ENRE: A Tool Framework for Extensible eNtity Relation Extraction

Abstract

Understanding the dependencies among code entities is fundamental to many software analysis tools and techniques. However, with the emergence of new programming languages and paradigms, the increasingly common practice of writing systems in multiple languages, and the increasing popularity of dynamic languages, no existing framework can reliably extract this information. That is, no tools exist to accurately extract dependencies from systems written in multiple and dynamic languages. To address this problem, we have designed and implemented the Extensible eNtity Relation Extraction (ENRE) framework. ENRE supports the extraction of entities and their dependencies from systems written in multiple languages, enables the customization of dependencies of interest to the user, and makes implicit dependencies explicit. To demonstrate feasibility of this framework, we developed two ENRE instances for analyzing Python and Golang programs. Our experiments on 12 Python and Golang projects demonstrated the effectiveness and flexibility of ENRE. By comparing with a commercial static analysis tool, we show that we can extract dependencies from Golang programs which are not supported by existing tools and we can reveal implicit dependencies in Python. (Demo Video: https://youtu.be/BfXp5bb1yqc)