June 2008 - Haskell-Cafe

CouchDB module in Yhc source tree: clarification, and small problems with other packages
by Dimitry Golubovsky 21 Aug '08

21 Aug '08

Hi, Don asked: > Are we likely to see the couchdb bindings released as a standalone > library for the wider community? Let me explain a little bit why the CouchDB module appears in the Yhc repo. As the next stage of the Yhc/Javascript project, I am trying to set up a web service where people might upload some Haskell code (or reference to such code, these details haven't been clarified yet), and get it compiled by background Yhc/ycr2js batch job, returning a HTML page with Javascript in response. CouchDB was picked* as intermediate storage for this project. Therefore an interface module was needed, so I wrote one. This module, `Database.CouchDB' is not 100% finished; it was coded up to satisfy the needs of the first utility I wrote, `cdbup' which is currently capable of uploading a text (utf-8 expected) or a binary file as a document attachment to CouchDB. So some functionality hasn't been coded yet, although it will definitely be, as the project develops. Regarding packaging, I hope I will eventually put it on Hackage, but there are some problems with another packages which I'll try to explain here, so maybe those packages' developers suggest something. 1. Text.JSON. This module was picked from Jun Mukai's haskellnet package (http://darcs.haskell.org/SoC/haskellnet/) One problem with this package, it needs too many dependencies (while JSON module relies only on Packrat parser). I had to extract the JSON module into Yhc tree; can the haskellnet package be split somehow into orthogonal sub-packages, and maybe one toplevel package which would pull all parts together if used? 2. Text.JSON again. I had to change the pNumber function (parser for numeric nodes) replacing the '1'..'9' digits range with '0' .. '9'. Otherwise it did not handle JSON-encoded numbers equal to zero. I am not sure I was 100% right, so perhaps some JSON experts here might correct me. 3. Packrat parser does not seem to be packaged anywhere else, does it? 4. The dataenc package (http://code.haskell.org/dataenc/devo/) from which I pulled the Codec.Binary.Base64 module. The package may need some adjustment to GHC 6.8.x packages structure as only "base" listed in build-depends does not seem to be enough: when building it, Cabal complains for hidden packages such as collections (which I had to add manually to the cabal file), etc. Is it some problem with my GHC setup, or does just the dataenc.cabal need to be updated? So, in order to get things done quicker, I pulled the said pieces from their packages into Yhc tree. For packaging the CouchDB interface, definitely these duplicates need to be removed. Other packages that CouchDB relies upon, are**: base, network, HTTP, filepath, base64-string, utf8-string Thanks for the interest in CouchDB interface, and feel free to ask any questions. --- * before support form IBM/Apache was announced, but this recent development definitely increases interest to CouchDB ** Just from the -package options I used to compile it; I may have something omitted here. -- Dimitry Golubovsky Anywhere on the Web

5 5

Aren't type system extensions fun?
by Andrew Coppin 23 Jul '08

23 Jul '08

Haskell 98 provides a simple and clean type system, which I feel I understand very well. GHC provides a vast zoo of strange and perplexing type system extensions, which I don't understand at all. (Well, some of it is simple enough - e.g., multiparameter type classes. But GADTs? FDs? ATs? Hmm...) Anyway, it seems there is a large set of such type system extensions that involve writing "forall" all over the place. I have by now more or less managed to comprehend the fact that data Thing = forall x. Thing x allows a type variable to appear on the RHS that is *not* present on the LHS, thus "hiding" the type of something inside the structure. And for some reason, they call this "existential quantification" [which I can't spell never mind pronounce]. Today I was reading a potentially interesting paper, and I stumbled across something referred to as a "rank-2 type". Specifically, class Typable x => Term x where gmapT :: (forall y. Term y => y -> y) -> x -> x At this point, I am at a complete loss as to how this is any different from gmapT :: Term y => (y -> y) -> x -> x Can anybody enlighten me? This is probably the first real use I've ever seen of so-called rank-2 types, and I'm curios to know why people think they need to exist. [Obviously when somebody vastly more intelligent than me says something is necessary, they probably know something I don't...] At this point, I don't think I even wanna *know* what the hell a rank-N type is... o_O

