ((Vertex p _ ):vs) = putStr (show p) >> > printPointList vs > > --- main > main = > do > let q = createQuad (0,0,0) (0,1,0) (1,1,0) (1,0,0) > let f = (\(Body _ _ (face:faces)) -> face) q > let verts = getVertsFromFace f > printPointList verts > ----------------- > ---- End ----

2 1