I’ve been tinkering with Padrino this week. Padrino is a ruby web framework built on top of the delightful sinatra microframework. I found it attractive because

• it appears to have a simple design, as you’d expect of something built on top of sinatra,
• it rolls in the various elements that you’d expect for a fully fledged web framework (i.e. orm, database, mailer) but makes no commitments to the particular modules used for each *
• it had a default Admin interface (which is a huge win for Django IMO).

There’s quite a bit of supporting material on the web, but my impression after a few hours is that it’s still pretty beta and has a strong leaning towards RHEL linux. So this post details the steps I had to take to be able to work through the basic project guide with Ubuntu Lucid Lynx.

For starters I couldnt use Ubuntu’s default version of rubygems, so I uninstalled it and installed the latest. This allowed me to apparently install the framework but there were still several hidden dependencies to uncover before I could walk through the guide.

I was able to generate a project but found lots of errors pointing to missing gems when I tried to create the in-built admin application. I found I had to install: rake, haml, rack-flash, thin, activerecord, sqlite3-ruby and datamapper gems and for sqlite3-ruby I needed libsqlite3-ruby and libsqlite3-dev via apt-get.

With these installed, when I attempted to create an admin app I wasnt getting errors, but it wasnt working:

me@laptop:~/padrino/first_go$ padrino g admin
=> Located unlocked Gemfile for development
Please specify generator to use (project, app, mailer, controller, model, migration)

Fortunately, someone else had got there first. Installing libssl-dev and libopenssl-ruby through apt-get allowed me to generate an admin app.  It’s cute and I’m able to do everything I’ve tried so far.  I’m still getting “=> Located unlocked Gemfile for development” messages when I interact with padrino which suggests something is up, but I haven’t worked out what it is yet or whether it’s something I need to worry about.

* it remains to see if this approach is successful. It could be that maintaining agnosticism towards the various modules means that the API becomes too hard to use, but it’s certainly interesting to watch them give this a go.

• A script for backing up all your MySQL databases to file
• Updating dates in a demo database refresh restore

A script for backing up your databases to file

In my experience it’s remarkably difficult to restore MySQL databases from the raw database files.  On the other hand, it’s very easy to restore databases from a dump script.  This bash script creates dump scripts for all your non-system mysql databases in a directory that can then be backed up every night.  [sarcasm] It’s probably not appropriate for production servers.

#!/bin/sh
#A bash script that backs up each database into a sepearate file, database_name.sql

for i in `echo "show databases" | mysql -u backup -pbackuppassword`;
do
  if [ ! "$i" = "Database" ] && [ ! "$i" = "mysql" ] && [ ! "$i" = "information_schema" ]
  then
    mysqldump -u backup -pbackuppassword $i > /backup/mysql/$i.sql
  fi
done

Note that the success of this approach depends on you having a backup mysql user set up with the appropriate permissions for all databases.

Nightly restores of a demo database with updated dates

In one of our demos, we back our application with some demo entries and allow interested folk to play with those entries and create new ones.  In order to maintain a sense of order in the demo database, it gets refreshed every night.  Bash’s date manipulations can be useful here.  Having restored the database, i.e.:

# step 1. clear database and add example patients
mysql -u mydbuser -pmydbuserpassword mydb < mydb.sql

We can tweak the dates to make the data seem fresh…

# step 2. update dates
mysql -u mydbuser -pmydbpassword mydb -e "update mytable set last_seen = '`date --date='1 day ago' '+%y-%m-%d %H:%M:%S'`'"

Note how the date program allows you to use natural language modifers like “2 weeks ago” to define your date.

I needed latex this morning. My quick tip for installing it on Ubuntu Karmic Koala is to install Texmaker via the “Ubuntu Software Centre” (which you should be able to find at the bottom of the Applications menu) and then install the

• texlive-fonts-recommended
• texlive-fonts-extra

packages on the command line with aptitude.  This sorted me out very quickly, and the interface for the new texmaker version in karmic seems excellent – attractive and intuitive.

The first is that the linux download of eclipse doesnt work very well in the latest version of Ubuntu (Karmic Koala). A work around is described here with further links to Ubuntu’s issuelist. But actually the work around is only a partial fix because Eclipse sometimes starts a copy of itself and, when it does, it ignores the work around. This affects me as I’m currently trying to develop plugins.

One alternative approach that has worked for me so far is to use the version of eclipse that’s in the Ubuntu repositories. However, this version wasn’t taken from the final release of the eclipse codebase and the groovy plugin install, which patches the JDT (Java Development Toolkit) just wont install in this version.

So to use the groovy plugin it looks as though I’ll have to wait for Eclipse 3.5.2 (which is due on February 29th).

