- Haskell-Cafe - Haskell.org

Defining Haskell using "arrows"
by Andy Elvey 06 Nov '04

06 Nov '04

Hi all - I'm a first-timer here, and am *very* much attracted by Haskell's elegance and power ... :-) I have only poked around briefly with Haskell so far (at the "hello world" level). One thing that I have come across, and which really got me thinking, was the page on the Haskell website about "arrows". They really seem to take elegance and the "generic-ness" of algorithms to a new level. I was thinking - arrows are a relatively recent concept. Haskell's grammar (as defined in the Haskell report) has now been around for quite some time. So - I thought "what if?" What if Haskell were to be defined (er, *re*-defined) in terms of arrows? In other words, what if it were possible to start totally "from scratch", re-defining Haskell's grammar using arrows? (and preferably, *only* arrows ... ). This would presumably involve somehow coding up the arrows first, and then (haivng got them), using them to create/define the rest of the language. My reason for suggesting this - simply to see if it were possible to do a very small (but still quite comprehensive) Haskell implementation. Small meaning maybe 2-3 MB or so ... As an interesting "benchmark" of this, the Rebol language (though proprietary - sigh!) - comes in at 0.5 Mb or so, including a GUI. So, "very small but very powerful" is possible (although undoubtedly pretty difficult ... ) As a Haskell newbie, I have to defer to everyone else on this list (in terms of my Haskell ability), but just thought I'd post this idea anyway. I was thinking " I wonder if anyone has thought of this? ....... " . Arrows seem to be a recent "add-on", as it were, so I thought -given their power and flexibility, why not turn things totally around, and put them at the *very core* of the language, using them to define the language itself. Anyway, I'm keen to hear people's thoughts on this! Very many thanks in advance ... :-) - Andy

1 0

mkEmptyWeakPtr?
by David Roundy 06 Nov '04

06 Nov '04

I was wondering whether there might be any sort of trick one could use to create an empty "Weak a", i.e. one that has already been GCed? It should be possible to create such a pointer without having an object of type a. mkEmptyWeakPtr :: IO (Weak a) I know I could get this effect by using instead a "Maybe (Weak a)", but that's ugly and adds an additional layer of redirection. The point is that the weak pointer is being used to cache a value, but I'd like to not fill the cache until I actually need to read it. i.e. I have a data type AntiMemo (IO a) (IORef (Weak a)) where the second term is a cache of the result of the first, which is really a pure function, but I'm using the IO monad to force reevaluation, so the result, which is large, can be discarded and recomputed. So I could change this to AntiMemo (IO a) (IORef (Maybe (Weak a))) but it just seems a shame when Weak already effectively has a "Nothing" value type. I suppose one could also make it a "Weak (Maybe a)", but in any case, I'd rather be able to just initialize the IORef to an empty weak pointer, so there wouldn't need to be any extra code when reading the value. -- David Roundy http://www.abridgegame.org

1 0

Making MVar and Chan Instances of Typeable
by Benjamin Franksen 06 Nov '04

06 Nov '04

Hello Experts, I need MVar and Chan to be instances of Typeable. Any hint on how this is most easily done would be greatly appreciated. I could change the libraries and add 'deriving Typeable' but I hesitate to do so. Cheers, Ben

4 6

Re: [Haskell-cafe] Re: Double -> CDouble, realToFrac doesn't work
by MR K P SCHUPKE 05 Nov '04

05 Nov '04

>My guess is because irrationals can't be represented on a discrete computer Well, call it arbitrary precision floating point then. Having built in Integer support, it does seem odd only having Float/Double/Rational... Keean. ..

2 1

Literate Haskell and piecewise construction of classes/instances
by Graham Klyne 05 Nov '04

