I wanted to try out a persistence API for Cassandra. There’s a JPA implementation for Cassandra: Kundera, as well as JDO implementation, on top (or using) datanucleus: datanucleus-cassandra.

Just to clarify why I would want such a thing: nosql solutions such as Cassandra are, in essence distributed key-value stores. I know, from following a few classes on databases and distributed databases when attending the university that’s a whole lot more complicated to implement then it may seem.
Nevertheless, to do anything useful with that you need a layer on top of the key-value store. Basically all SQL databases like Oracle or MySQL have SQL layer around there key-value store. But the nice thing about stuff like Cassandra you can put something more convenient on top.

I wanted to try out both, starting with the JDO version, as that seemed more like a natural fit, and I once started ORM with JDO 1.0 many years ago.
I’d need a git client on Windows, and apparently there’s already a Tortoise Git client, based on the truly great TortoiseSVN, on of the tools I really miss on my Mac. The client version of git first had to be installed, but that was quite easy as well, just download and install msysgit.
The tortoisegit client looks very well, and seems easy to use. At least I got the code of datanucleus-cassandra in a breeze. Well now it’s kind of late, I’ll try out later.

Other stuff I want to try out: creating a Rest service using Restlet on Google AppEngine. Just to find out how that works. But later.

Finally I saw a demo on VMForce. Looks very promising, as I like Spring for development and that’s what you’d use. Nothing is released however, now there’s just articles and a demo.

Of course Oracle has something like SQLPlus, but that would mean SQLPlus has to be installed on the machine from where the package is installed. Another problem with SQLPlus is that it’s not so userfriendly.

Using JDBC and some programming I can easily create something that would check what current version of my application is installed, and then run the needed SQL scripts to upgrade the database. However, even easy things take some time and I’d think there are already some existing solutions.

I was about to ask a question on stackoverflow, but before I finished typing my question stackoverflow already came up with a similar question: Update a backend database on software update with Java. I’ll have to try out on of the suggested solutions, Liquibase, dbmigrate and maybe Autopatch.

Also, this allows for fine-grained security at database level: Virtual Private Database.

(c) Oracle

As described in the above image and referenced article, the user identified by userid 106 (let’s say that’s user john) will access the database using a private database connection. The user can only access rows that have that userid as primary or foreign key.  John can’t see the passwords, orders, credit card data or anything of other users even if he would somehow hack the webapplication. How to set up this finegrained securiy is beyond this article, but I hope the above example explains what the purpose VPD is.

If you develop your JEE software using Oracle software  proxy authentication requires only a bit of configuration, for example, see the following article how to setup proxy authentication using JHeadstart.

However, what if you’re JEE applicication consists of non-oracle software? Can you still use proxy authentication when you use software like JBoss, Hibernate, IBatis, MyFaces, Wicket etc? Yes you can! I’ll explain below how to set up proxy authentication using JBoss in such a way you don’t have to modify any of the code that uses JDBC, directly or indirectly.

• First a way is needed to set a username for each (JDBC) database connection retrieved. It-eye weblog explains how to open a connection to a database using java, and then switch to another username.
• Secondly, in your application users should authenticate them self using the default mechanism of J2EE 1.4 (and higher) application, using JAAS. Usually creating a security policy inside your web.xml or inside your ear is enough. Here’s the information how to do this under JBoss: Secure a webapplication.
  Since we’re using a database, the best option would be to use database based authentication, meaning user information comes out of a database table. There’s a lot of documentation on the web how to do add security, so I won’t repeat that here.
• Database connection in JBoss are retrieved using connection pooling, as is custom in any JEE server.  You can create a custom connection pool, that changes the switches to the user name that is currently logged in at the webapplication. That way, every action on the database is done under a database user that is currently logged in.Do to be able to do that, first, you’ll need a custom connection factory that extends the default connection factory. Our connection factory will return a customized datasource that modifies code.Here’s a code listing:

  package nl.gerbrandict.dbconnaudit;
  
  import java.sql.SQLException;
  import javax.resource.ResourceException;
  import javax.resource.spi.ConnectionManager;
  
  import org.apache.log4j.Logger;
  import org.apache.commons.lang.StringUtils;
  import org.jboss.resource.adapter.jdbc.local.LocalManagedConnectionFactory;
  
  /**
   * An extended connection factory, that uses the Oracle feature to change the username of an existing connection
   *
  
   * When a user authenticated on the application server, the database connection will switch to that username.
   * This allows for better auditing and potentially for improved security.
   *
   * Properties (get'ers and set'ters) can be set via the configuration section of the -ds file
   * @author gvdieijen
   */
  public class DBConnAuditConnectionFactory extends LocalManagedConnectionFactory {
  
      private String defaultProxyUser;
  
      public DBConnAuditConnectionFactory() throws SQLException {
          super();
  
      }
  
      @Override
      public Object createConnectionFactory(ConnectionManager cm) throws ResourceException {
          return new OracleWrapperDataSource(this, cm);
      }
  
      public void setEnableProxySession(Boolean enableProxySession) {
          this.enableDbConnAudit = enableProxySession;
      }
  
      public void setDefaultProxyUser(final String defaultProxyUser) {
          if (StringUtils.isEmpty(defaultProxyUser)) {
              this.defaultProxyUser=null;
          } else {
               this.defaultProxyUser = defaultProxyUser;
          }
      }
  
