July 2010 - Haskell-Cafe

Applying a value to a function generically
by Arnaud Bailly 14 Jul '10

14 Jul '10

Hello, I would like to construct a collection of function-like objects on which I could apply some value, in a typesafe and clean way. Let's say I have something like this: > data Fun = forall a b . F (a -> b) > type Callables = Map String Fun I would like to be able to write: > invoke :: Callables -> a -> String -> b > invoke m d k = case lookup k m of > Just (F f) -> f d > Nothing -> error $ "unable to find function for key " ++ k which of course does not compile nor even make sense. I suspect I need to have more information in Fun to be able to apply functions correctly, but don't know where to look at. I read recently an article about implicit configurations that used some type wizardry to ensure correct and typesafe use of external data and sounded pretty cool, (http://www.cs.rutgers.edu/~ccshan/prepose/prepose.pdf) but I suspect this is 1) too complex for me an 2) overkill for my problem. Could someone point me in the right direction ? Thanks in advance Arnaud

3 2

Collecting MonadError errors generically
by oleg＠okmij.org 14 Jul '10

14 Jul '10

Leon Grynszpan wrote: > What I want, instead, is to run a whole bunch of > computations that may throw errors. If there are any errors, I want to > collect them all into one big master error. If not, I want a list of > results. Here's an example of usage: > > couldThrowError :: (Error e, MonadError e m) => t1 -> m t2 > getParams :: (Error e, MonadError e m) => [t1] --> m [t2] > getParams = groupErrors . map couldThrowError > > I found it pretty easy to implement groupErrors for Either String: > ... > The problem, though, is that running this function now causes type > inference to provide "Either String" as my MonadError implementation... > It would be nice if I could stay generic. You might find the following two functions useful: > reify :: (Error e, MonadError e m) => m a -> m (Either e a) > reify m = (m >>= return . Right) `catchError` (return . Left) > > -- Not needed here, but very nice to have anyway > reflect :: (Error e, MonadError e m) => m (Either e a) -> m a > reflect m = m >>= either throwError return Then groupErrors can be written almost the same way you wrote for the Either String monad -- but now generically > groupErrors :: (Monoid e, Error e, MonadError e m) => [m a] -> m [a] > groupErrors lst = mapM reify lst >>= \lst -> > case partitionEithers lst of > ([],xs) -> return xs > (errs,_) -> throwError (mconcat errs) If you don't like the Monoid constraint, this is fine. But you have to provide then some other way to group errors. For example, you could use the new extensible exceptions. Here is the rest of the code: > couldThrowError :: (MonadError String m) => String -> m Int > couldThrowError s = case reads s of > [(n,"")] -> return n > _ -> throwError $ "parse error: " ++ s ++ "\n" > > getParams :: (Monoid e, Error e, MonadError e m) => > (t1 -> m t2) -> [t1] -> m [t2] > getParams f = groupErrors . map f > > test :: Either String [Int] > test = getParams couldThrowError ["1","2","a","b"] If you wish to know more about reify and reflect, a good paper to start is the following. Section 1.2 talks directly about exception monads. @InProceedings{ filinski-representing, author = "Andrzej Filinski", title = "Representing Monads", pages = "446--457", crossref = "popl1994", url = "http://www.diku.dk/~andrzej/papers/RM-abstract.html http://www.diku.dk/~andrzej/papers/RM.dvi http://www.diku.dk/~andrzej/papers/RM.ps.gz", abstract = "We show that any monad whose unit and extension operations are expressible as purely functional terms can be embedded in a call-by-value language with ``composable continuations''. As part of the development, we extend Meyer and Wand's characterization of the relationship between continuation-passing and direct style to one for continuation-passing vs.~general ``monadic'' style. We further show that the composable-continuations construct can itself be represented using ordinary, non-composable first-class continuations and a single piece of state. Thus, in the presence of two specific computational effects---storage and escapes---\emph{any} expressible monadic structure (e.g., nondeterminism as represented by the list monad) can be added as a purely definitional extension, without requiring a reinterpretation of the whole language. The paper includes an implementation of the construction (in Standard ML with some New Jersey extensions) and several examples." } There is a follow-up: @InProceedings{ filinski-layered, author = "Andrzej Filinski", title = "Representing Layered Monads", pages = "175--188", crossref = "popl1999", url = "http://www.diku.dk/~andrzej/papers/RLM-abstract.html http://www.diku.dk/~andrzej/papers/RLM.dvi http://www.diku.dk/~andrzej/papers/RLM.ps.gz", } The term reification has been introduced 26 years ago: @InProceedings{ friedman-reification, author = {Daniel P. Friedman and Mitchell Wand}, title = {Reification: Reflection without Metaphysics}, pages = {348--355}, crossref = "lfp1984", abstract = {We consider how the data structures of an interpreter may be made available to the program it is running, and how the program may alter its interpreter's structures. We refer to these processes as reification and reflection. We show how these processes may be considered as an extension of the fexpr concept in which not only the form and the environment, but also the continuation, are made available to the body of the procedure. We show how such a construct can be used to effectively add an unlimited variety of ``special forms'' to a very small base language. We consider some trade-offs in how interpreter objects are reified. Our version of this construct is similar to one in 3-Lisp [Smith 82, 84], but is independent of the rest of 3-Lisp. In particular, it does not rely on the notion of a tower of interpreters.} }

