- Haskell-Cafe - Haskell.org

YOrd and ysort
by Krzysztof Skrzętnicki 21 Mar '08

21 Mar '08

The ysort function I've written for YOrd class was incredibly inefficient. Here is better one: ysort :: (YOrd a) => [a] -> [a] ysort = head . mergeAll . wrap wrap :: [a] -> [[a]] wrap xs = map (:[]) xs mergeAll :: (YOrd a) => [[a]] -> [[a]] mergeAll [] = [] mergeAll [x] = [x] mergeAll (a:b:rest) = mergeAll ((merge a b) : (mergeAll rest)) merge :: (YOrd a) => [a] -> [a] -> [a] merge [] [] = [] merge xs [] = xs merge [] xs = xs merge xs ys = let ((sm:smS),gts) = xs `ycmp` ys in smS `seq` gts `seq` sm : (merge smS gts)

1 0

ANN: The Disciplined Disciple Compiler - alpha 1
by Ben Lippmeier 20 Mar '08

20 Mar '08

Hi All, I'm pleased to announce the initial alpha release of the Disciplined Disciple Compiler (DDC). Disciple is an explicitly lazy dialect of Haskell which includes: - first class destructive update of arbitrary data. - computational effects without the need for state monads. - type directed field projections. All this and more through the magic of effect typing. More information (and download!) available from: http://www.haskell.org/haskellwiki/DDC or http://code.google.com/p/disciple DDC: more than lambdas. Onward! Ben.

7 8

Help me understand this pattern
by Justin Bailey 20 Mar '08

20 Mar '08

All, Yesterday I wrote some code I thought was very clever (even though I stole most of it) and I wonder what the pattern is. I have seen it in at least one other place, and I suspect it's well known to long-time Haskellers. The problem I wanted to solved was creating a combinator that could be defined to take any number of differently typed arguments and convert them to a more fundamental representation. To be concrete, I have a data type Expr: data Expr a = ... where a is a phantom type. I wanted users of the library (haskelldb) to be able to define functions[1] such as: -- Trim whitespace from the expression given. trim :: Expr a -> Expr String -- Pad the expression given to the correct length with the given string, if one is given. rpad :: Expr a -> Expr Int -> Maybe String -> Expr String I wanted them to be able to do this using one combinator from the library and I did not want to expose the guts of the library so they could define the functions directly in the untyped, primitive representation. That is, I did NOT want to define my combinator as: func :: String -> [PrimExpr] -> Expr a Using ideas already in haskelldb, I came up with this: data ExprNil = ExprNil data ExprCons h tl = ExprCons h tl func :: String -> (ExprNil -> ExprCons (Expr e) tl) -> Expr o arg :: (Expr a) -> (tl -> ExprCons (Expr a) tl) Using these two functions, I could define trim and rpad via composition and I did not have to expose a primitive representation or use an existentially quantified data type[2]: trim str = func "trim" $ arg str rpad str len (Just pad_str) = func "rpad" $ arg str . arg len . arg (toExpr pad_str) where toExpr :: String -> Expr String rpad str len Nothing = func "rpad" $ arg str . arg len Notice how in the second case of 'rpad', the function defined only has two arguments since no padding character was provided. With that background, I'm hoping others can tell me what this generalizes too. It's pretty clear I'm working with type-level lists, but I'm sure there is more. Is this encoding existentials? Is the (ExprCons hd tl) type really a continuation? I'm positive this is a well-known pattern and I'd really like to learn more about it (history, other uses, libraries that depend on it, etc.). Thanks in advance! Justin [1] These functions are used to generate SQL code and do not actually do the operations given. For example, trim causes the "trim" function to be applied to the column given when the SQL is actually generated. [2] Just a note that these functions are NOT in haskelldb right now - this is my own experimentation.

1 0

ANN: Network.MiniHTTP 0.2
by Adam Langley 20 Mar '08

20 Mar '08

Since many people are interested in a ByteString HTTP library, here's network-minihttp 0.2 [1]. * Has code for both clients and servers * HTTP and HTTPS support (the latter is a little shaky, but mostly works) * DNS lookups (in the client code) don't block the whole process The Haddock documentation is at: http://darcs.imperialviolet.org/minihttp-docs/index.html Half the interfaces are ill-thought out, and the rest are just as bad. However, if you send me your ideas for how it should look I'll certainly consider them. The major omission at the moment are defensive timeouts for talking to the network. However, these shouldn't be too hard. [1] http://hackage.haskell.org/cgi-bin/hackage-scripts/package/network-minihttp… -- Adam Langley agl(a)imperialviolet.org http://www.imperialviolet.org

