public class DemoApplication {
 
    public static void main(String[] args) throws Exception {
        DemoApplication calc = new DemoApplication();
        System.out.println(calc.divide(calc.add(17, 19), 2));
    }
 
    public int add(int i, int j) {
        return i + j;
    }
 
    public int divide(int i, int j) {
        return i / j;
    }
}

In earlier times, I stepped one step further to get into the method divide. Then I got the result of add(17,19) in my parameter.

But Eclipse is offering a better solution. In the Debugging Perspective, there is a view called Expressions. Just add a new expression add(17,19) and the Debugger is printing out the result.

The result is directly printed out:

A simple, but helpful feature in Eclipse!

I started about two months with working a little bit with OSGi. I want to write here these parts, which were unclear for me at the beginning. It should be an introduction for everyone, who has never heard about OSGi.

Okay, let’s start.

The setup of OSGi is something like EJB.

1. EJB has a specification from Sun.
2. OSGi has specification from the OSGi Alliance.
3. EJB need’s an application server as a platform (Weblogic, Glassfish, Websphere, JBoss, and so on) to run an “EJB” application.
4. OSGi need’s also a platform to run “OSGi” application. Like an application server, which is implementing the EJB contract, an OSGi platform implements the OSGi framework specification. Currently, there some OSGi implementation out there. The most known are Equinox, Apache Felix or Knopflerfish.

Okay, now, we know, how OSGi is built. But what does it? In one sentence, it is something like a JVM with some extra features. OSGi extends the JVM and a Java program is running on that OSGi Platform instead of running directly on the JVM. OSGi offers some advantages, which I will mention shortly.

 
(Source: http://de.wikipedia.org/wiki/OSGi)

 
The difference between a usual Java program (a JAR file) and an OSGi program (let us call that a bundle) are some information in the META-INF\MANIFEST.MF file. These information looks like:

Manifest-Version: 1.0
Bundle-ManifestVersion: 2
Bundle-Name: Example OSGi bundle
Bundle-SymbolicName: ch.stefanjaeger.osgi.example
Bundle-Version: 1.0.0

That’s it. An OSGi program (called a bundle) is nothing more than a JAR file with the MANIFEST.MF. Now, you can run this bundle in an OSGi environment like Equinox or Apache Felix.

Okay, know, we come to the advantages of OSGi. The biggest advantage is the ability to run a bundle in different versions at the same time. For example, bundle A has a dependency to bundle B in version 1.0.0 and bundle C has a dependency to bundle B in version 1.1.0.

With the usual JVM, you get a problem when running A and C in the same VM. The JVM will take the first “B” in the classpath it found. This means, it is possible, that for C gets the files from B in version 1.0.0 (instead of 1.1.0), which could result in “MethodNotFoundException” and such stuff.

With OSGi, this problem doesn’t exists anymore. In the MANIFST.MF you can define all dependencies explicitly with the version. Following example means, that our bundle has a dependency to the Apache Commons Lang Package in version between 2.0.0 and 2.1.0.

Require-Bundle: org.apache.commons.lang;bundle-version="[2.0.0,2.1.0]"

The OSGi platform checks at the beginning, that all required bundles are available and solves the classpath problem for us.

Okay, we’re almost finished. There is one important point I want to mention. Packages are not “open” as usual in JAR files. If we want to make some classes accessible for other bundles, we do have to export these packages:

Export-Package: ch.stefanjaeger.communiction.interfaces

Of course, OSGi can do more than that. But for this introduction, I think, it’s enough

• Study Guide
  http://java.boot.by/scbcd5-guide/
• Free Training Test:
  http://www.sun.com/training/certification/assessment/

I also wrote some “Factsheets”, which I used to sum up all topics. Besides Factsheet 0, they can also be used as an assistance on the daily EJB 3.0 work.

• Factsheet 0 – some stuff, which is only for the SCBCD relevant
• Factsheet 1 – Stateful and Stateless Session Beans
• Factsheet 2 – MDBs and JMS
• Factsheet 3 – Transactions and Security
• Factsheet 4 – JPA, part 1
• Factsheet 5 – JPA, part 2

If you find any mistakes in the Factsheet, just let me know.

I am currently preparing for the SCBCD 5.0 and in my opinion, it’s better than the book “Enterprise JavaBans 3.0” from O’Reilly. “Mastering EJB 3.0” covers many topics about Java EE much deeper. For example, it mentions poison messages or clustering as  has a very useful overview of all possible annotations in the appendix.