23 42

NW Functional Programming Interest Group
by Greg Meredith 18 Jul '08

18 Jul '08

All, Apologies for multiple listings. A small cadre of us are organizing a Northwest Functional Programming Interest Group (hey... NWFPIG, that's kinda funny). Our first official meeting is at the The Seattle Public Library 1000 - 4th Ave. Seattle, WA 98104 from 17:00 - 18:00 on February 20th. On the first meeting's agenda we'll be - giving people who are interested in or actively using FP for work or play a chance to meet - seeking to build up a pipeline of presentations and guest speakers - trying to keep organizational mishigosh to a minimum Hope to see you there. Monadically yours, --greg -- L.G. Meredith Managing Partner Biosimilarity LLC 806 55th St NE Seattle, WA 98105 +1 206.650.3740 http://biosimilarity.blogspot.com

2 7

ANN: Topkata
by c-bauer-olsbruecken＠t-online.de 16 Jul '08

16 Jul '08

Hi All, Topkata is a simple OpenGL Game written in Haskell. It's not very advanced. Goal so far is to guide a ball trough an labyrinth to the opposite corner. The web page shows an screenshot. It's only tested under Linux. http://home.arcor.de/chr_bauer/topkata.html Feedback&Patches are welcome. BTW, I had a lot of support in #haskell! Christoph Bauer

14 19

What is a rigid type variable?
by Xiao-Yong Jin 08 Jul '08

08 Jul '08

Hi all, I'm writing a short function as follows, but I'm not able to find a suitable type signature for `go'. It uses Numeric.LinearAlgebra from hmatrix. -- | Map each element in a vector to vectors and thus form a matrix -- | row by row mapVecToMat :: (Element a, Element b) => (a -> Vector b) -> Vector a -> Matrix b mapVecToMat f v = fromRows $ go (d - 1) [] where d = dim v go :: Element b => Int -> [Vector b] -> [Vector b] go 0 vs = f (v @> 0) : vs go !j !vs = go (j - 1) (f (v @> j) : vs) If I give the type signature to go as this, I got the following error Couldn't match expected type `b1' against inferred type `b' `b1' is a rigid type variable bound by the type signature for `go' at test.hs:36:20 `b' is a rigid type variable bound by the type signature for `mapVecToMat' at test.hs:31:35 Expected type: Vector b1 Inferred type: Vector b In the first argument of `(:)', namely `f (v @> 0)' In the expression: f (v @> 0) : vs So what is this rigid type variable all about and what is correct type of the function `go'? Thanks in advance, X-Y -- c/* __o/* <\ * (__ */\ <

5 7

A Monad for on-demand file generation?
by Joachim Breitner 07 Jul '08

07 Jul '08

Hi, for an application such as a image gallery generator, that works on a bunch of input files (that are assumed to be constant during one run of the program) and generates or updates a bunch of output files, I often had the problem of manually tracking what input files a certain output file depends on, to check the timestamps if it is necessary to re-create the file. I thought a while how to do this with a monad that does the bookkeeping for me. Assuming it’s called ODIO (On demand IO), I’d like a piece of code like this: do file1 <- readFileOD "someInput" file2 <- readFileOD "someOtherInput" writeFileOD "someOutput" (someComplexFunction file1 file2) only actually read "someInput" and "someOtherInput", do the calculation and write the output if these have newer time stamps than the output. The problem I stumbled over was that considering the type of >>= (>>=): Monad m => m a -> (a -> m b) -> m b means that I can not „look ahead“ what files would be written without actually reading the requested file. Of course this is not always possible, although I expect this code to be the exception: do file1 <- readFileOD "someInput" file2 <- readFileOD "someOtherInput" let filename = decideFileNamenameBasedOn file2 writeFileOD filename (someComplexFunction file1 file2) But assuming that the input does not change during one run of the program, it should be safe to use "unsafeInterleaveIO" to only open and read the input when used. Then, the readFileOD could put the timestamp of the read file in a Monad-local state and the writeFileOD could, if the output is newer then all inputs listed in the state, skip the writing and thus the unsafeInterleaveIO’ed file reads are skipped as well, if they were not required for deciding the flow of the program. One nice thing is that the implementation of (>>) knows that files read in the first action will not affect files written in the second, so in contrast to MonadState, we can forget about them, which I hope leads to quite good guesses as to what files are relevant for a certain writeFileOD operation. Also, a function cacheResultOD :: (Read a, Show a) => FilePath -> a -> ODIO a can be used to write an (expensive) intermediate result, such as the extracted exif information from a file, to disk, so that it can be used without actually re-reading the large image file. Is that a sane idea? I’m also considering to use this example for a talk about monads at the GPN¹ next weekend. Greetings, Joachim ¹ http://entropia.de/wiki/GPN7 -- Joachim "nomeata" Breitner mail: mail(a)joachim-breitner.de | ICQ# 74513189 | GPG-Key: 4743206C JID: nomeata(a)joachim-breitner.de | http://www.joachim-breitner.de/ Debian Developer: nomeata(a)debian.org

