- Haskell-Cafe - Haskell.org

ANN: psqueues
by Jasper Van der Jeugt 26 Jun '14

26 Jun '14

Here at Better, we have just a released a package which implements faster priority search queues for Haskell. We hope it proves to be useful to the community! More information can be found in this blogpost: http://medium.com/@bttr/announcing-psqueues-8a0fe9fe939 Grab it from Hackage or GitHub: http://hackage.haskell.org/package/psqueues https://github.com/bttr/psqueues On behalf of the engineering team at Better, Jasper Van der Jeugt

3 2

Quantification in Instance Contexts
by Julian K. Arni 26 Jun '14

26 Jun '14

Hi Cafe, I'm playing around with simple logic-programming at the type level. For instance, encoding trees: >> {-# LANGUAGE MultiParamTypeClasses >> , FunctionalDependencies >> , FlexibleInstances >> , UndecidableInstances >> , FlexibleContexts >> , OverlappingInstances #-} >> {-# OPTIONS_GHC -fcontext-stack=100 #-} >> >> >> -- *A >> -- / \ >> -- B* *C >> -- | >> -- D* >> >> >> data A >> data B >> data C >> data D >> class Child a b bool | a b -> bool >> instance Child A B TrueT >> instance Child B D TrueT >> instance Child B C TrueT >> class Path a b bool | a b -> bool Now the following obviously doesn't work (never mind for now that 'Path' needs a recursive definition, and that this is really just 'Grandchild'): >> instance (Child a b TrueT, Child b c TrueT) => Path a c TrueT Because 'b' is ambiguous. Fair enough. But I can't directly use a fundep, because 'b' *is* in fact ambiguous. What I want to tell the compiler is that it's really okay, since the RHS side (and any possible 'where' expressions) doesn't depend on what is instantiated for 'b'. I know I could switch to a class 'Children a bs bool' and encode all children of as a type-level list, and then have a fundep between 'a' and 'bs'. That's not a *bad* solution: it does give a good sense of the intention. But it's not very extensible; really I'd rather be using something like 'forall' - something, I would guess, along the lines of: >> instance forall a c. (forall b. (Child a b TrueT, Child b c TrueT)) => Path >> a c TrueT That, however, fails with: Malformed instance head: (forall b. (Child a b TrueT, Child b c TrueT)) -> Path a c TrueT In the instance declaration for ‘Path a c TrueT’ So: is there a way (presumably with 'forall') of telling the compiler that the ambiguous type 'b' actually won't "leak through" to the RHS, so stop bugging me about ambiguity? Thanks, Julian

4 7

direct-fastcgi and logging
by Denis Shevchenko 26 Jun '14

26 Jun '14

Hi all! I write a FastCGI-server based on excellent package 'direct-fastcgi' (http://hackage.haskell.org/package/direct-fastcgi) It works perfectly, but I want to inject a logging facility into a handler of requests. For example: import Control.Concurrent (forkIO) import Network.FastCGI main :: IO () main = acceptLoop forkIO handleRequest handleRequest :: FastCGI () handleRequest = do -- >>> In this place I want to log a message <<< setResponseStatus 400 setResponseHeader HttpContentType "application/json" fPutStr "{ "my": "problem_description" }" Yes, I know about this function in 'Network.FastCGI': fLog :: MonadFastCGI m => String -> m () but it "logs a message using the web server's logging facility", and I want to log into _my_own_ file. So which solution I can use here? I'd be very grateful for any help. ---------- Sincerely, Denis Shevchenko haskell(a)dshevchenko.biz

2 1

BNFC-meta being maintained? Is there public src repo?
by Ki Yung Ahn 26 Jun '14

26 Jun '14

Dear BNFC-meta developers and users, I've been using BNFC-meta for the parser of my experimental language Nax language implementation ( https://github.com/kyagrd/mininax ). It's very cool 'cause you don't need to write makefile or cabal entries to call the command line tools (bnfc, alex, happy), and you get quasiquoters for free, which really makes the early development pleasant. But due to the bug fixed in recent versions of BNFC, probably after latest BNFC-meta release, I'll have to switch back to using plain BNFC at some point. (FYI,the bug is that block comments arn't working). If there is a source repository that has ported more recent versions of BNFC into BNFC-meta, it'll be helpful for people like me. Or, if the source repository has open access and easy to participate (e.g., putting on github or something like that), we can draw community support maintaining BNFC-meta to keep up with the recent BNFC (and also happy and alex) versions. -- Ki Yung Ahn

2 1

determine hoolge database directory
by Tobias Pflug 25 Jun '14

25 Jun '14

Hi haskellers, i'm new to haskell and am currently working on some tool that is supposed to fiddle with hoogle database files. Now I'm looking for a sane way to determine the directory where hoogle stores its database directories (when you perform `hoogle data`). In my case this is: .cabal/share/x86_64-linux-ghc-7.8.2/hoogle-4.2.33/databases So this is [home directory]/.cabal/[ghc-version]/[hoogle-version]/databases Could someone suggest me a sane way how I can determine this path at runtime ? Maybe there is a way to extract this from somewhere ? I'm out of ideas on this. I could of course take the easy way out and let users declare the path as config parameter but it would be nice if there was a way ... thanks in advance. regards, Tobi

