Recently I've been doing some work with a client on WAS 6.1. Since we have a number of Spring users on WAS and I need to test the application, I decided it was time to get a copy of WAS running on one of my work laptops. I say 'one of' because I'm currently working on both my Mac (with OSX) and my ThinkPad (with Ubuntu) - more recently I've just been using the ThinkPad because I can have Oracle XE and WAS running without the need for a VM tool like Parallels. I still prefer the Mac, but to be honest there isn't much difference day-to-day - I just miss some of the more useful Mac tools like Spotlight, Quicksilver, TextMate and NewsFire.

Anyway, back to the main topic - installing WAS 6.1 on Ubuntu. I'm using Ubuntu Edgy and my first attempts at an install failed completely and I just couldn't figure out why. Thankfully a quick Google turned up this article. I was completely unaware that /bin/sh was linked to dash instead of bash - what on earth possessed them. I didn't really like the suggested solution of running the installer, letting it fail and then changing all the scripts in the installed directory. Instead, I just relinked /bin/sh with a quick sudo unlink /bin/sh followed by sudo ln -s /bin/bash /bin/sh. After that, the installer ran like a dream and I was up and running with a WAS install in about 15 minutes.

Even on my ThinkPad with Oracle XE running at the same time, WAS runs pretty quickly. One of the nicest things about WAS is that the tools provided (admin console, command-line tools) are really robust. The Admin Console is noticeable for its performance - many other servers have consoles that are painfully slow.

In my last entry I presented a technique for creating strategy classes that take full advantage of any generic metadata that is present in your application. At the end of that entry I showed this code snippet:

EntitlementCalculator calculator = new DividendEntitlementCalculator();
calculator.calculateEntitlement(new MergerCorporateActionEvent());

You'll remember that DividendEntitlementCalculator was defined as:

public class DividendEntitlementCalculator implements EntitlementCalculator<DividendCorporateActionEvent> {

    public void calculateEntitlement(DividendCorporateActionEvent event) {

    }
}

As such it is not correct to pass an instance of MergerCorporateActionEvent into the calculateEntitlement method of the DividendEntitlementCalculator class. However, as I mentioned in my last entry that code will compile. Why? Well, EntitlementCalculator.calculateEntitlement() is defined to accept any type that extends CorporateActionEvent so it should compile. So in this scenario what happens at runtime and how does Java enforce type safety? Well, as you might imagine running this code gives you a ClassCastException saying that you cannot cast a MergerCorporateActionEvent to DividendCoporateActionEvent. In this way, Java can enforce type safety for you application - there is no way that the MergerCorporateActionEvent can creep into a method where DividendCorporateActionEvent is expected.

The real question here is: 'Where does that ClassCastException come from?' The answer is pretty simple - the Java compiler adds the code to create and throw it as appropriate by introducing a bridge method. Bridge methods are synthetic methods that the compiler will generate and add to your classes to ensure type safety in the face of generic types.

In the case shown above EntitlementCalculator.calculateEntitlement can be called with any object that is type compatible with CorporateActionEvent. However, DividendEntitlementCalculator accepts only objects that are type compatible with DividendCorporateActionEvent, but, since you can call the DividendEntitlementCalculator via the EntitlementCalculator interface it too must accept CorporateActionEvent. So what does this translate to in the compiled class file? We have the user supplied method:

public void calculateEntitlement(DividendCorporateActionEvent event) {
    System.out.println(event);
}

Which translates to this bytecode:

public void calculateEntitlement(bigbank.DividendCorporateActionEvent);
  Code:
   Stack=2, Locals=2, Args_size=2
   0:   getstatic       #2; //Field java…

I was on site with some clients the other day, and they asked me about unit testing and mock objects. I decided to write up some of the discussion we had as a tutorial on creating dependencies (collaborators) for unit testing. We discuss two options, stubbing and mock objects and give some simple examples that illustrate the usage, and the advantages and disadvantages of both approaches.

It is common in unit tests to mock or stub collaborators of the class under test so that the test is independent of the implementation of the collaborators. It is also a useful thing to be able to do to…

Last month Rod Johnson presented at three Australian Spring User Group meetings a session entitled, "What's New and Cool in Spring 2.0". Rod mentioned during those meetings that I'd make his presentation available, so here it is.

There are some other recent presentations that people have also been emailing me about. In no particular order, here is the latest:

• If you're looking for Rod's other session, "Simplifying Enterprise Applications with AOP", you can download it directly from the Sydney Java User Group web site.
• The Real Object Oriented (ROO) presentation I gave at two Enterprise Java Australia meetings will not be published. These slides have been superseded by the more recent and detailed session I gave at The Spring Experience.
• My "Fast Hop into Real Object Oriented (ROO) Applications" session from The Spring Experience will be published on this blog shortly. I'm currently working on a video to go along with it.
• Coverage of Acegi Security 1.0.3's new ACL capability in my "Beyond Low-Hanging Fruit: Domain Access Control List" session will also be published here shortly. In the meantime, Acegi Security SVN now contains the sample application (SEC-415).

For those of you who attended the presentations, I hope you enjoyed them.

A few weeks ago I blogged about the origins of the name Spring. We also get many questions about the origins of the name Interface21.

For anyone who's read my books or considered the design of Spring, the interface part is hardly a surprise. It plays on both the OO concept of an interface (for which I've always had a deep love) and the notion of the interface to a system. For example, putting a web interface onto an existing green screen system--something I was actually doing when I first thought of the company name.

So far, so good. The real problem is with the numbers.

As with the name Spring, some of the theories are more interesting than the real explanation. So let's start with the theories I've heard regarding the 21 …

I was on site at a customer last week and a question came from the crowd, "Why isn't getConfigLocations() abstract anymore?" After working in front of customers for a while, it becomes rare that you're speechless, and yet I was. To be honest, my first thought was that there was no way the customer could be right. But lo and behold, in revision 1.3 of AbstractSingleSpringContextTests it clearly states that getConfigLocations() is no longer abstract. I hadn't created any new integration tests against 2.0.1, so I hadn't even seen the change.

Surprised by this, I put in an email to Juergen to…

These shots of our venue were taken yesterday (proximity to places like this is one of the perks of Interface21 having an office in Florida).

We are incorporating several of these shots into the main conference banners to be draped from the towering ceilings of the Diplomat. Everything is set for a great show. See you at The Spring Experience next week!

In the run up to The Spring Experience I've been busy but I've noticed that Rod's been really active on the blogging front. So in some spare time in airports and on planes today, I've decided to do a little blogging.

One of the biggest balancing acts that we in the Spring community have is to make sure that we stay backwards compatible while still innovating. Part of that innovation is taking advantage of new features and constructs in later versions of Java such as Java 5. Since the 1.2.x branch, we've seen some of this with things like the @Transactional annotation and our JMX auto-detection based on the @ManagedResource annotation. In the end these are great features and have greatly simplified development (at least mine anyway), but they really amount to moving metadata into the code. What we hadn't seen was…

Thanks to our philosophy of pluggability and a lot of hard work in the implementation, the Spring IoC container (like most of the rest of Spring) is extremely flexible.

One point that is often missed is that Spring configuration need not be in XML, although the XML format is by far the most commonly used. Spring has its own internal metadata format in the form of the BeanDefinition interface and subinterfaces. The BeanFactory and ApplicationContext implementations that represent IoC container instances are powered by this Java metadata, and are quite separate from metadata parsing, which is…