Only the part JPA is not covered as deep as in the O’Reilly book. EmbededId is only mentioned, but not described in detail.

There are interesting clean ups like “Remove unnecessary casts” or to add missing Annotations like “@Override”.

After creating your profile, simply right click to the project and choose Source – Clean up.

Nice feature. If you want to know more about this feature, you will find here more information’s.

But let me start with the simple problem. We do have two simple POJOs, Customer and Person, which do know each other. If you are using for example Hibernate, bidirectional linking is not a rarity. With the JiBX binding XML we used in our last examples, the bidirectional linking is not set. To demonstrate this, I added these two lines of code in the main application:

System.out.println(customer.toString()); 
System.out.println(customer.getPerson().getCustomer().toString());

The second line will result in a NullPointerException, because JiBX is not setting this bidirectional linking. To get that working without affecting the two classes Customer and Person (!), we have to create our own Marshaller and Unmarshaller.

First of all, we have to prepare our mapping file. The best way for a bidirectional mapping is to define an own mapping section for the class Person (which was previously part of the Customer mapping part).

The reason for this step is simple, but important. Our customized Unmarshaller should let JiBX unmarshal the Person into an object and our Unmarshaller should only extend the object with the bidirectional mapping to the Customer. For this reason, JiBX needs an own mapping section for this class.

After preparing the JiBX mapping, our next step is to create a Marshaller and Unmarshaller class. For the bidirectional mapping, we will just need an Unmarshaller. But because we do have a JiBX mapping file for both directions, we do have to define a Marshaller and a Unmarshaller. 
We create for this example a new class with the name BidirectionalMapper, which implements the interfaces IMarshaller, IUnmarshaller and IAliasable. These Interfaces are delivered with the JiBX libraries. In the implementation, there are two remarkable parts in the method unmarshal. The first one just forwards the real unmarshalling action to the JiBX context and the second one creates the bidirectional linking.

Object unmarshalledObject = ctx.unmarshalElement(); 
Person person = (Person) unmarshalledObject; 
... 
Object parent = ctx.getStackObject(0); 
person.setCustomer((Customer) parent);

In the marshal method, we also let JiBX marshal our person. Because the bidirectional linking has no affect to this method, this method is just simple.

ctx.marshalCollection(listOfElements);

After we created our Mapper, we need to define the marshallers in the JiBX mapping file. For this reason, we expand the element structure with the marshaller and unmarshaller class path.

If we run now our application again, there is no NullPointerException anymore and we can use our previously used domain model without changing anything at these classes.

If you are interested, you can download the whole eclipse project with all the sources from here.

In a software project, Software Engineers need to ensure, that the source code meets the architectural rules. One of a usual architecture rule are the dependencies between the layer and packages. There are a lot of tools, which can be used to ensure, that no class breaks this rules.

If you know, that there are some specific classes, which do not meet these rules, a dependency code analysis is needed. Therefore, I found a free tool called CDA (http://www.dependency-analyzer.org/), which is very helpful in such a situation. It can also be used for repeated check to ensure architecture compatibility, but in this blog entry, I will focus on code analysis.

First of all, just install this tool as described on the website (http://www.dependency-analyzer.org/#Installation). CDA doesn’t need any installation (just unzip the compressed file) and can be run with Java 1.5.

After finishing the installation procedure, you need to create for every project a Workset. A Workset can contain different code sources, which are analysed together. First, define a Name. After that, go to the Classpath register. The easiest way to use CDA is to analyse a JAR file. Add all JAR files, which you want to analyse.

If you are finished, save the Workset. CDA will load now all classes from the JAR files into a tree. With CTRL + F you can search for a specific class or you can navigate with the package name to a specific class. With right click, you can analyse all dependencies (on which classes/packages does the class depends on) or you can analyse all dependants (which classes/packages depends on the currently selected class).

If you are analysing the dependencies, you can show the classes or packages, on which the specified class depends on. Furthermore, you will see all third-party libraries, on which the class depends on. If you do not want to see some specific packages, you can apply a filter (button “edit filter”), on which you can define excluded packages.

The generated graph looks like following. It is almost the same as the textual dependency view, with the exception, that no third-party libraries are shown.

This tool is user friendly. If you are using this tool, you will very fast get to work with it. I use this tool to find dependants on a class and to draw simple UML diagrams of a class (very useful if you want to draw a UML diagram from just one class with all it dependencies).

For example:

Collections.<String> emtpyList();