I’m taking baby steps while I explore the space of possible options.  So far I’ve prototyped an expression evaluator that accepts data in the form of simple Java  framework objects and allows boolean expressions to be evaluated over them. In today’s baby step I thought I’d try and manipulate some non-framework Java objects built with an EMF model that I’m currently working on and two different scripting languages, Groovy and Rhino. Groovy is a scripting language that has grown out of the Java community in response to Python and Ruby.  Rhino is a Javascript engine implemented in Java from the Mozilla foundation that is bundled as the default scripting engine in the Java6 JDK.  I got somewhere with Groovy but got completely stumped with Rhino because I just couldn’t load my custom classes.

Groovy

First the success story.  I fired up the groovy shell (groovysh) in the directory that contains my classes and got started straight away:

Groovy Shell (, JVM: 1.6.0_15)
Type 'help' or '\h' for help.
-------------------------------------
groovy:000> dec = new argkit.decision.impl.DecisionImpl()
ERROR java.lang.NoClassDefFoundError: org/eclipse/emf/ecore/EObject
at java_lang_Runnable$run.call (Unknown Source)

Oh, yeah, I needed to add the EMF classes to the classpath. I could do that by restarting the shell with the EMF jars included in the classpath or add them in directly…

groovy:000> this.class.classLoader.rootLoader.addURL(new URL("file:///.../org.eclipse.emf.ecore_2.5.0.v200906151043.jar"))
===> null
groovy:000> this.class.classLoader.rootLoader.addURL(new URL("file:///.../org.eclipse.emf.common_2.5.0.v200906151043.jar"))
===> null
groovy:000> dec = new argkit.decision.impl.DecisionImpl()
===> argkit.decision.impl.DecisionImpl@5878d2 (name: decision0, caption: null)
groovy:000> cand = new argkit.decision.impl.CandidateImpl()
===> argkit.decision.impl.CandidateImpl@b9132a (name: candidate1, caption: null)
groovy:000> dec.candidates.add(cand)
===> true
groovy:000> cand.status
===> UNKNOWN
groovy:000> arg = new argkit.decision.impl.ArgumentImpl()
===> argkit.decision.impl.ArgumentImpl@ce796 (name: argument2, caption: null) (support: +)
groovy:000> cand.arguments.add(arg)
===> true
groovy:000> cand.status
===> RECOMMENDED
groovy::000> arg.support = argkit.decision.Support.AGAINST
===> -
groovy::000> cand.status
===> REJECTED
groovy::000> exit


Having told Groovy about the EMF jars my code depends on, I create a decision object with one candidate and watch the recommended status of the candidate change with the addition of an argument and it’s subsequent manipulation.

Rhino

Then I fire up the equivalent for Rhino (jrunscript), this time remembering to include the EMF jars on the classpath, and start my shell:

js> dec = argkit.decision.impl.DecisionImpl()
script error: sun.org.mozilla.javascript.internal.EcmaError: ReferenceError: "argkit" is not defined. (<STDIN>#1) in <STDIN> at line number 1

Ah.  This was an unexpected issue as argkit is definitely on the classpath.  And there’s pretty much where I got to.  Some digging lead me to this bug report for Java which suggests that the problem is with the non-standard top level domain name (TLDN) in my package and points the finger at Rhino.  Adjusting my model (by setting .org as the TLDN) for this perplexing issue got me a little further…

js> dec = org.argkit.decision.impl.DecisionImpl()
script error: sun.org.mozilla.javascript.internal.EcmaError: TypeError: [JavaPackage org.argkit.decision.impl.DecisionImpl] is not a function, it is sun.org.mozilla.javascript.internal.NativeJavaPackage. (#1) in at line number 1

Ummm, it’s a what?

js> org.argkit.decision.impl.DecisionImpl
[JavaPackage org.argkit.decision.impl.DecisionImpl]

It’s certainly not a Java Package. It seems to me that Rhino was really not cut for this particular use case.

Conclusions

So to recap, I tried to use Groovy and Rhino to manipulate some non-framework objects, as the expressions in my scripting engine embedded app will need to range over custom state objects and Groovy beat Rhino hands down in terms of ease of integration with the custom code. Now, if only I could get the Groovy plugin for Eclipse to install, but that’s another story…

When I save a link to delicious I have two rules that I’m not always able to keep:

1. Always write a description and
2. Never use the word “interesting”.

Came across siftables today for the first time via a link from O’Reilly Radar, which referenced a post from that very nice Phil chap’s coverage at ETech (wish I was there).  This is very nice kit.  The video on the siftables site is really quite exciting.   And it shows footage from a media lab – another awesome project from MIT that I’d forgotten about.  All very life affirming.

Apparently there was a New Scientist article about it a month or so ago so I was surprised I’d missed it.  I suppose it must have been in that issue that got lost in the heavy snow

• Big dog video – impressive, very expensive looking robotic quadruped from DARPA
• Blowing up – a 2002 article from New York Times focussing on Nassim Nicholas Taleb – a hedge fund investor looking for black swans.
• Shell shock – a write up of a researcher at MIT interested in biologically inspired self-assembling materials.