August 2010 - Haskell-Cafe

A cabal odyssey
by Andrew Coppin 22 Aug '10

22 Aug '10

And so today, just for giggles, I tried to get Sifflet to work. Along the way, I encountered a number of... "glitches", if you will. First of all, I tried to get it to work on Windows. I fired up a new Windows VM and installed Haskell Platform 2010.1.0.0. It seems that (finally) this includes the cabal-install tool, which is nice. It seems it includes cabal-install 0.8.0, so as soon as I tried "cabal update", it tells me a new version is available. It claims I just need to "cabal install cabal-install" (which is amusingly riddle-like). Unfortunately, although doing this *does work*, the new cabal.exe is installed lower down the search path than the existing one, so you still get the old version (unless you manually fiddle with the search path). In fact, it seems that the HP install folder is higher up the search path than the "global" binary target, which is higher than the "local" binary target. It seems to me like this ought to be the other way around. The next problem is that "cabal install sifflet" gets mighty confused and outright fails. Basically it can't figure out how to resolve all the dependencies. It seems that Sifflet demands GTK 0.11.0 (i.e., gtk-0.11.0, pango-0.11.0, glib-0.11.0, etc.) However, Cabal looks at gtk-0.11.0, sees it depends on (say) "glib-0.11.*", and wants to use glib-0.11.1 (the latest one). But then Sifflet wants glib-0.11.0, not glib-0.11.1, and it seems Cabal just can't figure out what the heck to do. Which is slightly surprising, really. The solution (of sorts) is to painstakingly resolve the dependencies by hand, by asking Cabal to install the correct packages one at a time in the correct order. (I still love the way Gtk2hs *actually compiles* on Windows now. That's pretty sweet!) That reminds me. What the heck is actually *in* file 126? I don't know why, but compiling file 126 (Graphics.UI.Gtk.Gdk.Cursor) takes up 75% of the entire Gtk2hs build time! What's that all about? Anyway, having finally built Gtk2hs version 0.11.0 successfully, I continued trying to get Sifflet working... only to discover it wants the "curl" package. And when I ask Cabal to build it, it just retorts that it has a configure script. *sigh* So that's the end of that. I have absolutely no idea why a tool like Sifflet would need access to the Curl library. Presumably this is just another one of those obscure dependency-chasing artifacts that happen from time to time? (After dealing with Linux, I'm used to this kind of weirdness.) OK, so it's only possible to run Sifflet under Linux. Let's give that a try... So I fire up a new OpenSUSE VM. I quickly discovered that gtk seems to want Alex and Happy, but cabal-install is defaulting to doing a "local" rather than "global" install, and thereafter it can't "find" Alex or Happy, even though they're installed. (Yeah, great, thanks for that...) So I rewind the VM, build the latest version of cabal-install, and edit the configuration to do global installs instead of local. The amusing part is, if you "sudo cabal install" so it has permission to put the installed files into place, it then uses root's configuration file instead. *sigh* Well anyway, I managed to work around that. But... Cabal *still* fails to find Alex or Happy, even though they're now in the search path. Oh, wait. They're in *my* search path. They're not in root's search path. (As per good security practise, root's search path is rather short.) I can pass some CLI switches to tell it where these are, but then gtk2hs-buildtools makes a whole bunch of stuff which Cabal also can't find. Eventually, the easiest thing I could come up with was to do "cabal unpack" to get a source tree, configure and build as me, and sudo for the install. Except that then it tries to reconfigure...? Wuh? So instead of "sudo cabal install", I tried "sudo runhaskell Setup install", which works just fine (although obviously it's rather wordy!) So, I manually hold Cabal's hand through the process of building all the 0.11.0 packages one at a time, in a way that it can find all the stuff at configure-time. Gosh this is a faff! Well anyway, it worked. Trips over when I reach Curl, but that's because I need to ask YaST to install the curl-devel package. And, finally, I can build and install Sifflet itself. After 4 hours or so of typing commands, it was nice to do some stuff with my mouse. ;-) I spent about 20 minutes trying to figure out how to build high-order functions before I discovered that you can't. Still, the stated design goal of Sifflet is "to teach beginners about recursion". Clearly it's design achieves this goal quite well. Equally clearly, Haskell has damaged my mind. I actually cannot think of *any* interesting low-order functions! 5 years of programming in Pascal with low-order *monomorphic* functions, and today I can't think of a single interesting use for such a thing. ;-) Heh, well anyway... I don't know how many of the things I've mentioned are known or likely to be fixed. Just thought I'd share.