2 1

accessing a ByteArray from FFI
by Mihaly Barasz 25 Jun '14

25 Jun '14

Hello, I want to access the contents of a Data.Vector.Primitive from FFI. I came up with the following so far: {-# LANGUAGE ForeignFunctionInterface #-} module FFIExample (exampleFn) where import Control.Monad.Primitive (touch) import Control.Monad.ST (runST) import Data.Primitive.Addr (Addr(..)) import Data.Primitive.ByteArray (mutableByteArrayContents) import qualified Data.Vector.Primitive as P import qualified Data.Vector.Primitive.Mutable as PM import GHC.Ptr (Ptr(..), plusPtr) import Data.Word (Word64) foreign import ccall unsafe "my_external_fn" myExternalFn :: Ptr Word64 -> Int -> Int exampleFn :: P.Vector Word64 -> Int exampleFn v = runST $ do PM.MVector off len mba <- P.unsafeThaw v let ptr = case mutableByteArrayContents mba of Addr addr -> Ptr addr `plusPtr` off result = myExternalFn ptr len result `seq` touch mba return result This seems to work, but I have questions that I haven't been able to find the answers to on my own: 1. Is this actually OK to do? I guess that this wouldn't be OK if the foreign import were "safe", as then the GC could move the contents of the byte array while the foreign function is running. But is it safe like this? 2. If the answer to the previous question is no, then is there a way to do it properly? Or there is just no way to pass an unpinned byte array to a foreign call? What about foreign import prim? 3. If the answer to Q1 is no, then would it be OK if the underlying byte array were pinned? 4. Any other simplifications? Pointers to resources on these topics would be more than welcome! But, I haven't been able to find any. Thanks, Mihaly

2 4

Fwd: 自动回复：Re: ANN: hmatrix-0.16
by Dominic Steinitz 25 Jun '14

25 Jun '14

Hello Cafe, Everytime I post something via gmane I get what I assume is spam - see below. Is this a gmane feature or a haskell-cafe feature? Dominic Steinitz dominic(a)steinitz.org http://idontgetoutmuch.wordpress.com Begin forwarded message: > From: <zhangliu529(a)sina.com> > Subject: 自动回复：Re: [Haskell-cafe] ANN: hmatrix-0.16 > Date: 22 June 2014 11:37:49 BST > To: <dominic(a)steinitz.org> > > 您发的邮件我已收到，我将尽快给您答复。

5 4

ANN: hmatrix-0.16
by Alberto Ruiz 24 Jun '14

24 Jun '14

Hi! I am happy to announce the release of hmatrix-0.16, a Haskell package for matrix computations and numeric linear algebra. New features: - BSD3 license. The modules depending on GSL have been moved to a new package hmatrix-gsl (GPL). Now hmatrix depends only on BLAS/LAPACK. - Simpler reexport modules with improved documentation and usage examples. There are a few API changes but the traditional modules are also exposed for backwards compatibility. - Alternative interface using type-level literals for static dimension checking and inference (work in progress). - Initial support for sparse linear systems. - Minor improvements and bug fixes (see the changelog for details). http://hackage.haskell.org/package/hmatrix https://github.com/albertoruiz/hmatrix Some examples: http://dis.um.es/~alberto/hmatrix/hmatrix.html Suggestions, contributions, and bug reports are welcome. Thanks! Alberto

4 4

ANN: Graphics.OpenSCAD
by Mike Meyer 24 Jun '14

24 Jun '14

Graphics.OpenSCAD provides an algebraic data type for creating OpenSCAD 3d models, as well a a function to generate an OpenSCAD program from a data structure. OpenSCAD is a "programmers" CAD tool. It provides a declarative language for 3d models with a primitive functional language for doing computation. The Graphics.OpenSCAD module tries to preserve the declarative parts by providing tools for building data structures that match the OpenSCAD declarative models, then embedding that in a powerful, modern functional language. The critical problem with OpenSCAD is the error handling. At it's best, it will say "syntax error" and provide sufficient location information to easily find the error. At times, the location information points outside the file. Errors in dealing with variables may have no error messages, or messages that are easy to overlook, resulting in models that are radically different from what was expected with no way to determine the problem except searching the complete program text. Graphics.OpenSCAD attempts to solve that by leveraging Haskell's type system as far as possible. While not every run-time OpenSCAD error can be detected in Haskell, as many as possible are. Hackage page with pointers to source and documentation: http://hackage.haskell.org/package/OpenSCAD

1 0

how to upgrade to a new version of Haskell?
by Johan Larson 24 Jun '14

24 Jun '14

I am currently running an old installation of the Haskell environment from 2012, and I would like to upgrade. Can I just grab the installer from https://www.haskell.org/platform/windows.html and run it? This is on Windows. -- Johan Larson -- Toronto, Canada

4 4