- Haskell-Cafe - Haskell.org

Re: [Haskell-cafe] haltavista - look for functions by example
by Roel van Dijk 19 Sep '10

19 Sep '10

Impressive! I didn't think you could implement it so quickly. Now someone needs to add this functionality to Leksah or hack something up for Emacs. Regards, Roel On Sun, Sep 19, 2010 at 8:27 PM, Paul Brauner <paul.brauner(a)inria.fr> wrote: > It works: > > brauner@worf:/tmp$ cat test.hs > import Test.QuickCheck > import Test.QuickCheck.Arbitrary > import Control.Monad(forM_) > > intMul :: Integer -> Integer -> Integer > intMul x n | n < 0 = -(intMul x $ abs n) > | n == 0 = 0 > | n > 0 = x + intMul x (n - 1) > > main = do xs <- sample' arbitrary > ys <- sample' arbitrary > zip xs ys `forM_` \(x,y) -> > putStrLn (show x ++ " " ++ show y ++ " " ++ (show $ intMul x y)) > > > brauner@worf:/tmp$ runghc test.hs | haltavista > Prelude (*) > > I will include this functionality in the next version. Thank you for > this nice idea. > > Paul

1 0

haltavista - look for functions by example
by Paul Brauner 19 Sep '10

19 Sep '10

Hello, I just hacked together something I've been talking about a while ago on that mailing list. It's a program that looks for functions given a set of input/outputs. Example session 1: brauner@worf:~$ haltavista 2 2 4 <EOF> Prelude (*) Prelude (+) Prelude (^) Example session 2 (refining previous search): brauner@worf:~$ haltavista 2 2 4 1 2 3 <EOF> Prelude (+) Example session 3 (higher-order functions): brauner@worf:~$ haltavista (+1) (+2) (1,1) (2,3) <EOF> Data.Graph.Inductive.Query.Monad (><) Under the hood, uses: - hint for type inference; - hoogle to get a list of candidate functions; - hint for testing. Hoogle calling facility has been copy-pasted (and later modified) from the Yi project. It's availaible on github (http://github.com/polux/haltavista) and I plan to release it on hackage as soon as I catch stack overflows that occur during testing using hint. So far I didn't manage to do it, even by catching asynchronous exceptions. Every suggestion/help is welcome. Also, if I got something wrong with the licences (Yi uses GPL-2 and code is copy-pasted, Hint BSD3 and is linked, Hoogle is called as an external process, haltavista is GPL-3 for now) please tell me. Paul

2 5

Re: [Haskell-cafe] haltavista - look for functions by example
by Paul Brauner 19 Sep '10

19 Sep '10

It works: brauner@worf:/tmp$ cat test.hs import Test.QuickCheck import Test.QuickCheck.Arbitrary import Control.Monad(forM_) intMul :: Integer -> Integer -> Integer intMul x n | n < 0 = -(intMul x $ abs n) | n == 0 = 0 | n > 0 = x + intMul x (n - 1) main = do xs <- sample' arbitrary ys <- sample' arbitrary zip xs ys `forM_` \(x,y) -> putStrLn (show x ++ " " ++ show y ++ " " ++ (show $ intMul x y)) brauner@worf:/tmp$ runghc test.hs | haltavista Prelude (*) I will include this functionality in the next version. Thank you for this nice idea. Paul On Sun, Sep 19, 2010 at 08:16:08PM +0200, Paul Brauner wrote: > That's a great idea! > > In the same vein, have you had a look at quickspec by Koen Claessen, > Nicholas Smallbone and John Hughes? > > www.cse.chalmers.se/~nicsma/quickspec.pdf > > This reminds me of another idea, suggested by Jun inoue: look for > functions by specification instead of examples. > > I will try your idea ASAP. As you say, I think that might be helpful for > beginners, as you suggest, or even when you're not a beginner anymore > but you start using a new library. > > Paul > > > On Sun, Sep 19, 2010 at 07:41:21PM +0200, Roel van Dijk wrote: > > Very interesting! > > > > It got me thinking: if you combine this with the Arbitrary class [1] > > of QuickCheck you can use it check if you have defined a function that > > is "equal" to an already defined function. > > > > Let's say I write the following function: > > > > intMul ∷ Integer → Integer → Integer > > intMul x 0 = 0 > > intMul x n = x + intMul x (n - 1) > > > > No you can automatically apply this function to a list of 100 > > generated inputs to get a list of input output pairs. Feed this into > > haltavista and it should tell you that you can replace your definition > > with Prelude (*). While such an observation is certainly not a proof > > it is still useful. > > > > It would be a nice addition to a Haskell editor. Especially for those > > new to the language. > > > > Regards, > > Roel > > > > 1 - http://hackage.haskell.org/packages/archive/QuickCheck/2.3/doc/html/Test-Qu…