1 0

Collecting MonadError errors generically
by Leon Grynszpan 13 Jul '10

13 Jul '10

Hello, This is my first post to the Haskell Café, and I'm hoping the issue I'm tackling here isn't one that's been thoroughly explored elsewhere. If that's the case, I'll apologize in advance for my insufficient google chops. Otherwise, here goes... I do a lot of work within the error monad. Having read Eric Kidd's blog on error reporting in Haskell (http://www.randomhacks.net/articles/2007/03/10/haskell-8-ways-to-report-err…) I'm always a bit paranoid about dealing with errors in the most generic, flexible way possible. As a result, a lot of the functions I use that could result in errors have signatures along these lines: couldThrowError :: (Error e, MonadError e m) => t1 -> t2 -> m t3 The way I figure it, this means that I can use these functions within any particular MonadError, which keeps me happy if I decide to change implementations at any particular point during development. Usually, I use these couldThrowError functions in a standard do block. Something along the lines of: couldThrowError :: (Error e, MonadError e m) => t1 -> t2 -> m t3 couldThrowError = do x <- processWithPossibleError t1 y <- processWithPossibleError t2 ... return (z :: t3) Things stay generic at this level. I can still choose any MonadError implementation I want, and if any of the computations in the do block throws an Error, the whole function short-circuits. Often enough, however, I don't want this kind of short-circuiting behavior. What I want, instead, is to run a whole bunch of computations that may throw errors. If there are any errors, I want to collect them all into one big master error. If not, I want a list of results. Here's an example of usage: couldThrowError :: (Error e, MonadError e m) => t1 -> m t2 getParams :: (Error e, MonadError e m) => [t1] --> m [t2] getParams = groupErrors . map couldThrowError I found it pretty easy to implement groupErrors for Either String: groupErrors :: (Error e, MonadError e m) => [Either String a] -> m [a] groupErrors eithers = case partitionEithers eithers of ([], xs) -> return xs (es, _) -> strError $ unlines es The problem, though, is that running this function now causes type inference to provide "Either String" as my MonadError implementation. If any of the potential error sources are dealing with a different concrete implementation, I'm stuck! It would be nice if I could stay generic. I made an effort at this that you can see below, but it strikes me as very awkward. collectErrors :: (Monoid e, Error e, MonadError e m) => m a -> m [a] -> m [a] collectErrors m1 m2 = do m1r <- (m1 `catchError` (\e1 -> (m2 `catchError` (\e2 -> throwError $ mappend e1 e2)) >> throwError e1)) m2r <- m2 return $ m1r : m2r groupErrors' :: (Monoid e, Error e, MonadError e m) => [m a] -> m [a] groupErrors' = foldr collectErrors (return []) As you can see, I now require an error type that implements monoid. An example I've tested is: newtype ErrorString = ES String deriving (Read, Show, Eq, Ord, Error) instance Monoid ErrorString where mempty = ES "" mappend (ES s1) (ES s2) = ES $ s1 `mappend` "\n" `mappend` s2 This seems to work, but something strikes me as being very hackish here. Maybe it's just that collectErrors looks very ugly, and in Haskell my intuition tells me that what looks ugly on the surface is often ugly down below. I suspect that I might really not be approaching this problem the right way, but after a lot of time spent wrestling with it, I'm eager to read your suggestions. Is there a better way to do groupErrors, or should I be looking at an entirely different philosophy? Thanks, Grynszpan

