| The crawl-based method of testing web applications test the web applications automatically by exploring the state space of user interfaces of the applications.However,the state spaces of most Web applications have exponential complexity and exhaustive exploration can easily become mired in irrelevant regions.None of the existing guided-crawl methods take the semantic information of a page into consideration,which results in that these methods can't effectively cover the specified functionalities of a web application.For this,in this thesis,we present a keywords-guided crawl strategy which prefers to cover the specified functionalities of a Web application.The crawler will search for the specified functionalities by calculating the similarity between a state and the keywords given by users,and then construct a subset of the state flow graph(SFG)of user interfaces of the application,which will be regarded as the test model.This model only contains parts of the states and events of the application,but can cover the specified functionalities effectively.Besides,we propose a method called "mark-select" to generate test cases.This method achieves full coverage of the edges of the model,as well as reduces the size of test suites.Based on the above ideas,we implemented a tool of testing Web applications automatically called KeyjaxTest.And empirical study is conducted on 10 open source Web applications.The results show that compared with DFS and BFS,the number of states of the specified functionalities covered by KeyjaxTest is 31.0%and 36.5%higher respectively in a given crawl time,and compared with another tool of guided-crawl—FeedEx,the number is 31.0%higher.Besides,when generating test cases from the SFG constructed by these different methods,the test cases generated from the model constructed by KeyjaxTest can fully cover the specified functionalities,while the number of test cases is 12.75%less in average than those generated from models constructed by DFS and BFS. |
PDF Full Text Request