- Haskell-Cafe - Haskell.org

(automatic) type classes context inference
by João Paulo 28 Aug '10

28 Aug '10

Hello everyone, I am developing a toolset in which I have several (multiparameter) type classes; It is often the case that I can only define a data-type X as an instance of one such class (say A), if X is an instance of another class (say B); The thing is that, while it is hard for me, because of all the type parameters that I have to deal with, to add 'X is an instance of B' to the context cxt_A in 'instance cxt_A => A X' ghc is always able to correctly infer all type parameters; In fact, I always get: 'Could not deduce (B X t1 ... tn) from the context cxt_A arising from ... Probable fix: add (B X t1 ... tn) to the context cxt_A ...' In my case, this is the fix that I always need: most of the times, I am just copy-pasting (B X t1 ... tn) to cxt_A! Is there a way, say a compilation option, to avoid this? can anyone please help me here? :) thank you very much -- João Paulo Fernandes Universidade do Minho www.di.uminho.pt/~jpaulo

3 2

type classes and logic
by Patrick Browne 28 Aug '10

28 Aug '10

Hi, I am trying to understand type classes and sub-classes in purely logical terms From the literature I gather that: 1)a type class is a predicate over types (using type variables) 2)type sub-classing is a logical implication between these predicates. But I feel that 1) and 2) does not give enough detail and I am missing the logical significance of the function definitions in the class and their implementations in the instances. This following example outlines my current (mis?)understanding: data Person = Person Name deriving Show data Employee = Employee Name Position deriving Show class BEING being where The class BEING is logically a predicate with type variable being(i.e. BEING(being)) BEING(being) is true for types that are members of the BEING type class. Any types that instantiate the type class BEING will have all the BEING class functions defined on them. Such a type is a member of the type-class BEING Actual function signatures and/or default implementations omitted A proof for the predicate BEING(being) must show that there is an inhabited type (a type which has values) If we find a value for a type that is a proof that a type exists (it is inhabited) that is a member of the class instance BEING Person where All the functions from the class BEING must be defined for the type Person It is required the functions in the instance respect the types from the class and type from the instantiated parameter. t is required that when the functions are run they produce values of the required type. Implementations omitted With at least one instantiation existing (e.g. BEING(Person) = true) we can define a sub-class In the sub-class definition we can assume BEING(human) to be true (type variable name does not matter) And based on that assumption we can write a logical implication BEING(human) => HUMAN(human) class BEING human => HUMAN human where At this point there is no additional functionality is defined for the subclass HUMAN Sub-classing is logical implication BEING(human) => HUMAN(human) All types t that make BEING(t) = true also make HUMAN(t)=true In general for the HUMAN sub-class to be useful additional constraints are added after the where keyword. These are similar in purpose to those described in an ordinary class (not a sub-class) Another example from Bird[1.Intro. to Functional Prog. using Haskell] class Enum a where toEnum :: a -> Int fromEnum :: Int -> a A type is declared an instance of the class Enum by giving definitions toEnum and fromEnum, functions that convert between elements of the type a and the type Int. In instance declarations fromEnum should be a left inverse to toEnum That is for all x fromEnum(toEnum x) = x. This requirement cannot be expressed in Haskell My questions are: a) Is the above *logical view* of type classes correct? b) What is the logical role of the functions in the classes and instances. Are they constraints on the types in predicates? c) Apart from signature matching between class-to-instance is there any logical relation between the functions in a class and the functions in an instances. From Birds[1] example I suspect that while the types are subject to logical rules of 1) and 2) the actual functions in the instance do not have to obey any logical laws. Any instantiation that respects types is fine? Regards, Pat This message has been scanned for content and viruses by the DIT Information Services E-Mail Scanning Service, and is believed to be clean. http://www.dit.ie

4 6

On to applicative
by michael rice 28 Aug '10

28 Aug '10

fmap seems oddly named because no "mapping" takes place, except in the fourth example, where the map is "passed in." Just sayin'. Michael 1) Prelude Control.Monad Control.Applicative> fmap (++ "abc") getLine xyz "xyzabc" 2) Prelude Control.Monad Control.Applicative Data.Char Data.String> fmap (splitAt 3) getLine qwertyuio ("qwe","rtyuio") 3) Prelude Control.Monad Control.Applicative Data.Char> fmap toUpper getLine <interactive>:1:13: Couldn't match expected type `Char' against inferred type `[Char]' Expected type: IO Char Inferred type: IO String In the second argument of `fmap', namely `getLine' In the expression: fmap toUpper getLine 4) Prelude Control.Monad Control.Applicative Data.Char Data.String> fmap (map toUpper) getLine qwertyuio "QWERTYUIO"

