July 2007 - Haskell-Cafe

AngloHaskell 2007, dates and venue confirmed
by Neil Mitchell 15 Jul '07

15 Jul '07

Hi, We are pleased to announce AngloHaskell 2007 http://www.haskell.org/haskellwiki/AngloHaskell Dates: 10th-11th of August (Friday-Saturday) Location: Cambridge, with talks at Microsoft Research on Friday All the details are on the wiki page, along with free registration. Everyone is invited, we will have a day of talks at MSR, then a day of other activities. There will be plenty of chance for general discussions on anything. If anyone in Cambridge is able to accommodate a few people for the Friday or Saturday night, please add your name to the wiki, and accept our thanks in advance. All that is needed is floor space. Thanks Neil and Philippa

4 5

Re: [Haskell-cafe] Maintaining the community
by Andrea Rossato 15 Jul '07

15 Jul '07

On Sun, Jul 15, 2007 at 07:05:53PM +1000, Donald Bruce Stewart wrote: > So how would you think we can approve? We have to help in more specific > ways, and listen more carefully to what people are asking? I usually do this when I want a newcomer to join my community (for instance my research community): I study him, I read his stuff, I try to get to know him, so that I can find for him a place that can suit his needs, feed his curiosity, and help him improve while maximizing his contribution to the community. I treat him as she were my fellow researcher, and not a student I'm supposed to teach to. Just an example: I find it so much irritating when you give a paper to someone and, before even reading it, the receiver starts asking you if you have already read the just published papers of authors x and y about a topic that could slightly resemble the one of your paper (even if actually does not)? This a the kind of behaviour I always try to avoid with new comers. I suppose they know more then me, even if in different area of knowledge. We are all experts of almost nothing. But I'm talking about a small research community, with one or two new members a year. I have no experience in managing such a large community as this one. I just thought that the fact that I felt repelled by the behaviour of some of the members of this community, and please believe me when I say that that behaviour was far from being intentionally offensive or even disrespectful in my regards, could be useful for someone that has such a difficult task. All my best wishes. Andrea

1 0

[weird stuff] The Dodgy Diagonal
by Conor McBride 14 Jul '07

14 Jul '07

Scary words warning: Polynomial, Functor, Bifunctor, unsafeCoerce# Folks A peculiar query for folks who know more about the internals of Haskell compilers than I do. I attach the full code with all the bits and pieces, but let me pull out the essentials in order to state the problem. I've been mucking around with polynomial types in one parameter > newtype Id x = Id x -- element > newtype K1 a x = K1 a -- constant > data ((Sum1 p q)) x = L1 (p x) | R1 (q x) -- choice > data ((Prod1 p q)) x = P1 (p x) (q x) -- pairing all of which give unremarkable instances of Functor < class Functor p where < fmap :: (s -> t) -> p s -> p t > instance Functor Id > instance Functor (K1 a) > instance (Functor p, Functor q) => Functor (Sum1 p q) > instance (Functor p, Functor q) => Functor (Prod1 p q) I've also been mucking around with polynomial types in two parameters > newtype Fst x y = Fst x > newtype Snd x y = Snd y > newtype K2 a x y = K2 a > data ((Sum2 p q)) x y = L2 (p x y) | R2 (q x y) > data ((Prod2 p q)) x y = P2 (p x y) (q x y) which give unremarkable bifunctors, doing two maps at once > class Bifunctor p where > bimap :: (s1 -> t1) -> (s2 -> t2) -> p s1 s2 -> p t1 t2 > instance Bifunctor Fst where > bimap f g (Fst x) = Fst (f x) > > instance Bifunctor Snd where > bimap f g (Snd y) = Snd (g y) > > instance Bifunctor (K2 a) > instance (Bifunctor p, Bifunctor q) => > Bifunctor (Sum2 p q) > instance (Bifunctor p, Bifunctor q) => > Bifunctor (Prod2 p q) Now, I'm interested in collapsing the diagonal. What? Er, this: > class (Bifunctor b, Functor f) => Diag b f | b -> f where > diag :: b x x -> f x > gaid :: f x -> b x x If the two parameters to a bifunctor are instantiated with the same thing, we should be able to exchange with the functorial representation. I'll just do one way. > instance Diag Fst Id where > diag (Fst x) = Id x > instance Diag Snd Id where > diag (Snd x) = Id x > instance Diag (K2 a) (K1 a) where > diag (K2 a) = K1 a > instance (Diag pb pf, Diag qb qf) => Diag (Sum2 pb qb) (Sum1 pf qf) where > diag (L2 p) = L1 (diag p) > diag (R2 q) = R1 (diag q) > instance (Diag pb pf, Diag qb qf) => Diag (Prod2 pb qb) (Prod1 pf qf) where > diag (P2 p q) = P1 (diag p) (diag q) That looks like a whole lot of doing very little. So, can I (in practice, in this or that compiler) get away with... > dodgy :: Diag b f => b x x -> f x > dodgy = unsafeCoerce# > ygdod :: Diag b f => f x -> b x x > ygdod = unsafeCoerce# ...dodgy for diag and ygdod for giad? Minimal nontrivial experiments in ghc give grounds for cautious optimism, but I'd be delighted to hear from better informed sources. Cheers Conor ------------------