2 1

library for objective-j?
by Günther Schmidt 13 Jul '10

13 Jul '10

Hi all, is there a haskell library for generating Objective-J? Günther

2 3

AngloHaskell 2010
by Derek Wright 13 Jul '10

13 Jul '10

Are you interested in AngloHaskell 2010? AngloHaskell is a gathering of all people Haskell-related from beginners to seasoned hackers. Previously it has been held in Cambridge (three times) and London (once). Last year I volunteered to organise it but I want know when you can attend. Traditionally it is the first Friday and Saturday in August, which this year is the 6th and 7th August. But we may only have enough time to organise one day so the question for you is: Q) Which day is better for you? Friday, Saturday or either? I am looking for venues in London but if you can suggest or offer a venue then let me know. As well as attendees we are looking for talks on all and any Haskell related subjects. There is a section on the wiki page http://www.haskell.org/haskellwiki/AngloHaskell/2010 for talks. Please go ahead and add your talk to it. Let me know by email, the wiki or the #anglohaskell IRC channel. I planning to have the venue confirmed this week and then a full announcement will be made. Thanks, Derek.

1 0

The site has been exploited (again)
by Christopher Done 13 Jul '10

13 Jul '10

http://haskell.org/ It says "TO BUY Cilamox ONLINE", etc. Whoever has power please fix this and upgrade the bloody wiki. This is ridiculous. Point the domain at tryhaskell.org or something. I'll put a holder page up. Anything. Cheers

6 10

Marshalling
by Phyx 13 Jul '10

13 Jul '10

Hello I'm wondering if anyone ever benchmarked marshalling in Haskell/GHC. No matter how much I optimize my Haskell code my program still seems to run slow, which leads me to beleive that Marshalling is painfully slow. Does anyone know a way I can test this and fix it? Regards, Phyx

3 4

ByteString, zipWith', and rewrite rules
by Thomas M. DuBuisson 13 Jul '10

13 Jul '10

Comments on the zipWith' function inside of Data.ByteString say: -- Rewrite rules -- are used to automatically covert zipWith into zipWith' when a pack is -- performed on the result of zipWith. This is only true internally to Data.ByteString because the zipWith' function could be inlined away by GHC once that module is compiled, right? If I want "pack (zipWith xor bs1 bs2)" to be efficient then I'll have to either get zipWith' exported or write my own lower level zipWith'? If I'm wrong here then the comment needs to be moved to somewhere that will get haddockized, perhaps with zipWith. Cheers, Thomas

2 1

Functional dependencies and Peano numbers
by Oscar Finnsson 13 Jul '10

13 Jul '10

Hi, there is a lot of buzz around functional dependencies on the mailing list and Planet Haskell. I've read some of the tutorials and I think I understand how they work but I still haven't figured out where they can be useful. * Can someone give me a real world (preferably hackagedb) example where functional dependencies are used? * Can someone give me a real world (preferably hackagedb) example where Peano numbers a la "data Zero; data Succ n" are used? Finally... * All the examples involving functional dependencies are on the form > a b c | a b -> c but can they also be on a form similar to > a b c d e f g h| b c -> d e f | b d g -> h (i.e. d,e,f are decided by the b,c-combination while h is decided by the b,d,g-combination)? -- Oscar

6 5

64-bit #const with hsc2hs?
by Dean Herington 13 Jul '10

13 Jul '10

Does anyone know if I can bring a 64-bit (long long or unsigned long long) value from C-land into Haskell via hsc2hs? #const on such a value seems to provide only the low 32 bits, at least in my environment (Haskell Platform 2009.2.0.2 on Windows). I've tried to create a 64-bit variant of #const without success. Dean

1 0