05 Nov '04

I'm experimenting with a Literate Haskell style to construct a tutorial about Description Logics [1][2]. I'm developing the material in small sections, and using examples in the document to test the code as I go along . I therefore find that I want to introduce the components of class and instance declarations in pieces, rather than having the entire declaration in a single place. Am I right in thinking this is not possible? For example, I find myself resorting to tricks like this: [[ So far, a single Description Logic (AL) has been considered. My next step is to generalize the function interface to a collection of type classes that can handle arbitrary description logics. The basic ideas of concepts, roles and interpretations are common to all description logics, so these are used as defined above. It is the concept descriptions, role descriptions and associated operations that need to be further abstracted: - class ConceptExpr c where - iConcept :: Ord a => TInterpretation a -> c -> Set a (The full and final definition of ConceptExpr is given later.) : ]] and later... [[ The class interface for concept expressions is therefore extended thus: > class ConceptExpr c where > iConcept :: Ord a => TInterpretation a -> c -> Set a > isSatisfiableWith :: TBox c -> c -> Bool > isSatisfiableWith t c = not (isSubsumedByWith t c emptyClass) > isSubsumedByWith :: TBox c -> c -> c -> Bool > isSubsumedByWith t c d = > not $ isSatisfiableWith t (intersectClass c (complementClass d)) > emptyClass :: c > complementClass :: c -> c > intersectClass :: c -> c -> c ]] What this suggests to me is that for Literate Haskell it is desirable to have multiple appearances of a class/instance declaration, constrained so that one version contains all of the definitions contained in any other appearance, and no two appearances contain differing definitions for any component (using a fairly strict syntactic comparison to determine difference here). Does this strike any chords with anyone? #g -- [1] http://dl.kr.org/ (Description Logics web page) [2] http://www.ninebynine.org/Software/HaskellDL/DLExploration.lhs (My tutorial, work-in-progress - implements "structural subsumption" for a simple Description Logic.) ------------ Graham Klyne For email: http://www.ninebynine.org/#Contact

1 0

Re: [Haskell-cafe] Making MVar and Chan Instances of Typeable
by MR K P SCHUPKE 05 Nov '04

05 Nov '04

My mistake: instance Typeable a => Typeable (MVar a) where typeOf (x::MVar x) = mkAppTy (mkTyCon "Control.Concurrent.MVar.MVar") [typeOf (undefined::x)] Keean.

1 0

Re: [Haskell-cafe] Re: Double -> CDouble, realToFrac doesn't work
by Benjamin Franksen 05 Nov '04

05 Nov '04

On Friday 05 November 2004 14:11, you wrote: > It's worse: Since according to IEEE +0 is not equal to -0, atan2 is not a > function! Sorry, I meant to write: Since according to IEEE +0 *is* to be regarded as equal to -0, atan2 is not a function. (Because it gives different values for argument combinations -0, +0.) Ben

2 1

Re: [Haskell-cafe] Making MVar and Chan Instances of Typeable
by MR K P SCHUPKE 05 Nov '04

05 Nov '04

nstance Typeable a => Typeable (MVar a) where typeOf (x::x) = mkAppTy (mkTyCon "Control.Concurrent.MVar.MVar") [typeOf (undefined::x)] Keean.

2 1

Re: Double -> CDouble, realToFrac doesn't work
by dpt＠lotus.bostoncoop.net 05 Nov '04

05 Nov '04

On Thu, Nov 04, 2004 at 08:32:52PM +0100, Sven Panne wrote: > It's an old thread, but nothing has really happened yet, so I'd like to > restate and expand the question: What should the behaviour of toRational, > fromRational, and decodeFloat for NaN and +/-Infinity be? Even if the report > is unclear here, it would be nice if GHC, Hugs, and NHC98 agreed on > something. > Can we agree on the special Rational values below? I would be very careful of adding non-rationals to the Rational type. For one thing, it breaks the traditional rule for equality a % b == c % d iff a*d == b*c You'd need to look at all the instances for Ratio a that are defined. For instance, the Ord instance would require at least lots of special cases. And when would you expect 'x/0' to give +Infinity and when -Infinity? For IEEE floats, there are distinct representations of +0 and -0, which lets you know when you want which one. But for the Rational type there is no such distinction. The behaviour that '1 % 0' gives the error 'Ratio.% : zero denominator' is clearly specified by the Library Report. In the meantime, there are utility functions for dealing with IEEE floats (isNaN, etc.) Peace, Dylan

1 0

Newbie Question on type constructors
by Brian Beckman 04 Nov '04

04 Nov '04

This is a tiny question on the "data" syntax. Consider the following (cribbed from Paul Hudak's "School of Expression" book, page 22): data Shape = Circle Float | Square Float I read this something along the lines of "'Shape' is a type constructor, for use in other type-defining expressions, and 'Circle' and 'Sqare' are its two data constructors, which should be used like functions of type 'Float -> Shape'". Indeed, typing "Circle" at the Hugs prompt reveals that Haskell has a "function" named "Circle" with type "Float -> Shape." However, I don't know of other circumstances where (1) I can declare functions with capitalized names (Hugs groans about syntax errors if I attempt the following: Circle2 :: Float -> Shape Circle2 = Circle And (2) where the argument-types of a function can be declared on the function's right-hand side. So, in some sense, am I right to think of "data" as a special syntax for special kinds of function declarations, namely data-constructor function declarations, and that this syntax is different in appearance, but not terribly different in meaning, from the ordinary syntax, typified by circle2 :: Float -> Shape circle2 = Circle (this is just the uncialized version of the one above, and is perfectly legal) Apologies if this is just too nitpicky and pedantic for words, but I try to notice everything.

8 13