1 0

Hackage hierarchical module naming conventions
by Jan Snajder 19 Sep '10

19 Sep '10

Hi folks, I'm almost ready to upload my first-ever package to Hackage, but there's one thing that's bothering me for quite some time now: Are there any hierarchical module naming conventions in Hackage?! Well, I've came across this: http://www.haskell.org/haskellwiki/Hierarchical_module_names This is nice, but obviously (intentionally?) underspecified, and consequently the structure we have now doesn't really seem all that consistent. For example, consider these three packages (just an example, no offense intended): Data.KMeans Data.Datamining.Clustering.Gsom Data.Clustering.Hierarchical (I actually prefer Data.Clustering.*) It seems like anyone is building his or her own taxonomies of modules, and we end up with a lot of space in between. Has this issue been brought up before? Are there any additional guidelines? Best, Jan

1 0

how to measure memory usage?
by Johannes Waldmann 19 Sep '10

19 Sep '10

How could I measure the amount of memory that my ghc-compiled program is actually using, from inside the program? It is expanding nodes in a (game) tree, and I need to stop this when it gets near the physical limits of the system (or some other limit that I want to set on the command line) The node size is varying so I cannot just pre-compute a maximal number of nodes. Thanks - J.W. PS: the program is this: http://www.littlegolem.net/jsp/info/player.jsp?plid=22146

1 1

Full strict functor by abusing Haskell exceptions
by Maciej Piechotka 18 Sep '10

18 Sep '10

I started experiment with strict functors. I come to: > import Control.Exception > import Foreign > import Prelude hiding (catch) > > data StrictMonad a = StrictMonad a deriving Show > > instance Functor StrictMonad where > f `fmap` StrictMonad v = return $ f v > > instance Applicative StrictMonad where > pure = return > (<*>) = ap > > instance Monad StrictMonad where > return x = unsafePerformIO $ do > (return $! x) `catch` \(SomeException _) -> return x > return $! StrictMonad x > StrictMonad v >>= f = f v It seems to be valid IMHO Functor and Monad (I haven't prove it) as long as functions terminates. Some time ago there was post stating that there is not possible strict 'interesting' functor - I guess that the above is 'interesting' (and due to halting problem I guess it is not possible to create strict Functor which would deal with that problem). Regards

8 12

"malicious" JS on haskell site
by Keith Sheppard 18 Sep '10

18 Sep '10

Hello cafe, Maybe malicious isn't the right word but there is a JS based web counter on http://www.haskell.org/complex/why_does_haskell_matter.html which likes to show pop up adverts. They must have switched over from counting visitors to showing adverts at some point since the web page was created. I'm sure this is something we want removed... It would probably be good to grep the whole server to make sure that this counter doesn't show up on some other page too. Sorry if this is a simple oversight on my part but I don't know who I should be contacting about this kind of thing. It would be great if there was some kind of webmaster(a)haskell.org address that was advertised or some form that we could submit for issues like this. -keith -- keithsheppard.name

