At least that’s what I think it was. It’s been too long. We generally go through the profiler for each major release so in between I sometimes forget some of the details. We focus mainly on optimizing the main bottlenecks instead of spending a lot of effort on unnecessary optimizations (which is why we really value this tool!).

You might want to doublecheck with the JProfiler website, I’m sure they can confirm it for you.

I’m not sure I correctly understand what you’re asking… Would you mind rephrasing it differently, and possibly adding an example?

Does sy have information about what means <class>[] in the primitive type list?

First, thanks! I’m glad you enjoyed the article.

In regards to lazy instantiation, I generally use “eager instantiation” when I want a pre-set default value, or the object is a primitive in Java. Otherwise I try to use lazy instantiation if it doesn’t look too costly in terms of coding effort (which sometimes it is). If it looks like it will be a little more complex to use lazy instantiation then I shortcut and use eager instantiation and just let a code profiler such as JProfiler let me know where I should tweak the code. Why go through all the trouble when 90+% of the time it won’t make any niticeable difference

As for test database, we created some database generating scripts. For example, for tenants we randomly pick a first and last name from a list of the top 1000 firstnames and top 1000 lastnames accross North America. As for addresses, we just took a list of all the streets from one city and randomly append a street number. Mostly what we do is try to grab realistic data and include some randomness into it so that we can have much better testcases!

Don’t get me wrong, creating all the test database scripts took some time initially, but what’s great about it now is that we can create new fairly random databases that contain real world looking data. And sometimes this brings forth issues that we might not have thought about before!

I am curious how you approach creating test databases, especially larger ones. Did you create a few properties and then copy them many times or write random stuff into the databases, or did you try to create more realistic datasets? Perhaps you took a combination of approaches?