Using ChatGPT to Generate Unit Tests

In this article, I will demonstrate how I am using ChatGPT to write unit tests for our system in a consistent and predictable fashion. I was curious if I could take the scopes from our code and use ChatGPT to generate unit tests for them. ChatGPT was able to come up with a couple of unit tests, but they needed some modifications to work within our system. By providing additional contextual information, I was able to improve the generated code. I decided to further test ChatGPT by using it to generate unit tests for more complex scopes. I provided it with a copy of our code and asked it to create tests for specific scopes. ChatGPT generated multiple tests, some of which worked perfectly and others that uncovered bugs in our code. This ability to uncover unknown problems is a powerful aspect of unit testing.

While ChatGPT was able to generate code that covered all scenarios, there were some areas where it could benefit from further training. For example, it didn't follow best practices for setting up data. However, this can be improved by training future bots with better patterns. I have also explored using ChatGPT to generate code in different programming languages, such as Python, Ruby, and JavaScript. With some pre-training and meta programming, I was able to get consistent code that followed desired patterns and practices. Overall, using ChatGPT to generate unit tests has been a valuable tool in our development process. It has saved time and helped uncover hidden bugs. With further training and refinement, it has the potential to become an even more powerful tool for code generation and testing.