10 32

Re: [Haskell-cafe] A cabal odyssey
by Mathew de Detrich 22 Aug '10

22 Aug '10

What haskell really needs is a big ' hit' where someone shows how powerful haskell is in some environment. Ruby had RoR, which is what splurged the language into the open On 22/08/2010 3:46 PM, "wren ng thornton" <wren(a)freegeek.org> wrote: Daniel Fischer wrote: > > On Saturday 21 August 2010 15:35:08, Ivan Lazar Miljenovic wrote: >> >> A ... Three! -- Live well, ~wren _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe(a)haskell.org ...

1 0

GADT and problems with rigid type variables
by Markus Barenhoff 22 Aug '10

22 Aug '10

Hello, playing with GADTs I ran into a problem with rigid type variables which is ilustrated by the following example. I think it should be pretty clear what I'am trying to express... Any suggestions? ---- snip ---- {-# LANGUAGE GADTs #-} data Foo where Foo :: (Eq t) => t -> Foo instance Eq Foo where (Foo a) == (Foo b) = a == b {- Scratch.hs:7:28: Couldn't match expected type `t' against inferred type `t1' `t' is a rigid type variable bound by the constructor `Foo' at /home/alios/src/lab/Scratch.hs:7:3 `t1' is a rigid type variable bound by the constructor `Foo' at /home/alios/src/lab/Scratch.hs:7:14 In the second argument of `(==)', namely `b' In the expression: a == b In the definition of `==': (Foo a) == (Foo b) = a == b Failed, modules loaded: none. -} ---- snip ---- thnx Markus

3 2

Error in enumerator when using interpreter instead of compiler
by Tobias Brandt 22 Aug '10

22 Aug '10

Hi all, I was trying out the enumerator package. I wanted to copy the contents of one file to another: module Main where import Data.Enumerator import Data.Enumerator.IO main = run (enumFile "foo" $$ iterFile "bar") If I compile this code with GHC, it works as expected. But if I run it with runhaskell I get the following error: Left bar: hPutBuf: illegal operation (handle is closed) If I replace the above code with this: module Main where import Data.Enumerator import Data.Enumerator.IO import System.IO main = withBinaryFile "bar" WriteMode $ \h -> run (enumFile "foo" $$ iterHandle h) It works both compiled and interpreted. Is this a bug in enumerator (probably in the iterFile function) or a deficiency of the interpreter?

2 3

Perform a research a la Unix 'find'
by Yves Parès 22 Aug '10

22 Aug '10

Hello, I would like to recode in Haskell a piece of bash program that uses find. The advantage of find is that it is quite easy to use and fast. Is there a package that let us access to find-like functionnalities, with similar performances? Or at least some C functions that could be used through the FFI? (I would like to avoid to recur through directories and match files myself, I tried it and the search lasts for ages).

3 6

Re: [Haskell-cafe] Code that writes code
by Andrew Coppin 22 Aug '10

22 Aug '10

Graham Klyne wrote: > Maybe not helpful to you at this stage, but... > > An alternative to generating source code is to factor out the common > "boilerplate" elements into separate functions, suitably > parameterized, and to use higher order functions to stitch these > together. Well, yeah, if you've got so much boiler plate that you have to automate generating the boilerplate, you're probably doing it wrong. ;-) All I'm actually using it to do is generate a set of fixed-size containers (each of which has a bazillion class instances). I've got a variable-sized container, but sometimes it's useful to statically guarantee that a container is a specific size. In addition, by being fixed-size you can get a few small performance gains. That's really all I'm using autogeneration for. I suppose instead of building an ADT for each container size, I could just write a newtype over the variable-size container and put a phantom type on it representing the size... That would give me the static guarantees but not the efficiency.

3 3

Re: [Haskell] Re: [Haskell-cafe] ANNOUNCE: enumerator, an alternative iteratee package
by John Millikin 21 Aug '10

21 Aug '10

On Sat, Aug 21, 2010 at 14:41, Michael Snoyman <michael(a)snoyman.com> wrote: > Hey John, > As I mentioned, I'm considering having persistent depend upon enumerator. Do > you think it's too early in enumerator's life to do so and I should wait > till the API stabilizes a bit more? Also, two other packages I would think > to bring into the enumerator family would be: > * yaml > * wai-extra, providing an enumerator layer for more easily dealing with the > Source and Enumerator datatypes in wai. I might just release a > wai-enumerator package instead. > Thanks again for your work on this, > Michael I think the API is pretty stable. Most of the significant research into iteratee-based APIs has already been performed by users of the "iteratee" library, and by Oleg. There might be a few backwards-compatible changes (new modules, new exports, etc). I'm not planning to make any large changes, such as Mr. Lato's transition to CPS-based iteratees. As long as you import the enumerator modules with "qualified" (to avoid Prelude name clashes), it should be safe to start porting libraries.

