Documentation Reuse: Hot or Not? An Empirical Study

Having available a high quality documentation is critical for software projects. This is why documentation tools such as Javadoc are so popular. As for code, documentation should be reused when possible to increase developer productivity and simplify maintenance. In this paper, we perform an empirical study of duplications in JavaDoc documentation on a corpus of seven famous Java APIs. Our results show that copypastes of JavaDoc documentation tags are abundant in our corpus. We also show that these copy- pastes are caused by four different kinds of relations in the underlying source code. In addition, we show that popular documentation tools do not provide any reuse mechanism to cope with these relations. Finally, we make a proposal for a simple but efficient automatic reuse mechanism.