2 2

FRP - style
by Maciej Piechotka 28 Aug '10

28 Aug '10

I wanted to learn something about FRP. I tried to use existing packages on hackage and write something in OpenGL but I failed - mainly due to problem of handling single occurrence of event[1]. I tried to write my own library (https://patch-tag.com/r/uzytkownik/reactive-event/home) and I'd like to ask if I'm doing it from right direction or do I miss some important aspects of FRP. I assume that the events are discrete and the value can change (along with action) only on event. For specific class of events there exists 'always' event associated with time [In practice it would mean that it is executed as fast as possible]. Regards [1] In retrospection - I should have rely more on immutable data to pass around information that I did handled event - but it seems for me an overkill.

1 0

ANN: mute-unmute-0.2.0.0
by Krzysztof Skrzętnicki 27 Aug '10

27 Aug '10

* Hello! I'm pleased to announce a small utility program I wrote recently. Today I pushed version 0.2.0.0 of mute-unmute to Hackage, from with you can cabal-install it right now (following the usual 'update' step). What it does? It's simple: it will mute the sound when you lock your screen and unmute it back again when you unlock it. Why anyone would need such program? Well, this is a little bit more complicated. I work in "open space" kind of office. It can get quite noisy there. I often sit with headphones on, but when I walk away from my desk I put them down and lock my screen. Here comes uncomfortable moment: when I remove my headphones suddenly people around me who doesn't use their headphones now start to hear the music I listen to. And they don't always like it, so I manually mute the sound to make happy. But now, instead of doing this task by hand I have a program do it for me :-) Unfortunately this kind of programs tend to be pretty much platform specific. Current version supports the recent versions of Gnome and KDE, but needs ALSA. It's quite common setup, but if you happen to have a different one you can easily hack the code (it's very straightforward) and send me patches. Patches are always welcome. The source repo is on Github and released versions are on Hackage: Github: http://github.com/Tener/mute-unmute <http://github.com/Tener/mute-unmute#> Hackage: http://hackage.haskell.org/package/mute-unmute Example use: ~/ mute-unmute # first session, it will ask for configuration options mute-unmute: Creating config directory... mute-unmute: Mute config is missing... Mute sound card and press RETURN mute-unmute: Unmute config is missing... Unmute sound card and press RETURN mute-unmute: Session: UnixAbstract {addrPath = "/tmp/dbus-1tB1AEUcgg", addrGuid = Just "19ec85e3778816ad38a77bbf0000003f"} mute-unmute: Connected mute-unmute: Handler added: [MatchType Signal,MatchInterface "org.gnome.ScreenSaver",MatchMember "ActiveChanged",MatchPath "/org/gnome/ScreenSaver"] mute-unmute: Handler added: [MatchType Signal,MatchInterface "org.freedesktop.ScreenSaver",MatchMember "ActiveChanged",MatchPath "/ScreenSaver"] ^C ~/ mute-unmute # remembers previous options which can be changed by calling with --store switch mute-unmute: Session: UnixAbstract {addrPath = "/tmp/dbus-1tB1AEUcgg", addrGuid = Just "19ec85e3778816ad38a77bbf0000003f"} mute-unmute: Connected mute-unmute: Handler added: [MatchType Signal,MatchInterface "org.gnome.ScreenSaver",MatchMember "ActiveChanged",MatchPath "/org/gnome/ScreenSaver"] mute-unmute: Handler added: [MatchType Signal,MatchInterface "org.freedesktop.ScreenSaver",MatchMember "ActiveChanged",MatchPath "/ScreenSaver"] mute-unmute: Message {mType = Signal, mFlags = [NoReplyExpected], mSerial = 31, mPath = Just "/org/gnome/ScreenSaver", mInterface = Just "org.gnome.ScreenSaver", mMember = Just "ActiveChanged", mErrorName = Nothing, mReplySerial = Nothing, mDestination = Nothing, mSender = Just ":1.19", mBody = [Variant (True) {- b -}]} mute-unmute: Lock screen mute-unmute: Message {mType = Signal, mFlags = [NoReplyExpected], mSerial = 32, mPath = Just "/org/gnome/ScreenSaver", mInterface = Just "org.gnome.ScreenSaver", mMember = Just "ActiveChanged", mErrorName = Nothing, mReplySerial = Nothing, mDestination = Nothing, mSender = Just ":1.19", mBody = [Variant (False) {- b -}]} mute-unmute: Unlock screen ^C * You can get usage information with --help switch: *~/ mute-unmute --help *mute-unmute --store ask for mute and unmute configurations --store-mute ask for mute configuration --store-unmute ask for unmute configuration -d --daemon wait for screensaver state changes -h, -? --help show help There isn't much more I can say. I hope you find the program at least educational (as a simple example of DBus usage) if not useful. Best regards, Krzysztof Skrzętnicki

