July 2010 - Haskell-Cafe

State of the Hackage: Q1, Q2 2010
by Don Stewart 01 Jul '10

01 Jul '10

Downloads and popular packages on Hackage for Q1 and Q2 this year. http://donsbot.wordpress.com/2010/06/30/popular-haskell-packages-q2-2010-re… -- Don

4 5

Associated types
by Andrew Coppin 01 Jul '10

01 Jul '10

Consider the following: class Path p where ... class Path p => CompletePath p where ... class Path p => IncompletePath p where type CompletedPath p :: * Obviously, the idea is that CompletedPath Foo (where Foo is an IncompletePath) should yield some type which is a CompletePath. However, the source code does not actually state this, so GHC (rightly) complains that it was unable to deduce this constraint from the actual code written. Is it possible to state this constraint? If so, how? (And if not, the answer, presumably, is to fundamentally redesign the whole thing...)

6 8

chart "broken" under 6.12 according to criterion
by Tom Doris 01 Jul '10

01 Jul '10

According to the criterion.cabal file shipped with the latest (0.5.0.1) version of criterion, the Chart package is broken under GHC 6.12: flag Chart

4 5

Type-Level Programming
by Walt Rorie-Baety 01 Jul '10

01 Jul '10

I've noticed over the - okay, over the months - that some folks enjoy the puzzle-like qualities of programming in the type system (poor Oleg, he's become #haskell's answer to the "Chuck Norris" meme commonly encountered in MMORPGs). Anyway,... are there any languages out there whose term-level programming resembles Haskell type-level programming, and if so, would a deliberate effort to appeal to that resemblance be an advantage (leaving out for now the hair-pulling effort that such a change would entail)? Or, better yet, is there an Interest Group or committee (Working, or not), that is looking at a coherent architecture or design for a possible future version of Haskell (no offense to Tim Sheard's excelent Ωmega project)? Walt "BMeph" Rorie-Baety "A mountain that eats people? I want one!." - Richard, of LFGComic.com

15 37

ghc api printing of types
by Phyx 01 Jul '10

01 Jul '10

I was wondering how given a Type I can get a pretty printed type out of it. I'm currently using showSDocUnqual . pprType . snd . tidyOpenType emptyTidyEnv But this has the problem that predicates don't get printed, anyone know how GHCi does this? Thanks, Phyx

1 0

Inferring the most general type
by Job Vranish 01 Jul '10

01 Jul '10

Esteemed fellow haskellers, I recently ran into a very simple real life case where Haskell's rules for inferring the types for mutually recursive definitions resulted in a type that was less general than it could be. It took me a while to realize that the type error I was getting wasn't actually a problem with my code. I understand why Haskell does this (it infers the strongly connected mutually recursive definitions monomorphically), but I think it _could_ infer the more general type even with recursive definitions like this. Here is a simplified example that illustrates the problem: > import Data.Maybe > -- The fixed point datatype > data Y f = Y (f (Y f)) > -- silly dummy function > maybeToInt :: Maybe a -> Int > maybeToInt = length . maybeToList > -- f :: Y Maybe -> Int > f (Y x) = g maybeToInt x > g h x = h $ fmap f x This is the type it wants to infer for g g :: (Maybe Int -> Int) -> Maybe (Y Maybe) -> Int This is the type I think it should have, note you can't force the type with a typesig without -XRelaxedPolyRec g :: (Functor f) => (f Int -> b) -> f (Y Maybe) -> b If I use -XRelaxedPolyRec I can manually specify the more general type, but then I have to convince myself that there isn't a more general type that I'm missing. Are there other known algorithms that yield a more general type? and if so, what was the rational for Haskell keeping the current method? I worked out an alternative algorithm that would give a more general type (perhaps the most general type) but it has factorial complexity and probably wouldn't be good for strongly connected groups with 7 or more members. Even so, I would much rather have the inferred types always be the most general ones and be required to add type signatures for mutually recursive groups with 7 or more members (which probably need to be redesigned anyway) than be always required to manually figure out the more general signatures. What do you think? - Job

2 1

RE: [Haskell-cafe] checking types with type families
by Simon Peyton-Jones 01 Jul '10

