Web application testing as an automated service

点击翻译按钮获取中文摘要

Facilitated by recent Web 2.0 technologies, people are increasingly using the web as a programming platform. Yet, due to the distributed and dynamic nature of web applications, testing them is still a largely manual and daunting task. Consequently, the dependability risks of such applications become progressively significant. This project aims to study how software testing can be made readily available to web developers in a more feasible and interactive way. The intent is to experimentally investigate new ways of maximizing test automation and minimizing the effort required to express the desired program behavior for testing. In particular, this project proposes to provide web testing as an interactive automated service, in which the web application is automatically executed in a real browser to infer run-time observations. These observations are subsequently augmented by the user and used to generate test cases. The challenges consist of automatically analyzing and inferring web program behavior and generating robust regression comparators. Successful completion of the proposed research will establish a novel analysis and testing method for the domain of web applications. Key innovations include (1) new techniques for automatically finding run-time structural and code-level observations; (2) new methods for generating robust and reusable test suites to prevent regressions; (3) implementation of a prototype framework that can be entirely accessed as an automated service; and (4) engaging web developers in the analysis and testing process through interactive visualization techniques. The empirical evaluation, through a range of case studies on open-source and industrial systems, will focus on whether the developed methods contribute to increasing web application dependability. The impact will be significant on web testing practice, not only for web developers, but also for professional software testers.