| With the popularity of Web Applications, Web Applications testing has played an important role in ensuring the quality and reliability of the application of Web system. Due to the distinguish characters of Web applications, the testing approaches for traditional software testing can not be directly applied to Web applications testing. In the studying of Web Applications testing, testing framework research could instruct the whole testing process from an abstract level, moreover, is crucial for the development and implementation of the testing process. Meanwhile, test case generation is also a key step to impact the testing effectiveness. Hence, based on the existing research achievement, this paper conducts in-depth researches on the studying of Web applications testing framework and the automatic test case generation methodology. By taking advantage of the previous testing concepts, approaches and theory system, we further explore the testing approaches in order to improve the testing strategies.In this paper, we first summarize the testing aspects and methodologies of Web applications, classify and illustrate six different testing approaches, which are Object-based testing, Agent-based testing, UML-based testing, Statistic-based testing, User-Session-based testing and Slicing-based testing, as well as analyze the strength and weakness of each testing approach. Through comparing the existing testing approaches, we systematically understand and acknowledge the mainstream studying methodologies and guide the development of the research work of this paper.In the research of Web Applications testing framework, we emphasize on the design of intelligence and extensibility, integrate the intelligent-field Agent theory into the framework design and propose an Agent-based extended testing framework of Web applications, called AETFW. AETFW composes of two parts, named Web-Modeling and Web-Testing, and they take charge of analyzing Web Applications and test case generation, execution and evaluation respectively. The framework builds models from object, behavior and structure aspects and support both functional and structural testing requirements. According to the characters of Web Applications, the architecture of multi-agent system is divided into several levels and assigns different functional roles to different Agents. Considering to map the Web Applications elements to each Agent, we divide the Test Agent into Object Agents and Object Agent is kind of an atomic Agent which can no longer be divided. Multi-agent system can flexibly construct, delete functional Agent in response to different requirements and this makes the whole testing framework can handle different testing requirements and be generally applicable and extensible.In the research of test case generation, we present a predication-based automatic test case generation algorithm. Firstly, we analyze the dependency relationships of Web Application structure, construct the TG graph by using DUCC expression and transformed dependency relationship, then define the concepts of control node, definition node and k-distance control node. By reversing the TG graph, the proposed algorithm can efficiently figure out all the execution paths from the source node to the basic node. Considering the large scale of Web Application code, we generate the wanted test cases into phases and generate test cases from intra-page and inter-page aspects, aiming to decrease the complexity of testing case generation and enhance the applicability of the reported approach. The preliminary case study shows that the presented test case generation algorithm can produce expected test cases efficiently with certain automatic execution level and meet the prospective research expectation. |