4 3

Troubles with Exercise 4.9 from YAHT
by michael＠schmong.org 14 Jul '07

14 Jul '07

from YAHT Exercise 4.9 Write a function elements which returns the elements in a BinaryTree in a bottom-up, left-to-right manner (i.e., the ï¿½ï¿½ï¿½rst element returned in the left-most leaf, followed by its parentï¿½ï¿½ï¿½s value, followed by the other childï¿½ï¿½ï¿½s value, and so on). The re- sult type should be a normal Haskell list. here's the data type BinaryTree data BinaryTree a = Leaf a | Branch (BinaryTree a) a (BinaryTree a) Here's my first attempt nodeCount :: BinaryTree a -> [b] nodeCount = map putStrLn (show a) a is out scope. I started looking at what types putStrLn and map are. First thing that occurs to me is that I can't use map unless I pass it a list. I'm not allow to use fold yet, that's the next assignment. Any clues on how to proceed?

2 1

Functional dependencies *not* part of the next Haskell standard?
by peterv 14 Jul '07

14 Jul '07

I tried to do something in CAL that I could not solve without functional dependencies. In their support forum, it got mentioned that func.deps propably won't make into the next Haskell standard... Any comments on that? Now, the thing I tried to solve was: data Vector2 a = Num a => V2 a a class Vector a n | a -> n where dot :: a -> a -> n instance Num a => Vector (Vector2 a) a where dot (V2 x1 y1) (V2 x2 y2) = x1 * x2 + y1 * y2 test1 = dot (V2 1.0 2.0) (V2 3.0 4.0) Please note this is just some testing code (the math people in here might be horrified by it ;), although any hints of how to do this in a clearer way are welcome. As side-note, does Haskell have some easy to use good linear algebra package for doing 2D, 3D and 4D math? Then I would not be writing any of this code, although I did find it useful for learning. Now without the funcdep a -> n in the type class declaration I get No instance for (Vector (Vector2 t) n) arising from use of `dot' at l:/Haskell/test/vector.hs:9:8-36 Possible fix: add an instance declaration for (Vector (Vector2 t) n) In the expression: dot (V2 1.0 2.0) (V2 3.0 4.0) In the definition of `test1': test1 = dot (V2 1.0 2.0) (V2 3.0 4.0) CAL seems to have the same problem, and without funcdeps I seem to be stuck in that language. PS(1): In the above I was using GHCI with extensions enabled. PS(2): IMHO CAL is in some aspects easier to get started than Haskell because of their nice Eclipse plugin and GEM Cutter environment for doing learning experiments. Thanks, Peter

15 29

Looking for final year project - using Haskell, or another functional language
by wp 14 Jul '07

14 Jul '07

Hi all, I will soon be doing my last year in computer science. One part of our last year encompasses quite a big project which will go over 3 terms and will account for 3 modules (45 credits). I was thinking in doing something using functional languages (preferably Haskell, as it's the one I know most). Does anybody know anyone who would have a task suitable for such a project which would encompass the whole development life cycle (maybe a sub-project?). I would do this obviously for free; the client can be anyone (industrial, academic, open source, etc. ... ), as long as the project is something serious and for practical usage. I would be happy for any suggestions ... Thanks walter.

