January 2005 - Haskell

Two lectureships in Nottingham
by Graham Hutton 12 Jan '05

12 Jan '05

Dear all, We are seeking two new lecturers (assistant professors) in Nottingham, England. Further details are available from: http://www.nottingham.ac.uk/hr/vacancies/academic.html#RUB671S The closing date for applications is 31st January 2005. Graham Hutton +---------------------------------------------------------------------+ | Dr Graham Hutton Email : gmh(a)cs.nott.ac.uk | | School of Computer Science and IT Web : www.cs.nott.ac.uk/~gmh | | University of Nottingham | | Jubilee Campus, Wollaton Road | | Nottingham NG8 1BB Phone : +44 (0)115 951 4220 | | United Kingdom Fax : +44 (0)115 951 4254 | +---------------------------------------------------------------------+ This message has been scanned but we cannot guarantee that it and any attachments are free from viruses or other damaging content: you are advised to perform your own checks. Email communications with the University of Nottingham may be monitored as permitted by UK legislation.

1 0

[ICCIMA/ICSE05] Call for Papers: ICCIMA'05
by ICCIMA 2005 12 Jan '05

12 Jan '05

We apologize if this is a duplicate email. This list is used solely for ICCIMA/ICSE. International Conference on Computational Intelligence and Multimedia Applications, August 16-18, 2005 University of Nevada, Las Vegas, USA (<http://www.iccima.org/>www.iccima.org) F I R S T C A L L F O R P A P E R S The International Conference on Computational Intelligence and Multimedia Applications will be held at the University of Nevada, Las Vegas, USA on August 16-18, 2005. The conference will provide an international forum for discussion on issues in the areas of Computational Intelligence and Multimedia for scientists, engineers, researchers and practitioners. ICCIMA'05 is organized jointly with International Conference on Systems Engineering (ICSE'05: <http://www.icseng.info/>www.icseng.info) and the registered participants of ICCIMA'05 will be able to attend ICSEng'05. The conference will include sessions on theory, implementation and applications, as well as the non-technical areas of challenges in education and technology transfer to industry. There will be both oral and poster sessions. Accepted full papers will be included in the proceedings to be published by IEEE CS Press. Selected papers will be published in "International Journal on Computational Intelligence and Applications" published by World Scientific Publishing Company Press. Several well-known keynote speakers will address the conference. Conference Topics Include (but not limited to): Artificial Intelligence Artificial Neural Networks Pattern Recognition Fuzzy Systems Genetic Algorithms Hybrid Systems Intelligent Control Intelligent Databases Knowledge-based Engineering Learning Algorithms Memory, Storage and Retrieval Multimedia Systems Formal Models for Multimedia Interactive Multimedia Multimedia and Virtual Reality Multimedia and Telecommunications Multimedia Information Retrieval Multimedia and Security Multimedia Hardware Multimedia and Algorithms Special Sessions, Pre-Conference Workshops and Tutorial: Proposals for special sessions, pre-conference workshops and tutorials relevant to the conference topics are invited. The deadline for submitting the proposal is Feb 15, 2005. Special Poster Session: ICCIMA'05 will include a special poster session devoted to recent work and work-in-progress. Abstracts are solicited for this session (2 page limit) in camera ready form, and may be submitted up to 30 days before the conference date. They will not be refereed and will not be included in the proceedings, but will be distributed to attendees upon arrival. Students are especially encouraged to submit abstracts for this session. Invited Sessions Keynote speakers (key industrialists, chief research scientists and leading academics) will be addressing the main issues of the conference. Important Dates: Submission of papers received latest on: March 10, 2005 Submission of Papers Papers in English reporting original and unpublished research results and experience are solicited. Electronic submission of papers via www.iccima.info. Visit the web page for more information. Page Limits Papers for refereeing should be double-spaced and must include an abstract of 100-150 words with up to six keywords. Selected papers will have a limit of 6 pages in the proceedings to be published by IEEE. Evaluation Process All submissions will be refereed based on the following criteria by two reviewers with appropriate background. originality significance contribution to the area of research technical quality relevance to ICCIMA 2005 topics clarity of presentation Referees report will be provided to all authors. Check List Prospective authors should check that the following items are attached and guidelines followed while submitting the papers for refereeing purpose. * The paper and its title page should not contain the name(s) of the author(s), or their affiliation * The paper should have attached a covering page containing the following information -title of the paper -author name(s), title, affiliation, mail and e-mail addresses, phone and fax numbers -Conference topic area -up to six keywords * The name, e-mail, phone, fax and postal address of the contact person should be attached to the submission. Contact Information: ICCIMA' 05 Secretariat Department of Electrical and Computer Engineering University of Nevada, Las Vegas 4505 Maryland Parkway, Box 454026 Las Vegas, NV 89154-4026 USA Phone: +1 702 895 4184 Fax: +1 702 895 1115 email: secretary(a)iccima.org URL: http://www.iccima.org/ _______________________________________________ Conf mailing list Conf(a)Mail.Egr.UNLV.EDU http://Mail.Egr.UNLV.EDU/mailman/listinfo/conf