01 Jul '10

| Here's a concrete case I recently ran into: | | type family SomeOtherTypeFamily a | class SomeClass a where type SomeType a | instance a ~ SomeOtherTypeFamily b => SomeClass a where | type SomeType a = (b,Int) | -- (error) Not in scope: type variable `b' | | The same thing with fundeps: | | class SomeClass a b | a -> b | instance a ~ SomeOtherTypeFamily b => SomeClass a (b,Int) | -- works fine It's not the same thing at all! The code you give certainly should fail. Suppose you use fundeps thus: class C a b | a->b where op :: a -> b -> b The idiomatic way to replace this with type functions is to remove the 'b' parameter, thus: class C a where type B a op :: a -> B a -> B a Sometimes you don't want to do this. For example, you might have a bidirectional fundep. Then you can (or rather will be able to) use a superclass thus: class (B a ~ b) => C a b where type B a The superclass says that you can only instantiate this class with a second argument b that is equal to (B a). Thus you might have an instance instance C Int Bool where type B Int = Bool Meanwhile we are working hard on the new type inference engine, which will allow superclass equalities like these. Simon

1 0

dirs in Emacs/Vim in Cabal
by Chris BROWN 01 Jul '10

01 Jul '10

Dear Haskellers, I am in the process of creating a cabal HaRe and I have run into a minor problem. I have an emacs and a vim script that HaRe uses to invoke the refactorer from the two editors. In those scripts, I point HaRe to a directory containing some libraries it needs to chase the imports of the modules. Cabal installs these libraries as data files into whatever prefix the user requests at install time. However, the problem I have is that it then requires a manual user intervention to manually configure emacs/vim to point to this new directory in the 'customizable' panel of HaRe. My question is: Is it possible to allow the emacs/vim scripts that I have know where this directory is automatically? I hope someone can help, Kindest regards, Chris Brown.

2 1

functional dependencies question
by Patrick Browne 01 Jul '10

01 Jul '10

Hi, My understanding of functional dependencies is that they can be used to ensure that one type depends on another type. For example, the type of location depends could on the type of the object at that location. Consider two models 1) The location of an aircraft landing should have location of the type AirportCode, 2) The location of a car's destination should have a type CartesianCoordinate. AirportCodes should not be used as locations in 2) and CartesianCoordinates should not be used as locations in 1). Haskell class, instances, and fundeps were used to represent the above requirements. I generalized the requirement a bit, testing on a variety of types (below). Obviously, some cases are ambigous and/or conflicting and the Haskell compiler correctly flags this. Why do some cases such as 1) fail to run even if they are the only instantiation. Regards, Pat {- C:\GHC\ghc-6.8.3\bin\ghci.exe -XMultiParamTypeClasses -XFunctionalDependencies -fglasgow-exts -fallow-undecidable-instances -} class LocatedAt object location | object -> location where spatialLocation :: object -> location -- 1) Compiles but does not run instance LocatedAt Int String where spatialLocation(1)="home" -- 2) Compiles and runs OK instance LocatedAt String Int where spatialLocation("home")=1 -- 3) Compiles and runs OK, but obviously not in conjunction with 2 instance LocatedAt String String where spatialLocation("home")="home" -- 4) Compiles and runs OK instance LocatedAt Bool String where spatialLocation(True)="home" -- 5) Compiles and runs OK, but obviously not in conjunction with 2 instance LocatedAt String Bool where spatialLocation("home") = True -- 6) Not OK instance LocatedAt Float String where spatialLocation(2.3)="home" -- 7) Compiles but does not run instance LocatedAt Int Char where spatialLocation(1)='1' -- 8) Compiles and runs OK instance LocatedAt Char Int where spatialLocation('1')=1 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 5

Newbie question about using WinGHCi
by Christopher Tauss 01 Jul '10

01 Jul '10

Hello - I just a day or so ago downloaded Hakell and am playing around with it, and I came upon this problem with WinGHCi: I am able to enter a multi-line "do" statement that works if I use brackets and semi-colon like so: Prelude> :{ Prelude| let main2 = do { Prelude| putStrLn "Please enter your name: "; Prelude| name <- getLine; Prelude| putStrLn ("Hello, " ++ name ++ ", how are you?") } Prelude| :} Prelude> main2 Please enter your name: CT Hello, CT, how are you? Prelude> Note there is no indentation. This makes sense to me because the :{ :} just take all the lines in between and make it one line. But it seems to me to be IMPOSSIBLE to input this into WinGHCi using indentation like most of the code samples seem to do. Should I just be satisfied that it works using brackets/ semi-colons? Or is there something obvious that I am missing that allows for indentation? Thanks in advance. It's issues lik this that keep me up into the depths of night. Best Regards, Chris

2 1