      /**
       * Default db user to open proxy session for, when no authenticated user is active
       * @return
       */
      public String getDefaultProxyUser() {
      	return this.defaultProxyUser;
      }
  }

  As you can see, the file returns a OracleWrapperDatasource. That’s custom class, that extends the default WrapperDatasource of JBoss, and changes the user of jdbc connection just before the connection is handed of to the application. To speak in terms of the fine book

• Now, How can you use that new class? They have to be packed into a rar file. Functionally, that’s a Resource Adapter, a module that allows a J2EE application to use resources. Technically (and practically), it’s just a jar-archive similar to a war, with a different extension. Maven can create these files automatically for you, if you set the packaging type to rar instead of jar (which is the default).
• When you download JBoss, you’ll get a sample connection pool for a in-memory database: default-ds.xml, located in the server/default/deploy directory of jboss. To use the custom classses, copy the file into (for example) myoracleproxy-ds.xml and create a minor modification so a custom connection factory is used – update the managedconnectionfactory property, that a custom connectionfactory is used, that returns the proxied connections:

  <managedconnectionfactory-class>nl.gerbrandict.dbconnaudit.DBConnAuditConnectionFactory</managedconnectionfactory-class>

All in all, after some tweaking, all queries and updates to your Oracle database are done under the J2EE username. This means when user Joe logs in, all his database access will be logged under user Joe as well. This can improve auditing as well as security.

In the coming postings I want to investigate that claim using my knowledge of experience in Java/J2EE and my complete lack of experience in Scala. Can you say to you’re developers, or yourself, from this day on, we’ll use Scala instead of Java to do our development?

For now, I found this nice presentation I want to share: Introduction To Scala For Java Developers by Miles Sabin.

• Hitting ‘Run’ and waiting a few seconds,
• to hitting ‘Run and deploy’ in my IDE and waiting about 20-30 seconds,
• to executing a ant script, restarting the application server and waiting 5 minutes
• to compiling a class file, uploading the class and accompanied libraries using a cumbersome webinterface and then executing a PLSQL script to run that class, wasting in total time amount 15 minutes (yes, that was a a very big and professional company)
  

In my opinion, the programmer is productive when he either codes, or when he tests his codes. Any time spent on waiting or performing some troublesome actions like accessing a web interface is pure waste. In short the above time ranging from a few seconds to 15 minutes is dead time.

Of course all that time wasted not only frustrates developers, it’ll cost quite some time and thus money. Now how can you limit that dead time, that is wasted by developers? One way is your have your developers write perfect code at once. Of course that’s not very realistic. Recently I found a very nice tool to limit that dead time, when you develop Java software: JRebel. What does that mean for you as java-developer:
JRebel runs as a software agent within your virtual machine, that in turn runs your application server. What JRebel does, is taking over the class-loader.
You can specify the output directories of your IDE (target/classes, /bin, etc.). Now whenever you make a change in your code, IDE’s like Eclipse can compile that code into a new class file in the output directory. JRebel will automatically detect the changed class file. Thus no waste! Since JRebel plugs into the VM, it’ll work with virtually any application server.

we’re also finishing up a new version of the Dropbox desktop software that features numerous performance improvements and our new “LAN sync” feature. LAN sync knows when Dropboxes are on the same network and will automatically exchange files directly between computers instead of downloading them from our servers – this makes sharing large files in an office environment much faster than was previously possible.

So all in all Dropbox now has that feature. Either Zumodrive follows fast, or they already do that. Seems kind of obvious feature, LAN-sync, so maybe they already do that.

class Address {
    String firstName;
}

List 
 myAddresses ;
...
List 
 onlyJohns = (List 
) ListUtils.predicatedList(myAddresses,new Predicate() {
   public boolean evaluate(Object o) {
       Address a=(Address)o;
       return a.firstName.equals("John");
   }
});
 

Quite a lot of casting as you can see in the above code. Less code would be needed if generics support would be added to the collection framework. Added generics support to the collections library should be too hard, and would improve my code.

I’d think other people would have that idea, and I quickly found the following posting on devx. Turns out there’s a sourceforge project that has modified the collections library to have generics support: Commons Collections with generics.
The project is even added to the central maven repository, so to use the new collection library adding the following dependency is enough:

net.sourceforge.collections
collections-generic
4.01

The new code now becomes (excluding the Address class, which is the same):

List 
 onlyJohns = ListUtils.predicatedList(myAddresses,new Predicate
() {
        @Override
   public boolean evaluate(Address a) {
       return a.firstName.equals("John");
   }
});

The code looks quite better! A lot less code would be needed if closures would be added to Java too, but that’ll have to wait until JDK 1.8 or something is released.

I started noticing other publications and advertisements on similar technologies. Ordina, a Dutch IT company, started an open source project Mod4j. At first glance, Ordina is using a traditional approach to MDD: using a model to generate source code, which can then be customized.  This approach had previously always failed, because in the end the customization was at the very least as costly and cumbersome as developing the software from scratch, but with less flexibility. Better approach seems to be using a DSL specifically. This approach is used by the two companies of the two former UU-students. Unlike Ordina, both companies don’t develop their software as Open Source so there’s little for me to download and try out.