During the summer of 2008, Lucy Simko '11 worked on research with Professor Sara Sprenkle and Professor Lori Pollock from the University of Delaware.
Abstract: Web application code is often large and complicated because it must handle millions of user requests dynamically and process massive amounts of information quickly and efficiently. Therefore, web application code is prone to errors and requires effective testing to expose errors. Testing based on user requests is favorable because user requests are cheap to record and testing focuses on what users actually do. In 2005, Sant et al. proposed a promising user-based test case generation approach. However, their algorithms for choosing parameter values were limited to value combinations from the original user requests, and their evaluation was based on only one application. In this poster, we propose two new algorithms that address the limitation and evaluate Sant et al.'s and our algorithms on four applications. Our results indicate that Sant et al.'s algorithms more closely model user sessions and demonstrate that there are benefits to creating parameter combinations beyond those in the original user sessions.