1 0

warn-incomplete-patterns and GADTs
by Tom Nielsen 27 Aug '10

27 Aug '10

Hi, is warn-incomplete-patterns (in GHC 6.10.3) less clever than it could be? {-# OPTIONS_GHC -fglasgow-exts #-} {-# OPTIONS_GHC -fwarn-incomplete-patterns #-} module Vec where data Z data S a data Vec n a where VNil :: Vec Z a VCons :: a -> Vec m a -> Vec (S m) a instance Eq a => Eq (Vec n a) where VNil == VNil = True VCons x vx == VCons y vy = x==y && vx == vy give the warning: Warning: Pattern match(es) are non-exhaustive In the definition of `==': Patterns not matched: VNil (VCons _ _) (VCons _ _) VNil but of course VNil and VCons can never have the same type. Tom

2 1

phantom types for inheritance between types
by Andrew U. Frank 27 Aug '10

27 Aug '10

there are packages which use phantom types to express inheritance (subtype) relations. unfortunately i cannot find a simple example and seem to misunderstand something. can somebody help with the following simple example? i have a most general type Dated, a subtype Note and a subsubtype Task. i have only instances for Note and Task, but operations for tasks and dated. my primary question is: why should i have the contexts in the classes (the code compiles without them)? they do not say more than what is already expressed in the instance declarations. i do also not think that i have used phantom types properly (eg. gtk2hs uses phantom types differently). what would be the proper solution? would it work for cases with multiple inheritance? thank you! andrew class CDated a class CDated a => CNotes a class CNotes a => CTasks a instance CDated Task instance CDated Note instance CNotes Note instance CNotes Task instance CTasks Task class Dated n where datedop :: n -> n class Notes1 n where noteop :: n -> n class Tasks1 n where taskop :: n -> n data Note = Note Int data Task = Task Int instance CTasks x => Tasks1 x where taskop n = n instance CDated x => Dated x where datedop n = n t:: Task = Task 1 n :: Note = Note 1 tt = taskop t -- nt = taskop n -- compiler error -- as expected! tx = datedop t nx = datedop n

1 0

Re: [Haskell-cafe] Haskell and SciTE (was: Hackage on Linux)
by C. McCann 27 Aug '10

27 Aug '10

On Fri, Aug 27, 2010 at 12:57 PM, Andrew Coppin <andrewcoppin(a)btinternet.com> wrote: > C. McCann wrote: >> >> On Fri, Aug 27, 2010 at 7:40 AM, Andrew Coppin >> <andrewcoppin(a)btinternet.com> wrote: >> >>> >>> Unfortunately, I haven't found anything for Windows yet which has syntax >>> hilighting for Haskell. >>> >>> I use SciTE, which has hilighting for a bazillion languages (including >>> XML >>> and JSON), but not Haskell sadly. >>> >> >> Veering somewhat offtopic, but last time I checked, SciTE does have >> lexer support for Haskell, it just doesn't actually include (for >> unknown reasons) a language properties file to go with it. > > OK. Well maybe it's just the version I've got then? Or maybe, as you say, > because it's not enabled I don't know it's there. Going by the Scintilla release history, it's been in there for about five years, but you wouldn't find it just by using SciTE, since all the menu options and such are controlled by the .properties files, none of which mention Haskell... Anyway, if the version you have is less than five years old, you should be able to drop the file I linked to alongside the other language files, make a few global properties changes to add Haskell to the menus, probably a couple other minor things, and be good to go. I'd offer more detail, but I set mine up when I first started learning Haskell (about a year ago) and have since forgotten what needed to be done. > Ah yes. The reason I seldom update SciTE is that it then takes hours to put > all the configuration back to the way I like it. (Especially if option names > have changed or defaults are different now.) > > SciTE is a nice editor, but not especially well documented. (And, what, they > haven't made a configuration editor yet? :-P ) Yes, well. SciTE isn't so much an editor as a bit of trivial code to demonstrate the functionality of the Scintilla editor widget, which people (such as myself) insist on using as if it were a proper editor. If you want something friendlier, I'm afraid you're on your own. Unfortunately, there don't seem to be any Haskell bindings for Scintilla... - C.

