October 2008 - Haskell-Cafe

Haskell on mobile platforms
by Janis Voigtlaender 13 Oct '08

13 Oct '08

Hi all, there were several discussions on this list about different implementations of Haskell running on the iPhone and other mobile platforms over the last three months or so. I would very much like to see the current state of what is and is not yet possible here reflected in the "Platforms" section of the upcoming edition of the Haskell Communities and Activities report. So please, if you have been working on getting Haskell to run on your phone, and even more so if you were successful, write a short HCAR entry about it. For further details, see: http://article.gmane.org/gmane.comp.lang.haskell.cafe/45967 Thanks, Janis. -- Dr. Janis Voigtlaender http://wwwtcs.inf.tu-dresden.de/~voigt/ mailto:voigt@tcs.inf.tu-dresden.de

1 0

the ghc "reflection" thing?
by Galchin, Vasili 13 Oct '08

13 Oct '08

hello, Several months ago I saw on the wiki or maybe it was a discussion on mechanism to get the ghc compiler's state. I can't remember enough to ask even well. I know there is a wiki entry. Sorry ... I can only "hint" at this ... ?? Thanks, Vasili

2 2

Constraints at construction
by Iain Barnett 12 Oct '08

12 Oct '08

If I were to create an object in C#, for instance, I could add code to the constructor that might limit the type further e.g. public class Car { string model; int wheels; public Car ( string model, int no_of_wheels ) { if ( no_of_wheels <= 2 ) { throw new Exception( "a car must have at least 3 wheels"); } this.model = model; this.no_of_wheels = no_of_wheels; } } or I could specify all sorts of things, like the size of an array could only be a certain length, or whatever. Similarly, when creating a column with SQL I'd be able to specify further constraints on the column than just it's type. I don't see how to do this with Haskell and the data keyword. Is there a tutorial someone could point me to, or an explanation of this? Or is it that I have to wrap the creation in an accessor function that checks the inputs first? type Model = String type Wheels = Int data Car = Car Model Wheels car_maker model wheels = if wheels <= 2 then error ... else Car model wheels ? Most of my programming career (if you can call it that:) has been primarily with C# and SQL, and perhaps this heady mixture of OOP and hacked-functional is confusing me! Iain

5 6

Re: [Haskell] Abusing quickcheck to check existential properties
by Sterling Clover 12 Oct '08

12 Oct '08

My understanding is that SmallCheck and lazy SmallCheck[1] are much better suited to handle existential properties, and indeed have combinators especially for that task. --Sterl. [1] http://www-users.cs.york.ac.uk/~mfn/lazysmallcheck/ smallcheck.pdf, http://hackage.haskell.org/cgi-bin/hackage-scripts/ package/smallcheck, http://hackage.haskell.org/cgi-bin/hackage- scripts/package/lazysmallcheck On Oct 12, 2008, at 1:38 PM, Norman Ramsey wrote: > I recently used QuickCheck to check on some calculations for image > compression. > (I love exact rational arithmetic!) But I thought only to check for > inverse properties, and I realized afterward I had failed to check for > ranges. For example I should have checked that > > boundedB block = -1 <= b && b <= 1 > where Cos a b c d = cos_of_block block > > which turns out to be correct. But actually my calculations were > wrong, and the real bounds on b are not +/- 1 but rather +/- 0.5, so I > might equally well have passed a more stringent test. It's quite > embarrassing as I have only 5 bits of precision to represent b, and > throwing away a bit on values that don't exist is not the best idea I > have had recently. > > What I would really like to do is write some combinators for interval > checking that say > > * It is a *universal* property that for *every* input, the output > lies within the stated interval. > > * It is an *existential* property of *some* input that the output > lies close to the ends of the stated interval. > > But how do I use QuickCheck to check an existential? > I realize I could run > > boundsNotAchieved block = -0.9 <= b && b <= 0.9 > where Cos a b c d = cos_of_block block > > and then if the test of boundsNotAcheived fails, my test passes > (by exists x : P(x) iff not forall(x) : not P(x)). > > But if I set up a test suite in which some tests are supposed to fail > and others are supposed to succeed, I will never keep track of which > is which. I would like to build an existential test. Is it > sufficient for me to write > > boundsAchieved block = expectFailure (-0.9 <= b && b <= 0.9) > where Cos a b c d = cos_of_block block > > and then run 'quickCheck boundsAchieved'? > > In the long run I'd love something like > > fillsIntervalWithin :: (Num a, Ord a, Arbitrary a) => (a, a) -> a > -> a -> Property > > with the idea that I want to test the partial application > > fillsIntervalWithin (lo, hi) epsilon > > and have the test succeed if every input x satisfies lo <= x <= hi > and if some input x satisfies not (lo + epsilon <= x <= hi - epsilon). > > I'm betting someone on this list has already thought about similar > problems and can advise me. (Because I have yet to write the > specialized instance declaration for Arbitrary that guarantees the > numerical invariants of the inputs, I have yet to test any of the > examples in this email.) > > > Norman > _______________________________________________ > Haskell mailing list > Haskell(a)haskell.org > http://www.haskell.org/mailman/listinfo/haskell

