Leiningen and VimClojure

Leiningen seems cool. It will take care of installing any jars you need for your clojure application, and setting up the classpath, so you don't have to worry about all the Java stuff and can get down to clojure hacking.

I spent some time yesterday getting Leiningen and VimClojure to work together. Here are the notes in case someone else has the same problem:

> mkdir code/clojure
> cd code/clojure

LEININGEN

I downloaded the install script from http://github.com/technomancy/leiningen and put it
in ~/code/clojure/lein.

> chmod a+x lein
> ./lein self-install
> ./lein new helloworld
> cd helloworld

NAILGUN server

Next step was to add the nailgun server needed by vimclojure. To do this I added these lines
to project.clj (which was generated by "lein new helloworld" before)
[org.clojars.ato/nailgun "0.7.1"]
[org.clojars.gilbertl/vimclojure "2.1.2"]

Now update the dependencies to generate a lib directory with all jars we need:

> ../lein deps

VIMCLOJURE

I already had the vimclojure vim plugin installed, but needed to add the nailgun client for VimClojure. For this I downloaded VimClojure (http://www.vim.org/scripts/script.php?script_id=2501), unpacked it somewhere and ran "make". This generated a file called "ng" which I copied to ~/code/clojure. I then added this line to .vimrc:
let vimclojure#NailgunClient = "/home/johan/code/clojure/ng"

I then put these alias definitions in .bashrc:
alias lein='/home/johan/code/clojure/lein'
alias ng='java -server -cp src:classes:lib/* com.martiansoftware.nailgun.NGServer 127.0.0.1'

TESTING AND RUNNING

> cd ~/code/clojure
> ng &
> vim src/helloworld.core.clj
You should now be able to type sr to open a REPL buffer (\sr if using the default settings)

VimClojure on Ubuntu

I have just installed vimclojure-2.0.0 on Ubuntu. Here is my log, in case someone wants to do something similar

Download vimclojure: http://kotka.de/projects/clojure/vimclojure.html

 cd
 mkdir clojure
 cd clojure/
 svn co http://clojure.googlecode.com/svn/trunk clojure
 cd clojure/
 ant
 cd ..
 mkdir jars
 cp clojure/clojure.jar jars
 git clone git://github.com/kevinoneill/clojure-contrib.git
 cd clojure-contrib/
 ant -Dclojure.jar=../clojure/clojure.jar
 cp *jar ../jars
 tar xvfz vimclojure-2.0.0.tar.gz 
 cd vimclojure-2.0.0/
 vi local.properties
clojure.jar=/home/johan/clojure/jars/clojure.jar
clojure-contrib.jar=/home/johan/clojure/jars/clojure-contrib.jar
nailgun-client=ng
vimdir=/home/johan/.vim
 ant
 mkdir ~/.vim/autoload
 mkdir ~/.vim/doc
 mkdir ~/.vim/indent
 mkdir ~/.vim/syntax
 mkdir ~/.vim/ftdetect
 mkdir ~/.vim/ftplugin
 mkdir ~/.vim/ftplugin/clojure
 cp autoload/vimclojure.vim /home/johan/.vim/autoload/
 cp doc/clojure.txt /home/johan/.vim/doc/
 cp indent/clojure.vim /home/johan/.vim/indent/
 cp syntax/clojure.vim /home/johan/.vim/syntax/
 cp ftdetect/clojure.vim /home/johan/.vim/ftdetect/
 cp ftplugin/clojure.vim /home/johan/.vim/ftplugin/
 cp ftplugin/clojure/completions-clojure.* /home/johan/.vim/ftplugin/clojure
 vi ~/.vimrc
filetype plugin indent on
syntax on
let clj_highlight_builtins = 1
let clj_highlight_contrib = 1
let clj_paren_rainbow = 1
let clj_want_gorilla = 1
let vimclojure#NailgunClient = "/home/johan/clojure/vimclojure-2.0.0/ng"
autocmd BufRead,BufNewFile *.clj nmap  ClojureEvalFile
autocmd BufRead,BufNewFile *.clj nmap  ClojureEvalLine
autocmd BufRead,BufNewFile *.clj nmap  ClojureEvalToplevel
autocmd BufRead,BufNewFile *.clj nmap  ClojureStartRepl
autocmd BufRead,BufNewFile *.clj nmap  ClojureDocLookupWord
autocmd BufRead,BufNewFile *.clj nmap  ClojureDocLookupInteractive
 vi ~/.bashrc
# Clojure
export CLOJURE_EXT=/home/johan/clojure/jars
export LD_LIBRARY_PATH=/home/johan/clojure/bin-linux
alias clojure='/home/johan/clojure/clojure-contrib/launchers/bash/clj-env-dir'
if ! (ps -ef | grep com.martiansoftware.nailgun.NGServer > /dev/null)
then
    sh /home/johan/clojure/vimclojure-2.0.0/bin/ng-server &
fi

And for the final step, start vim and type ":helptags ~/.vim/doc" to make the documentation available with ":help clojure"

You can also add this Nailgun startup script in /etc/init.d/nailgun

#! /bin/sh
# Install: ln -s /etc/init.d/nailgun /etc/rc3.d/S99nailgun

do_start () {
    export CLOJURE_EXT=/home/johan/clojure/jars
    export LD_LIBRARY_PATH=/home/johan/clojure/bin-linux
    sh /home/johan/clojure/vimclojure-2.0.0/bin/ng-server &
}

do_stop () {
    `ps -ef | grep java | grep NGServer | awk '{print "kill -9 "$2}'`
}

case "$1" in
  start)
    do_start
 ;;
  stop)
    do_stop
    exit 0
 ;;
  restart)
    do_stop
    do_start
    exit 0
 ;;
  *)
 echo "Usage: $0 start|stop|restart" >&2
 exit 3
 ;;
esac