1 0

Re: philosophy of Haskell
by Patai Gergely 27 Aug '10

27 Aug '10

> I don't agree. A concurrent change is the effect of an IO action, not > of the thread. For example if a concurrent thread writes to an MVar, > then that change becomes the effect of the next takeMVar, which gets > executed. If a concurrent thread changes a file on disk, then that > changing becomes the effect of the next readFile, which reads the > changed file. But that's exactly what the model cannot handle. Look at the following snippet again: > let (x, world1) = getLine world0 > world2 = print (x+1) world1 This clearly says that the world returned by getLine and the world consumed by print is the same one, since the state monad model is pure, therefore world1 is immutable. However, this is not true, since someone else could have modified it in the meantime. The state monad can only describe a single thread, but that's a non-existent situation in the case of I/O, since the world keeps changing outside the program even if the program itself is single-threaded. Gergely -- http://www.fastmail.fm - IMAP accessible web-mail

8 13

FW: [TYPES/announce] CUFP 2010 Call for Participation
by Simon Peyton-Jones 27 Aug '10

27 Aug '10

Dear Haskellers CUFP should be a lot of fun this year. Do consider going. Simon Commercial Users of Functional Programming Workshop (CUFP) 2010 Call for Participation Sponsored by SIGPLAN Co-located with ICFP 2010 _________________________________________________________ 1 - 2 October 2010 Baltimore (MD), USA http://cufp.org/conference/schedule/2010 Reservation will be available through ICFP's website: http://www.icfpconference.org/icfp2010 _________________________________________________________ Functional programming languages have been a hot topic of academic research for over 35 years, and have seen an ever larger practical impact in settings ranging from tech startups to financial firms to biomedical research labs. At the same time, a vigorous community of practically-minding functional programmers has come into existence. CUFP is designed to serve this community. The annual CUFP workshop is a place where people can see how others are using functional programming to solve real world problems; where practitioners meet and collaborate; where language designers and users can share ideas about the future of their favorite language; and where one can learn practical techniques and approaches for putting functional programming to work. CUFP 2010 will feature three hour Functional Programming tutorials given by language experts on the first day and Experience and Technical Talks on day two. Attendees may attend either or both days. Talks Program, October 2nd 2010 Luke Hoban (Microsoft) Keynote: F#: Embracing Functional Programming in "Visual Studio 2010" Sally A Browning (Galois Inc) Cryptol, a DSL for Cryptographic Algorithms Marius Eriksen (Twitter) Scaling Scala at Twitter Michael Fogus (Science Applications International Corporation) Naïveté vs. Experience - or, How We Thought We Could Use Scala and Clojure, and How We Actually Did It Neal Glew & Leaf Petersen (Intel) Functional Language Compiler Experiences at Intel Warren Harris (Metaweb) Functional Programming at Freebase Warren A. Hunt, Jr. (U. of Texas) Eating One’s Own Dog Food Rusty Klophaus (Basho Technologies) Riak Core: Building Distributed Applications Without Shared State Howard Mansell (Credit Suisse) Eden: An F#/WPF frameworok for building GUI tools Erik Meijer (Microsoft) Reactive Extensions (Rx): Curing Your Asynchronous Programming Blues Tutorial Program, October 1st 2010 Morning: Clojure (Aaron Bedra) Building robust servers with Erlang (Martin Logan) High Performance Haskell (Johan Tibell) Afternoon: F# 2.0 - A day at the beach (Rick Minerich) Implementing web sites with Scala and Lift (David Pollak) Camlp4 and Template Haskell (Nicolas Pouillard, Jake Donham) There will be no published proceedings, as the meeting is intended to be more a discussion forum than a technical interchange. For more information, for more information, including presentation abstracts and the most recent schedule information, visit http://cufp.org See you there!

1 0