1 0

Arch Haskell News: Oct 12 2008
by Don Stewart 12 Oct '08

12 Oct '08

News about Haskell on Arch Linux http://archhaskell.wordpress.com/2008/10/12/arch-haskell-news-oct-12-2008/ Arch now has 624 Haskell packages in AUR. That’s an increase of 15 new packages in the last 7 days. Planning is also underway for the GHC 6.10 release, and the 6.10 release candidate is in testing. Noteworthy * cabal-install-0.6.0: “The command-line interface for Cabal and Hackage.” * haskell-cabal-1.6.0.1: “A framework for packaging Haskell software” * haskell-bytestring-0.9.1.3: “Fast, packed, strict and lazy byte arrays with a list interface” * haskell-transactional-events-0.1.0.0: “Transactional events, based on Concurrent ML semantics” Full details on the site.

1 0

Could not deduce .. why ?
by Marc Weber 12 Oct '08

12 Oct '08

Surely I've overseen a small point. But I can't see it: Why doesn't ghc recognize taht elc_ (class decl), elc (instance decl) and elc_1 are the same type To help you find the connecting pieces faster I've marked them with ## ## Marc data PT a b = PT b -- phantom type containing state a and the result b class AddEl el_ el2_ elc_ where addEl :: el_ -> elc -> el2_ -- el, child -- ========== adding sub elements (tags) ============================= class AddElT est el_ estc ##elc_## est2 el2_ | estc est -> est2 , est est2 estc el2_ -> el_ , est est2 estc el2_ -> elc_ where addElT :: PT est el_ -> PT estc ##elc_## -> PT est2 el2_ -- first child ? attrs ok? instance ( AddEl el el2 ##elc## <-- this is not recognized because elc != elc_1 , Consume st (Elem celType) st' , DetermineElAddEl (NYV (Element elType AttrsOk st HFalse)) el (Valid celType) elc (NYV (Element elType AttrsOk st' HTrue)) el2 ) => AddElT (NYV (Element elType AttrsOk st HFalse)) el (Valid celType) ##elc## <-- elc_ from class declaration (NYV (Element elType AttrsOk st' HTrue)) el2 where addElT (PT t) (PT ##c##) = PT $ addEl t c -- <<<<<<<<< line 435, c should have type elc, not elc_1 src/Text/XML/Validated/Types.hs|435 col 32 error| || Could not deduce (AddEl el el2 elc_1) || from the context (AddElT || (NYV (Element elType AttrsOk st HFalse)) || el || (Valid celType) || elc || (NYV (Element elType AttrsOk st' HTrue)) || el2, || AddEl el el2 elc, || Consume st (Elem celType) st', || DetermineElAddEl || (NYV (Element elType AttrsOk st HFalse)) || el || (Valid celType) || elc || (NYV (Element elType AttrsOk st' HTrue)) || el2) || arising from a use of `addEl' || at src/Text/XML/Validated/Types.hs:435:32-40 || Possible fix: || add (AddEl el el2 elc_1) to the context of the instance declaration || In the second argument of `($)', namely `addEl t c' || In the expression: PT $ addEl t c || In the definition of `addElT': || addElT (PT t) (PT c) = PT $ addEl t c