10 19

ANN: Gtk2Hs 0.9.13 released
by Peter Gavin 07 Jul '08

07 Jul '08

Hello everyone, Gtk2Hs version 0.9.13 is now available. [1] New features: * bindings for Gnome VFS and GStreamer * a new Gtk+ tutorial has been adapted by Hans van Thiel * cairo image stride support * many new demos * compiles with GHC 6.8.3 * lots of bug fixes This release has been tested on a variety of platforms with different versions of Gtk+ and GHC, so you should have no trouble compiling it if you're using an older version of Gtk+. Note that the binaries for Win32 for this release are only provided for GHC 6.8.3 and Gtk+ 2.12. As with older releases, all the C libraries needed are included in the installer, so you don't need to download anything else to get up an running. I've also created zip files containing only the C libraries that can be used for redistribution. The sources for these binaries are available at [2] Links: [1] http://sourceforge.net/project/showfiles.php?group_id=49207&package_id=42440 [2] http://ftp.gnome.org/pub/GNOME/sources/ -- Pete

2 2

[WARN] Bug fix release of pureMD5 (Was: pureMD5)
by Thomas M. DuBuisson 06 Jul '08

06 Jul '08

Cafe, Daniel Larsson noticed a correctness issue with the pureMD5 package. This issue would affect you if you built the value incrementally via the 'updateMD5' function (vs just using 'md5') and didn't provide 512 bit long bytestrings (an MD5 block of operation). As you can probably tell, I didn't invest enough into the non-performance aspects of pureMD5. Faced with actual users ;-), I have released version 0.2.0 which has the bug fix, a new API (type prevention from re-finalizing a digest), and a reasonable set of quickchecks (covering Show / Binary instances, known answer and incremented hashing). Oh, also the module name has changed to place it inline with 'Crypto' package naming while not colliding. Sorry if this causes anyone headaches. Daniel, Good catch, I hope this didn't consume much of your time. Thanks a bunch! TomMD On Fri, 2008-06-13 at 04:14 +0200, Daniel Larsson wrote: > Hi Thomas, > > > I was fiddling around with your pureMD5 package, and encountered some > problems with using the md5Update/md5Finalize sequence. I tried to > calculate the md5 of some scattered data, but it kept returning the > wrong values. It seems that each md5Update must supply an exact > blockSize number of bits, since the mdLeftOver part isn't taken into > account in subsequent calls to md5Update. > > > I wrote a small patch, and a simple QuickCheck property, to support > calculating md5 of scattered data, attached to this mail. Hopefully I > didn't mess up something... > > > -- > Daniel

