On the Variations and Evolutions of API Usage Patterns: Case Study on Android Applications

Abstract

Software developers can reduce the implementation cost by calling already provided functions through accessing library Application Programming Interface (API). APIs are often used in combination but how to combine them are not well-documented. Existing researches focused on how to extract API usage patterns or how to detect API misuse from existing software. This kind of research might be affected by dataset to analyze, so to improve mining results and to understand how the difference of API usage patterns affect the software health are important tasks. We conducted an analysis on variations of API usage pattern among software projects and their version history with Android SDK APIs and Android applications. Based on our analysis results, we made some suggestions for further API analysis. For example, there are many project-specific API usage patterns and long-life uncommon API usage patterns so that they might affect the mining result or checking software health status.