Software Developer Activity as a Source for Identifying Hidden Source Code Dependencies

Abstract

Connections between source code components are important to know in the whole software life. Traditionally, we use syntactic analysis to identify source code dependencies which may not be sufficient in cases of dynamically typed programming languages, loosely coupled components or when multiple programming languages are combined. We aim at using developer activity as a source for identifying implicit source code dependencies, to enrich or supplement explicitly stated dependencies in the source code. We propose a method for identification of implicit dependencies from activity logs in IDE, mainly of switching between source code files in addition to usually used logs of copy-pasting code fragments and commits. We experimentally evaluated our method using data of students’ activity working on five projects. We compared implicit dependencies with explicit ones including manual evaluation of their significance. Our results show that implicit dependencies based on developer activity partially reflect explicit dependencies and so may supplement them in cases of their unavailability. In addition, implicit dependencies extend existing dependency graph with new significant connections applicable in software development and maintenance.