9 17

Re: [Haskell-cafe] Type system madness
by Andrew Coppin 14 Jul '07

14 Jul '07

Bryan O'Sullivan wrote: > Andrew Coppin wrote: > >> Oh well, the problem is easily fixed... *sigh* > > I doubt that anybody minds having you talk about Haskell. You've been > responsible for spawning a lot of interesting threads. [And that one about compression that's still going on somewhere... lol!] > All I would suggest is that you take your cue from the other people > who post to the list, and try a few tactics before you post: > > - If you have a chatty one-line comment, do 2,000 other people need to > see it? Mmm. This is why I prefer NNTP. (If a thread becomes tangent, everybody just marks it "ignore" and they don't have to waste time downloading it or reading it.) But yeah, point taken... > - If you have a question to ask, try to spend 2 minutes with Google or > the Haskell wiki to find the anwer. I continue to be surprised at the things that don't seem to be on the Wiki... Google is typically no help at all with anything Haskell-related, because Haskell is so completely obscure. The various haddoc documentation is also frustratingly sparse in places. (E.g., Control.Concurrent.STM.TVar contains *nothing* but terse type signatures. And concurrent programming is already a tricky thing to get right.) Some things seem to be "well known" yet not actually written down anywhere - e.g., the finer points of using "seq" to make stuff go faster. But sure, I do like to try to puzzle a thing out first before posting here. (If something else, I kind of enjoy a challenge...) > - Join us on #haskell on IRC. It's extremely chatty, and you'll be > welcome. Not in my experience, no. (Maybe I ask the wrong way... but almost everybody seems to simply ignore me. Actually, usually when I go there absolutely nobody is speaking at all. What time zone do these people live in?)

4 4

Displaying fractions of a second from a datediff?
by Hugh Perkins 14 Jul '07

14 Jul '07

Hi, Pretty dumb question I know, but..how to I display fractions of a second from a datediff? ie: main = do starttime <- gettime -- do something here that takes a few seconds endtime <- gettime let timediff = diffClockTimes endtime starttime let timediffstring = show( tdSec timediff) ++ " " ++ show( realToFrac(tdPicosec timediff) / 1000000000 ) putStrLn( show(timediffstring) ) return () tdPicosec is negative??? ... but sometimes it is positive... so it's really unclear what the appropriate way to handle this is! Example outputs: "1 -62.0" "0 921.0" ... it could be that the time in seconds in the second example is "0.921", but for the first, is it "1.062" (adding the absolute milliseconds) or is it "0.938" (adding the negative milliseconds)? Tried to figure this out by generating timeDiffs myself then normalizing them, but this doesnt seem to do anything useful ;-) : Prelude> normalizeTimeDiff $ TimeDiff 0 0 0 0 0 0 1000000000000000000000000000000 0000 TimeDiff {tdYear = 0, tdMonth = 0, tdDay = 0, tdHour = 0, tdMin = 0, tdSec = 0, tdPicosec = 10000000000000000000000000000000000}

1 1

Newbie question about tuples
by peterv 14 Jul '07

14 Jul '07

Hi, I have a couple of questions about tuples. Q1) Is it possible to treat a tuple of N elements in a generic way? So instead of writing functions like lift1 e1, lift2 (e1,e2), lift3 (e1,e2,e3) just one function liftN that works on tuples of any length? Q2) (Maybe related to Q1) Can I convert a tuple of length N to a heterogeneous list (using "forall" aka existentially quantified types) and vice versa? Q3) Suppose I want to declare an instance of Num on all tuple types having (Num instances) as elements; is this possible? I tried instance Num a => Num (a,a) where . but this fails I also tried instance Num a => Num ((,) a a) where . but that also fails. I can of course create a new type like newtype Num a => Vector2 a = Vector2 (a,a) and then create an instance for Vector2, but I was wondering if it would be possible without introducing a new type. Thanks, Peter

12 23

Type system madness
by Andrew Coppin 14 Jul '07

14 Jul '07

OK, can somebody explain to me *really slowly* exactly what the difference between an existential type and a rank-N type is? (I couldn't find much of use on the wiki. I have now in fact written some stuff there myself, but since I don't understand it in the first place, I'm having difficulty trying to explain it to anybody else...)

22 73