The analysis of the different characteristics of commits between developers with different experience level: An archival study

Background: With the development of software, its quality is increasingly valued by people. The developing technical ability was absolutely taken to underpin the performance of the developer, and code quality was raised as being related to developer performance, thus code quality could be a measure of developer performance. Developer performance is often influenced by multiple factors. Also, different factors have different impacts on developer performance in different project types. It is important for us to understand the positive and negative factors for developer performance in a certain project. If these factors are valued, developers will have better performance and project will have higher quality. Objectives: The objective of our study is to identify how factors (developer experience, task size, and team size) impact the developer performance in each case. Though understanding how factors impact the developer performance, developers can have a better performance, which is a big benefit to the quality of project. Methods: We decided to use the characteristics of commits during the Gerrit code review to measure the committed code quality in our research, and from committed code quality we can measure the developer performance. We selected two different projects which use Gerrit code review as our cases to conduct our archive study. One is the legacy project, another is the open-source project. Then we selected five common characteristics (the rate of abandoned application code, the rate of abandoned test code, abandoned lines of application code, abandoned lines of test code and build success rate) to measure the code quality The box plot is drawn to visualize the relationship between the factor experience and each characteristic of the commits. And Spearman rank correlation is used to test the correlation between each factor and characteristic of commits from the statistical perspective. Finally, we used the multiple linear regression to test how a specific characteristic of commits impacted by the multiple factors. Results: The results show that developers with high experience tend to abandon less proportion of their code and abandon less lines of code in the legacy project. Developers with high experience tend to abandon less proportion of their code in the open-source project. There is a similar pattern of the factor task size and the factor amount of code produced in these two cases. Bigger task or more amount of code produced will cause a great amount of code abandoned. Big team size will lead to a great amount of code abandoned in the legacy project. Conclusions: After we know about how factors (experience, task size, and team size) influence the developers' performance, we have listed two contributions that our research provided: 1. Big task size and big team size will bring negative impact to the developer performance. 2. Experienced developers usually have better performance than onboarded developers. According to these two contributions, we will give some suggestions to these two kinds of projects about how to improve developer performance, and how to assign the task reasonable