1 0

GADT and constraints [Was: Rebindable syntax for monads and arrows]
by oleg＠pobox.com 11 Jan '05

11 Jan '05

Simon Peyton-Jones wrote: |[ data MyVec a where |[ MkMyVec :: (Foo a) => a -> MyVec a > > f x = case x of > MkMyVec x -> <rhs> > > constraints are always solved *lazily* when inferring, so we'd infer > f :: Foo [a] => MyVec a -> ... > even without any overlap > > but they are solved *eagerly* when checking against a > user-supplied type sig > so when checking > f :: Eq a => MyVec a => ... > we'd accept the program. (If there were overlapping instances, > we'd complain about them.) I can't understand how constraints float at all. In the following code: > class Foo1 a where foo1 :: a -> String > class Foo2 a where foo2 :: a -> String > > instance Foo1 Bool where foo1 _ = "Bool" > instance Foo2 Char where foo2 _ = "Char" > > data U a where > Ub :: a -> U a > Uc :: a -> U a > > foo (Ub a) = foo1 a > foo (Uc a) = foo2 a What is the type of foo? GHC snapshot 6.3.20041106 happily accepts that code, and even infers *M> :t foo foo :: forall a. (Foo2 a, Foo1 a) => U a -> String but that seems odd. In any value of type U a, either Ub alternative is present, or Uc. They can never be both. Then why do we get the constraint that is the intersection of the two constraints rather than the union of them? BTW, when I tried to do > data U a where > Ub :: Foo1 => a -> U a > Uc :: Foo2 => a -> U a I received a message > Couldn't match kind `*' against `k_a1QK -> *' > In the class declaration for `Foo2' which I found quite interesting. Anyway, even though GHC accepts the code and infers the type for foo, the function is unusable. An attempt to evaluate 'foo (Ub True)' ends up in an error message that there is no instance of Foo2 Bool. Indeed, there isn't. But why that should matter if I'm only interested in the Ub alternative? Again, the function 'foo' seems legitimate, and can be easily expressed either in Martin Sulzmann's interpretation of GRDT via existentials, or using the typeclass representation of GADT. For example: > data Ub a = Ub a > data Uc a = Uc a > > class Foo t where foo:: t -> String > > instance Foo1 a => Foo (Ub a) where > foo (Ub a) = foo1 a > instance Foo2 a => Foo (Uc a) where > foo (Uc a) = foo2 a Here, the constraint Foo is clearly the `union' of Foo1 and Foo2 (the discriminated union at that).

2 1

RE: [Haskell] Re: Rebindable syntax for monads and arrows
by Simon Peyton-Jones 11 Jan '05

11 Jan '05

| You can to do this with GADTs: | | data MyVec a where | MkMyVec :: (FinSet a) => Vec a -> MyVec a | | instance Monad MyVec where | return a = MkMyVec (vreturn a) | etc. | | GADTs are scheduled for version 6.4 of GHC. The version of GHC now in | CVS does not currently allow this however, see GHC bug #1097046. I'm not sure how to fix this bug, or even if it is a bug. Let me explain why. Usually GHC is reluctant to perform context reduction by applying instance declarations. Two reasons: 1. Postponing context reduction give more opportunities for sharing 2. With overlapping instance declarations, eager context reduction may select an instance declaration "too early"; later, more type information may be available. In fact, when you force context reduction by giving a type signature, GHC complains if (2) applies with its "the instance selected depends on the instantiation of x" message. Now, suppose we have f x = case x of MkMyVec x -> <rhs> where (x::MyVec a) the <rhs> gives a constraint (Foo [a]). And suppose we have instance (Eq a, MyVec a) => Foo [a] then you may say that f should get type f :: Eq a => MyVec a -> ... because the MyVec part gets discharged by the MkMyVec pattern match, leaving only the Eq a part. Fine. Now suppose there's an overlapping instance declaration for Foo, say Foo [Char]. Then GHC would decline to select the Foo [a] instance, so we'd infer the type f :: Foo [a] => MyVec a -> .... There would be no error message in either case, just a different type inferred. A programmer might well find this odd. Sometimes the MyVec constraints are handled by the MkMyVec constructor, and sometimes they aren't. The root of the difficulty is that the MkMyVec constructor captures constraints that mention no existentially-quantified type variable, so it'd make perfect sense to float them outwards. (In contrast, constraints mentioning existentially-quantified type variables *must* be dealt with at the pattern match.) Here's one idea. If you specify a type signature for 'f' then matters are much clearer. In my example, if you specify f :: Eq a => MyVec a -> ... then it's absolutely clear what constraints are available at the pattern match (namely Eq a, MyVec a), and we solve all the constraints there. So the rule would be: constraints are always solved *lazily* when inferring, so we'd infer f :: Foo [a] => MyVec a -> ... even without any overlap but they are solved *eagerly* when checking against a user-supplied type sig so when checking f :: Eq a => MyVec a => ... we'd accept the program. (If there were overlapping instances, we'd complain about them.) But I suspect that programmers will prefer a more woolly rule that does as much constraint solving as it can (provided all instance selection is uniquely determined). That would give the slightly surprising behaviour I describe above if an overlapping instance declaration was added. Thoughts anyone? Simon

