Hello Seattle: GTAC 2008

Image via Wikipedia Here's what just showed up in my mailbox:

Congratulations! We are thrilled to be able to offer you a place to attend this year's Google Test Automation Conference in Seattle on October 23rd & 24th, 2008.

Thanks Google!!!

Why Developers Balk at Unit Testing

Unit testing is a topic near and dear to my heart. It was at about this time last year that I was really getting myself up to speed with Google Testing Blog and also Adam Goucher's QA blog.

All of this reading and testing has pointed out some weaknesses in unit testing experience.

1. A developer sitting down to write code is less likely to unit test if they don't already have some idea of unit tests they are to write. If you're not working in an Agile shop, (let's face it, there are lots of us that don't) this can be intimidating. What if you don't write enough tests or test the right way? How should this thing you're supposed to write be unit tested anyway?

2. Code coverage tools, such as this post. So what are they really telling us about tests anyway?

3. How many tests should developers be writing? This question, above all, seems the least answerable. Google had a post that addressed this very question, and even their answer was a very murky, "it depends." They qualify their "it depends," by saying, "it depends on how much confidence the tests can provide in the face of changes made by others." Um...sure. Thanks for the really practical answer (not, really).

Now that unit testing frameworks are more fleshed out, I'd like to see more tools surrounding the tests that address these issues. Have interesting tools or war stories? Please leave a comment.

Related articles by Zemanta

• 12 Unit Testing Tips for Software Engineers
• Quality assurance

Classifier angst

The Bayessian Classifier is working...now I'm onto the Decision Tree classifier. What I noticed immediately is that these two classifiers don't really appear to be useful for the same types of data. Dr. Hung gave me several sets of data and I think that they would all be classified more accurately with the Bayesian Classifier.

Before I started this project, I hadn't really thought about how different data categorizes in different ways. The book I've been using for my algorithms is the excellent book by Toby Seagaran, Programming Collective Intelligence. In the chapter where he compares these classifiers, he points out how data can masquerade as meaningful data, but if it's being manipulated in the wrong way, will be meaningless. It's very important to pick the right type of classifier for the right type of data.

My challenge for coding the decision tree, is to code it so that I can use the iris measurement data AND the data in the Collective Intelligence book. My time is starting to run very short, but so far, the DTC's logic is much easier to code and test.

The categories for both of these have been pretty simple. I'm wondering what would happen if my categories got more complex.