2 1

How to translate Haskell to other languages?
by Jason Dagit 12 Oct '08

12 Oct '08

Hello, I was thinking about translating Haskell to other languages, python being the main one at the moment. Here is my attempt at manually encoding Haskell in Python: \begin{code} import types class thunk: '''Thunks allow us to delay a computation and they also store their value inside themselves once they have been accessed.''' def __init__(self, v): self.v = v def value(self): '''Force the thunk to be calculated by referencing it.''' while self.isReducible(): self.reduce() return self.v def reduce(self): '''Reduces the thunk, by either calling the represented function or reducing the layers of thunk.''' if (type(self.v) == types.FunctionType): self.v = self.v() else: self.v = self.v.value() return self.v def isReducible(self): '''Returns True when the thunk is still callable.''' return isinstance(self.v, thunk) or \ type(self.v) == types.FunctionType class nil: '''Empty list element''' def __init__(self): pass class cons: '''Non-empty lists''' def __init__(self, head, tail): self.head = head self.tail = tail '''Unpack the cons cell''' def uncons(self): return self.head, self.tail def htail(t): '''This function works like Haskell's tail function.''' l = t.value() x, xs = l.uncons() return xs def plus(t1, t2): '''Adds numbers''' i1 = t1.value() i2 = t2.value() return thunk(i1+i2) def zipWith(f, t1, t2): '''This is like Haskell's zipWith function.''' l1 = t1.value() if isinstance(l1, nil): return thunk(nil()) l2 = t2.value() if isinstance(l2, nil): return thunk(nil()) x, xs = l1.uncons() y, ys = l2.uncons() zw = thunk(lambda: zipWith(f, xs, ys)) fxy = thunk(lambda: f(x,y)) return thunk(cons(fxy, zw)) def fibs(): '''This is the classic fibs: fibs = 1 : 1 : zipWith (+) fibs (tail fibs)''' f1 = thunk(1) f2 = thunk(1) fn = thunk(fibs) rest = thunk(lambda: zipWith(plus, fn, htail(fn))) restlist = thunk(cons(f2, rest)) fiblist = thunk(cons(f1, restlist)) return fiblist def hmap(f, t): '''map _ [] = [] map f (x:xs) = f x : map f xs''' l = t.value() if isinstance(l, nil): return thunk(nil()) x, xs = l.uncons() fx = thunk(lambda: f(x)) mapfxs = thunk(lambda: hmap(f, xs)) return thunk(cons(fx, mapfxs)) def show(t): '''show :: a -> String''' v = t.value() return thunk(str(v)) def printList(t): '''This just gives us a way to debug lists.''' v = t.value() print "[", while True: h,t = v.uncons() print "%s" % h.value(), if isinstance(t.value(), nil): break else: print ", ", v = t.value() print "]" def take(tn, tl): '''take n _ | n <= 0 = [] take _ [] = [] take n (x:xs) = x : take (n-1) xs''' n = tn.value() if n <= 0: return thunk(nil()) l = tl.value() if isinstance(l, nil): return thunk(nil()) x,xs = l.uncons() nminusone = thunk(lambda: plus(tn, thunk(-1))) takerec = thunk(lambda: take(nminusone, xs)) return thunk(cons(x, takerec)) \end{code} You can try this out in python with: tenfibs = take(thunk(10), fibs()) printList(tenfibs) This will print the first 10 fibs. Questions: I think the examples above are correctly lazy. Have I missed something? I noticed my thunks can get wrapped in each other, is this to be expected, or am I doing it wrong? Is there an easier encoding using generators? When I started I was using generators instead of thunk, but I found it was complicating my design so I removed it. And yet, since generators are python's version of thunks, it seems like there should be a more natural encoding there. I'm not explicitly using a graph reduction algorithm to reach WHNF, does this mean my translation is wrong? Are there some well known test cases I should try? Anyone know of a paper that discusses making this translation? I am trying to avoid writing an interpreter in Python for Haskell. My goal is to translate Haskell functions into the equivalent Python. I'm also hoping to avoid needing a G-machine. Thanks! Jason

6 6

Shooting your self in the foot with Haskell
by John Van Enk 12 Oct '08

12 Oct '08

There's the well known "How to shoot your self in the foot" list which I have it printed and taped on my desk at work. http://www-users.cs.york.ac.uk/susan/joke/foot.htm I had a co-worker ask me how you'd shoot your self in the foot with Haskell. Here's the best I could do: "You shoot the gun, but the bullet gets trapped in the IO monad." I fully realize how un-clever this is. Some one please give me something more worth of the original list. :) -- /jve

18 24

Real World Haskell Chapter 11 ... Testing
by Galchin, Vasili 12 Oct '08

12 Oct '08

Hi, Testing of "pure" code ... . way cool ... I am sure there literature from the first order logic, model theory, categorical logic viewpoint?? Kind regards, Vasili

1 0

[] vs [()]
by Daryoush Mehrtash 11 Oct '08

11 Oct '08

What is the difference between empty list [] and list with one unit element [()]? daryoush

10 12