3 3

translating bidirectional fundeps to TFs
by Gábor Lehel 18 Sep '10

18 Sep '10

Now that the new typechecking awesomeness has been committed to GHC HEAD, theoretically all of the things one can express with FunctionalDependencies (bar overlap) should now be expressible with TypeFamilies as well. This got me to thinking how this might actually be done. Here's an example of a previously-inexpressible class I might want to translate, which functional dependencies allow to be expressed fairly neatly: class TwoPhase frozen thawed | frozen -> thawed, thawed -> frozen where thaw :: frozen -> thawed freeze :: thawed -> frozen (Vaguely inspired by things I saw in the Vector API; it's an abstraction for types which have the same internal representation but can present either a pure, functional interface or a mutable, monadic one and be converted between the two. For the purposes of this message this is just an example.) As far as I can tell, the straightforward translation of this to type families and equality constraints would be like this: class (Frozen (Thawed a) ~ a) => Immutable a where type Thawed a :: * thaw :: a -> Thawed a freeze :: Thawed a -> a type family Frozen a :: * But I don't think this looks too nice. It loses the nice symmetry of the FD version. (You could take both type families outside the class to make it a bit less asymmetric, among other things, but I don't see a way to avoid the fact that you need to distinguish one end of the mapping and define the other in terms of it.) What I would *want* to write is this: class (Mutable (Thawed a), Frozen (Thawed a) ~ a) => Immutable a where type Thawed a :: * thaw :: a -> Thawed a class (Immutable (Frozen a), Thawed (Frozen a) ~ a) => Mutable a where type Frozen a :: * freeze :: a -> Frozen a This looks considerably nicer, but it's not valid -- I get a cycle in class declarations via superclasses error, on the one hand, and (if I remove that) an infinite loop in the typechecker (context reduction stack overflow) on the other (FlexibleContexts being a prerequisite for most of the interesting things you can do with TypeFamilies). So what should one, theoretically, do here? Is there a way to express this using TypeFamilies which preserves both the semantics and the symmetry of the FDs version, and is also valid? Or should one just go ahead and use the FDs because, in this case, they're simply better suited? *Or* should one just suck it up and go with the ugly TypeFamilies version because TypeFamilies are The Future*? * The abbreviation TF not being a coincidence. ~g -- Work is punishment for failing to procrastinate effectively.

2 3

ANN: happstack-auth-0.2
by Nils Schweinsberg 18 Sep '10

18 Sep '10

Hey! I'd like to announce the release of happstack-auth-0.2 [1]. It offers an easy (and secure) way to implement user authentication for Happstack web applications with a lot of high level functions and a solid Happstack-State back-end. The original project got started by MightyByte, until I took it over and improved the whole package. As a result, the API has changed a lot and if you were using the old happstack-auth you might consider rewriting your application to use the new api.* However, your applications state will be completely compatible with the new version and be migrated automatically. The package got uploaded to hackage and can now be installed via cabal: cabal install happstack-auth It should build with both, the current stable release of happstack (0.5.0.2) and the latest darcs version (0.5.1). If there are any errors/bugs during building/running, please contact me! A demonstration website is up and running with about 30 lines of code (+ html templates and the actual happstack server code) at [2]. If you don't want to wait until hackage rendered the haddock pages, you can take a look at the (latest git version of the) API at [3]. Hope you enjoy it! :) - Nils Schweinsberg / McManiaC * You can still use the old state-query/update constructors by importing "Happstack.Auth.Internal", although you'll miss a lot of features like session management etc. In addition, MightyBytes old repo should be still available at [4]. [1] http://hackage.haskell.org/package/happstack-auth [2] http://n-sch.de/happstack-auth [3] http://n-sch.de/hdocs/happstack-auth [4] http://github.com/mightybyte/happstack-auth

3 3

Uniting graphics libraries
by Tillmann Vogt 18 Sep '10

18 Sep '10

Hi, I have nearly finished writing a Collada-output library to make 3d animation files. Before doing this I have looked at two libraries on hackage that parse Collada to display it. For my library I have copied the types from gpipe-collada and extended them. Now I am thinking that it would be nice if all 3d libraries would use the same data type. Currently gpipe-collada, graphics-formats-collada, obj, ... parse 3d-formats. FieldTrip, lambdacube-engine, ... (to mention examples) can generate and/or display 3d-objects. But they all cannot be combined without tedious type conversions. The obvious argument is that every application has its own needs. This is of course true but I think that the Collada format shows how to deal with this (you can do quite a lot with profile_common). I would be willing to write patches for some libraries and make a discussable 3d-types library based on Collada. Hope you agree, Tillmann

2 4