Towards Code Improvements Suggestions from Client Exception Analysis

Modern software development heavily relies on reusing third-party libraries; this makes developers more productive, but may also lead to misuses or other kinds of design issues. In this paper, we focus on the exceptional behavior of library methods, and propose to detect client code that may trigger such exceptional behavior. As we demonstrate on several examples of open-source projects, exceptional behavior in clients often naturally suggests improvements to the documentation, tests, runtime checks, and annotations of the clients. In order to automatically detect client calls that may trigger exceptional behavior in library methods, we show how to repurpose existing techniques to extract a method's exception precondition-the condition under which the method throws an exception. To demonstrate the feasibility of our approach, we applied it to 1,523 open-source Java projects, where it found 4,115 cases of calls to library methods that may result in an exception. We manually analyzed 100 of these cases, confirming that the approach is capable of uncovering several interesting opportunities for code improvements.