Context-Aware Software Documentation

• Published in: 2018 IEEE International Conference on Software Maintenance and Evolution (ICSME) pp. 727 - 731
• Main Author: Aghajani, Emad
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.09.2018
• Subjects: Cloning; context-aware; Data mining; Documentation; Knowledge based systems; program comprehension; recommender system; Recommender systems; Software; software documentation; Task analysis
• DOI: 10.1109/ICSME.2018.00090

Software developers often do not possess the knowledge needed to understand a piece of code at hand, and the lack of code comments and outdated documentation exacerbates the problem. Asking for the help of colleagues, browsing the official documentation, or accessing online resources, such as Stack Overflow, can clearly help in this "code comprehension" activity that, however, still remains highly time-consuming and is not always successful. Enhancing this process has been addressed in different studies under the subject of automatic documentation of software artifacts. For example, "recommender systems" have been designed with the goal of retrieving and suggesting relevant pieces of information (e.g., Stack Overflow discussions) for a given piece of code inspected in an IDE. However, these techniques rely on limited contextual information, mainly solely source code. Our goal is to build a context-aware proactive recommender system supporting the code comprehension process. The system must be able to understand the context, consider the developer's profile, and help her by generating pieces of documentation at whatever granularity is required, e.g., going from summarizing the responsibilities implemented in a subsystem, to explaining how two classes collaborate to implement a functionality, down to documenting a single line of code. Generated documentation will be tailored for the current context (e.g., the task at hand, the developer's background knowledge, the history of interactions). In this paper we present our first steps toward our goal by introducing the ADANA project, a framework which generates fine-grained code comments for a given piece of code.