2 1

RE: [Haskell] Re: "Never" GADT Function?
by Simon Peyton-Jones 11 Jan '05

11 Jan '05

GHC deliberately rejects programs with unreachable code, as Ashley says, because a) they often are programmer errors b) I can't think of a time when they'd be useful. I don't see what's wrong with neverT :: T Char -> x neverT = error "urk" After all, since you can't construct a value of type (T Char) there's not much point in pattern-matching on it. Simon | -----Original Message----- | From: haskell-bounces(a)haskell.org [mailto:haskell-bounces@haskell.org] On Behalf Of Lennart | Augustsson | Sent: 08 January 2005 10:30 | To: Ashley Yakeley | Cc: haskell(a)haskell.org | Subject: Re: [Haskell] Re: "Never" GADT Function? | | Well, this compiles: | | data T a where | BoolT :: T Bool | IntT :: T Int | | neverT' :: T a -> x | neverT' BoolT = error "Bool" | neverT' IntT = error "Int" | | neverT :: T Char -> x | neverT = neverT' | | But it uses error for the unreachable cases, maybe not what you want. | | -- Lennart | | Ashley Yakeley wrote: | > In article <16863.36246.132716.788213(a)sf0.comp.nus.edu.sg>, | > Martin Sulzmann <sulzmann(a)comp.nus.edu.sg> wrote: | > | > You should be able to even write | > | > neverT (BoolT x) = x | > neverT (IntT x) = False | > | > Actually I didn't put in any arguments to my constructors. Apart from | > that I agree: this should compile, but doesn't: | > | > data T a where | > BoolT :: T Bool | > IntT :: T Int | > | > neverT :: T Char -> x | > neverT BoolT = "hello" | > neverT IntT = 37 | > | > Pick.hs:11:9: | > Inaccessible case alternative: Can't match types `Bool' and `Char' | > When checking the pattern: BoolT | > In the definition of `neverT': neverT BoolT = "hello" | > | | _______________________________________________ | Haskell mailing list | Haskell(a)haskell.org | http://www.haskell.org/mailman/listinfo/haskell

1 0

ANN: Codec 1.0
by Dominic Steinitz 09 Jan '05

09 Jan '05

Partly by way of experiment with darcs and Cabal and partly because I am restructuring the cryptographic library, I have set up a Codec package with some utilities in it, see http://www.haskell.org/Codec/index.html. Currently it contains only the module Codec.Utils. To get the package: darcs get http://www.haskell.org/Codec To build, you need the Cabal libraries installed: ghc -package Setup.lhs -o setup -package Cabal ./setup configure ./setup build ./setup install The latter needs to be done as root. base64 encoding / decoding should probably be included in Codec.Utils (I've seen copies of the same code all over the place) and if anyone has the time, they are very welcome to add it. I'd also welcome feedback on whether this is the right place in the hierarchy for such a package and modules. Finally, I don't want to tread on any toes and if anyone feels that this is their area then I would be happy to hand this over and submit it as a contribution. Dominic. PS I'd also like to thank the darcs and Cabal teams for their tools.

1 1

Enum class
by Jaime Nino 09 Jan '05

09 Jan '05

I would like to write a function with type next :: Enum a => a -> a which will wrap a values around; that is next value == toEnum 0 where value is the last value of a. I have not found any functionality on Enum types to do this. In particular how to specify the last element of any Enum type, or how to determine the number of elements in an Enum type. Thanks. Jaime Nino

3 3

"Never" GADT Function?
by Ashley Yakeley 08 Jan '05

08 Jan '05

OK, so I've been playing around with GADTs in the CVS GHC. Consider this: data T a where BoolT :: T Bool IntT :: T Int neverT :: T Char -> x Is it possible to write neverT without using undefined? Such functions are occasionally useful, especially when GADTs are being used as type witnesses. Then again, I suppose it's the same situation for types with no constructors. -- Ashley Yakeley, Seattle WA

3 3

Fail <<loop>> ?
by Andreas Marth 08 Jan '05

08 Jan '05

Hello! While executing a program of mine I get a message "Fail: <<loop>>" which is not an error which I throw somewhere. So where does it come from? And how can I proceed further to solve the problem? Any hints welcome, Andreas

3 2

The Implementation of Functional Programming Languages
by Simon Peyton-Jones 07 Jan '05

07 Jan '05

I'm happy to announce that my out-of-print 1987 book, The Implementation of Functional Programming Languages is now available online at http://research.microsoft.com/%7Esimonpj/papers/slpj-book-1987/index.htm Very many thanks to Marnie Montgomery, who did all the work. Happy reading Simon

1 0