8 11

I/O system brokenness with named pipes
by John Goerzen 05 Jul '08

05 Jul '08

So I have a need to write data to a POSIX named pipe (aka FIFO). Long story involving a command that doesn't have an option to read data from stdin, but can from a named pipe. I have created the named pipe from Haskell no problem. But I can't use writeFile to write data to it. Worse, it returns: *** Exception: /tmp/bakroller.zD0xHj/fifo: openFile: does not exist (No such device or address) which is completely false, as it *does* exist, and I can cat to it as expected. The call should block until everything is read on the remote end. I thought maybe writeFile is being weird, so I tried: openFile "/tmp/bakroller.zD0xHj/fifo" WriteMode Same thing. There is no logical reason I can see for this behavior. In fact, something must be going to *extra* effort to avoid writing to a named pipe, and I can't work out why. Named pipes are a standard, useful part of a system and shouldn't be ignored like this. Interestingly, readFile works fine on a named pipe. What's going on here? Am I going to have to resort to the System.Posix interface just to be able to write to a named pipe? -- John

10 20

Re: [Haskell-cafe] Access to Oracle database from Haskell
by Henning Thielemann 03 Jul '08

03 Jul '08

On Thu, 26 Jun 2008, Alistair Bayley wrote: >> Try this version of configOracle in Setup.hs: >> >> configOracle verbose buildtools = do >> if not (sqlplusProgram `isElem` buildtools) >> then return Nothing >> else do >> path <- getEnv "ORACLE_HOME" >> info verbose ("Using Oracle: " ++ path) >> makeConfig path libDir "oci/include" >> where libDir = if isWindows then "bin" else "lib" >> >> >> You'll also need to add this import: >> >> import System.Environment (getEnv) > > Did you get a chance to try this? I'm quite keen to fix the > non-Windows parts of the installation process, if I can. I have replaced configOracle by configOracle verbose buildtools = do if not (sqlplusProgram `isElem` buildtools) then return Nothing else do path <- getEnv "ORACLE_HOME" info verbose ("Using Oracle: " ++ path) let (libDir, includeDir) = if isWindows then ("bin", "oci/include") else ("lib", "/usr/include/oracle/10.2.0.4/client") makeConfig path libDir includeDir This is obviously a hack. We should get the Oracle include path from the user, maybe via another environment variable or a custom Cabal option. Now I can start GHCi with the example program you gave me. However I have to start with -lclntsh, otherwise symbol OCIEnvCreate cannot by found. I thought I do not need this option, because the installed Takusen package contains the library name: $ grep clntsh dist/installed-pkg-config extra-libraries: clntsh clntsh I wonder why it appears twice, because in dist/setup-config it exists only once: extraLibs = ["clntsh"] Nevertheless I can start GHCi with the example program and I can see 'hello world'! Great - many thanks for the hints that led me to that state! Next step is to fetch real data from the database. But now I encounter new problems. If I write a real 'select' statement I get at best: Unexpected null in row 1, column 1. *** Exception: (unknown) and at worst: 50001752 *** glibc detected *** /usr/lib/ghc-6.8.2/ghc-6.8.2: free(): invalid pointer: 0x08d10065 *** ======= Backtrace: ========= /lib/libc.so.6[0xb7dba4b6] /lib/libc.so.6(cfree+0x89)[0xb7dbc179] /usr/lib/oracle/10.2.0.4/client/lib/libclntsh.so(lfvclose+0x1c)[0xb5df9c1e] /usr/lib/oracle/10.2.0.4/client/lib/libclntsh.so(SlfMunmap+0x3f)[0xb5e70c3b] /usr/lib/oracle/10.2.0.4/client/lib/libclntsh.so(ldiutzd+0x37)[0xb5e49e7b] /usr/lib/oracle/10.2.0.4/client/lib/libclntsh.so(kputerm+0x51)[0xb54c96b9] /usr/lib/oracle/10.2.0.4/client/lib/libclntsh.so(OCITerminate+0x1c)[0xb5596888] [0xb6a33801] /usr/lib/ghc-6.8.2/ghc-6.8.2[0x8a600de] ======= Memory map: ======== ... then GHCi quits. > BTW, the location of your header files is still a puzzle. Oracle's docs: > http://download.oracle.com/docs/cd/B19306_01/appdev.102/b14250/ociabdem.htm… > > state that the header files shoulld be in $ORACLE_HOME/rdbms/public. > But perhaps things are different for the Instant Client. I don't know. I have just installed the RPMs and I expect that others will do so as well.

2 7