2 2

Hugs & Haskore loading
by Hans Aberg 21 Aug '10

21 Aug '10

When trying to load Haskore in Hugs Sept 2006 (say by :l), it does not work because it is in a subdirectory Haskore/Src/, despite that the Hugs searchpath includes /usr/local/lib/hugs/packages/* I could of course change this searchpath, but I'm curios about how to fix it without it - it seems to work for a lot of other packages.

2 8

Equality constraints and RankNTypes - how do I assist type inference
by DavidA 21 Aug '10

21 Aug '10

Hi, I have the following code, using equality constraints and (I believe) RankNTypes: {-# LANGUAGE MultiParamTypeClasses, TypeFamilies, RankNTypes, ExistentialQuantification #-} {-# LANGUAGE FlexibleInstances, TypeSynonymInstances #-} -- import Math.Algebra.Group.PermutationGroup -- Vector space over field k with basis b data Vect k b = V [(b,k)] deriving (Eq,Show) data TensorBasis a b = T a b deriving (Eq, Ord, Show) -- Tensor product of two vector spaces type Tensor u v = (u ~ Vect k a, v ~ Vect k b) => Vect k (TensorBasis a b) -- ** class Algebra k v where -- "v is a k-algebra" unit :: k -> v mult :: Tensor v v -> v type GroupAlgebra k = Vect k Int -- (Permutation Int) instance Num k => Algebra k (GroupAlgebra k) where unit 0 = V [] unit x = V [(1,x)] mult (V ts) = V [(g*h,x) | (T g h, x) <- ts] Everything is fine except for the last line, which causes the following error message: Couldn't match expected type `Tensor (GroupAlgebra k) (GroupAlgebra k)' against inferred type `Vect k1 b' In the pattern: V ts In the definition of `mult': mult (V ts) = V [(g * h, x) | (T g h, x) <- ts] In the instance declaration for `Algebra k (GroupAlgebra k)' But according to me, I've told it that these two types are the same at the line marked -- ** How do I help it out with type inference? (It, in this case, is GHCi 6.12.1) Any ideas?

3 3

System.Posix, forked processes and psuedo terminals on Linux
by Erik de Castro Lopo 21 Aug '10

21 Aug '10

Hi all, I've got a but it code below thats not quite working as I expect it to. The basic idea is that it opens a master/slave pair of pseudo terminals, forks a child process and then perform bi-directional communication between the parent and the child via the master/slave psuedo terminal. The thing does really need to forkProcess because once the comms is working I want to exec another process in the child. I also intend to dup the file descriptors so that stdin, stdout and stderr all point to the slave's end of the pty. The code below *almost* works. Its currently printing out: parent : Forked child was here! parent : Message from parent. Read 21 bytes while I think it should print: parent : Forked child was here! parent : Read 21 bytes Any clues on why 'Message from parent.' is also ending up on stdout? This is ghc-6.12.1 on Debian Linux. Cheers, Erik import System.Posix.IO import System.Posix.Process import System.Posix.Terminal import System.Posix.Types main :: IO () main = do (master, slave) <- openPseudoTerminal _childId <- forkProcess $ forkedChild (master, slave) closeFd slave runParent master runParent :: Fd -> IO () runParent fd = do (str, _) <- fdRead fd 1024 putStr $ "parent : " ++ str _ <- fdWrite fd "Message from parent.\n" (str2, _) <- fdRead fd 1024 putStr $ "parent : " ++ str2 forkedChild :: (Fd, Fd) -> IO () forkedChild (master, fd) = do closeFd master _ <- fdWrite fd "Forked child was here!\n" (_, count) <- fdRead fd 1024 _ <- fdWrite fd $ "Read " ++ show count ++ " bytes\n" closeFd fd -- ---------------------------------------------------------------------- Erik de Castro Lopo http://www.mega-nerd.com/

2 2