Xith3D with Clojure

I've been working a bit with 3D graphics with Clojure, and I have now tested various libraries. jME was a bit awkward to set up in Clojure, and Java3D is dying, so my current efforts are concentrated on Xith3D. I had some problems installing Xith3D in Ubuntu 8.10, so here is how to get started:

Download

  mkdir xith
  cd xith
  svn co https://xith3d.svn.sourceforge.net/svnroot/xith3d/trunk xith3d
  svn co https://xith-tk.svn.sourceforge.net/svnroot/xith-tk xith-tk

Compiling problems and how to solve them

xith3d only compiles with java-6-sun, not openjdk-6, so
remove all other java implementations (see /usr/lib/jvm) first.

  cd xith
  ant

Another problem is that ant is too old, and svnrevget fails. To fix
this, go to xith and do "svn log". Note the latest revision number,
and go to xith-tk. Edit build.xml, and find the svnrevget line.
Comment it out and hard-code the version number:

   [!-- svnrevget property="version.build" path="${project.xith3d}"/--]
   [property name="version.build" value="1668"]

Now you can do "ant" in the xith-tk folder.

Testing

Here is a basic test file (put it in the xith folder, call it test.clj)

(import '(org.xith3d.loop RenderLoop)
       '(org.xith3d.render Canvas3D Canvas3DFactory)
       '(org.xith3d.base Xith3DEnvironment))

(let [env (Xith3DEnvironment.)
     canvas (.createWindowed Canvas3DFactory 800 600 "My empty scene")
     rl (RenderLoop.)]
 (.addCanvas env canvas)
 (.setMaxFPS rl 120.0)
 (.setXith3DEnvironment rl env)
 (.begin rl))

And a script to start it with

#!/bin/sh
export LD_LIBRARY_PATH=xith3d/third-party/gluegen/linux-i586:$LD_LIBRARY_PATH
export LD_LIBRARY_PATH=xith3d/third-party/jogl/linux-i586:$LD_LIBRARY_PATH
java -cp ~/.clojure/clojure.jar:xith3d/dist/xith3d.jar:xith3d/third-party/jagatoo.jar:xith3d/third-party/math/openmali.jar:xith3d/third-party/jogl/jogl.jar:xith3d/third-party/gluegen/gluegen-rt.jar clojure.lang.Script test.clj

and an empty window appears on the screen.

XmlRpc with Clojure

I know, I know... I really should update this blog more often. Brief recap: last year I've made a fairly big Common Lisp project (funmv), and played quite a bit with both Haskell and Clojure. Today I want to share some experiences using XmlRpc with Clojure.

Clojure has the advantage of running on JVM, which means that it can use Java libraries. The problem is that, being a Lisp with a functional flavour, it sometimes doesn't mix well with the object-oriented programming style mandatory in Java development. I needed to implement a Xml-Rpc server in clojure, but my first attempts with apache xmlrpc and others ended in failure since their addHandler methods required an object, and would match the XmlRpc message methods with the name of the member functions. It is possible to instantiate Java-objects with the proxy special form in Clojure, but this only allows you to implement or subclass already declared methods, and creating an object completely from scratch seemed too complicated.

Finally I found an XmlRpc library from redstone (download redstone-simple-xmlrpc-1.0.zip), which allows you to define your own dispatch routine. This allows me to inspect the method name and call normal Clojure functions, rather than relying on Java reflection magic. The server code is short and sweet:

(def server (new redstone.xmlrpc.simple.Server 8080))
(def h (proxy [redstone.xmlrpc.XmlRpcInvocationHandler] []
       (invoke [method-name arguments]
          (cond
            (= method-name "add") (+ (nth arguments 0) (nth arguments 1))
            true (throw (new Exception "No such method"))))))
(doto (.getXmlRpcServer server) (.addInvocationHandler "test" h))
(.start server)

Here is a Java test client

import redstone.xmlrpc.XmlRpcClient;
public class SampleClient
{
    public static void main( String[] args ) throws Exception
    {
        XmlRpcClient client = new XmlRpcClient( "http://localhost:8080", false );
        Object reply = client.invoke( "test.add", new Object[] {1, 2} );
        System.out.println (reply); // should be 3
    }
}

You will need to add the jar files to the classpath:

javac -cp xmlrpc-1.1.1.jar SampleClient.java
jar cf SampleClient.jar *class
java -cp SampleClient.jar:xmlrpc-1.1.1.jar SampleClient

Is Qi the answer?

I have returned to functional programming after more than 10 years of C, C++, C# and Java, and it feels good. The old favorite is of course Lisp, but much has happened in the field since I was active, and I thought I should check out the statically typed languages, such as Haskell. Well, it turned out to be a fairly unpleasant experience. Not because it was hard to make programs - the program structure in Scheme is fairly similar to Haskell - but because the typing system keeps getting in the way of my iterative way of programming. I want to write some code, play with it, extend it, play with it some more, and let it grow step by step. In Haskell, like in statically typed imperative languages, I seem to spend most of the time fixing type errors instread. Of course they have to be taken care off, but not when I'm prototyping!

But perhaps I found a good compromise: Lisp syntax, Haskell-like functional programming, optional type checking, lots of libraries. Qi seems to mix the best of two worlds, and so far it has been a joy to use. First prototype your application, then turn on type checking, fix the errors and you should have a pretty robust program without the Haskell straitjacket. That is my hope, time will tell if it works out in practice.