1 0

Termination of substitution
by Neil Mitchell 20 Mar '08

20 Mar '08

Hi I'm trying to show that a system of rules for manipulating Haskell expressions is terminating. The rules can be applied in any order, to any subexpression - and there is a problem if there is any possible infinite sequence. The rule that is giving me particular problems is: (\v -> x) y => x[v/y] (I realise this rule may duplicate the computation of y, but that is not relevant for this purpose.) In particular, given the expression (\x -> x x) (\x -> x x) we can keep applying this rule infinitely. However, I don't believe this expression is type safe in Haskell. Are any such expressions that would cause this to non-terminate not type safe? What are the necessary conditions for this to be safe? Does GHC perform lambda reduction, probably by introducing a let binding? Does the combination of reducing lambdas and creating let bindings give similar problems, particularly if bindings are inlined? I'm wondering if this rule is genuinely unsafe in Haskell. If it is, why isn't it an issue in real simplifiers. If it isn't, what makes it safe. Thanks for any help, Neil

4 5

Three Cabal/Hackage GSoC project proposals
by Duncan Coutts 20 Mar '08

20 Mar '08

All, I've proposed three GSoC proposals on different aspects of Cabal and Hackage: * Cabal 'make-like' dependency framework http://hackage.haskell.org/trac/summer-of-code/ticket/1552 * cabal-install package dependency resolution http://hackage.haskell.org/trac/summer-of-code/ticket/1553 * Hackage build reporting http://hackage.haskell.org/trac/summer-of-code/ticket/1554 This is in addition to the project that Neil proposed: * Hoogle 4 (using it as the main interface to the hackage website) http://hackage.haskell.org/trac/summer-of-code/ticket/1429 I'd like to invite interested people to help improve the proposals and to invite students to look at these and consider them. I'd also like other people interested in mentoring any of these projects to sign up. In particular, students who are interested should email the cabal-devel and/or libraries list (though they'd have to subscribe I think) to express their interest and work with us on improving the detail of their proposal. Duncan

1 0

How to implement Read instance for user defined type
by Verma Anurag-VNF673 20 Mar '08

20 Mar '08

Hi, I am trying to understand how to define instances for Read class for a user defined type. This is the sample code I wrote, but I am not able to get it correctly. So, let me know what's wrong with this program: module Mark where data Mark = Mark Int deriving (Show) instance Read Mark where readsPrec _ str = [(Mark x, t') | ("mark",t) <- reads str, (x,t') <- reads t When I run it in ghci: *Mark> reads "mark 5" :: [ (Mark,String)] ( I am intentionally using mark in lowercase) [] Why is the output coming out as [] as against expected output [(Mark 5, "")] ? -Anurag

2 1

Scoping in template haskell
by Lars Oppermann 20 Mar '08

20 Mar '08

Hi all, I am trying to get familiar with template haskell and I'm wondering whether the following is possible and if so, how it would be done properly... I have two templates, and I'd like to use one from the other. For instance the first one might be gen1 :: Int -> ExpQ gen1 i = [| "<<" ++ (show i) ++ ">>" |] Thus, I gould do > $(gen1 42) and I'd get "<<42>>" now I'd like to use gen2 in another generated expression so that I could so > $(gen2 [42, 66]) and get "<<42>><<66>>" My naive intention was to write gen2 :: [Int] -> ExpQ gen2 is = [| map (\x -> $(gen1 x)) is |] which gives me "Stage error: `x' is bound at stage 2 but used at stage 1 So I guess my actual question would be: how do I pass a variable bound outside a quotation into another splice used in that quotation? Bests, Lars

2 3

Type parameters in type families
by Hugo Pacheco 20 Mar '08

20 Mar '08

Hi, I am trying to understand some differences of parameterizing or not some arguments of type families. I have some code such as *type family G a :: * -> * instance Functor (G Int) where fmap f (Left ()) = Left () fmap f (Right x) = Right (f x) ggg :: Functor (G a) => G a x -> G a x ggg = fmap id* and it works fine. However, I need to parameterize one extra argument (due to type equality): *type family F a x :: * class FunctorF d where fmapF :: (x -> y) -> F d x -> F d y fff :: (FunctorF a) => F a b -> F a b fff = fmapF id* This second scenario fails to compile because the compiler cannot unify types a and b with types d and x from the fmapF declaration. Is there any other option than passing dummy variables to fmapF? * class FunctorF d where fmapF :: d -> x -> (x -> y) -> F d x -> F d y fff :: (FunctorF a) => a -> b -> F a b -> F a b fff a b = fmapF a b id* Thanks, hugo

3 5

Error handling in calculations
by hjgtuyl＠chello.nl 19 Mar '08

19 Mar '08

L.S., When playing with exceptions, I noticed the following strangeness: *Main> 1 / 0 Infinity *Main> 1 `div` 0 *** Exception: divide by zero This is in GHCi 6.8.2; WinHugs Sep 2006 gives: Main> 1 / 0 1.#INF Main> 1 `div` 0 Program error: divide by zero Is this difference between fractional and integral calculation intentional? It makes error handling more difficult. -- Met vriendelijke groet, Henk-Jan van Tuyl -- http://functor.bamikanarie.com http://Van.Tuyl.eu/ --

2 1