The Design And Implementation Of A Sketch-based Android GUI Testing System

With the development of hardware technology and continuous optimization and upgrading of the Android System,the interfaces and functions of mobile applications are becoming more and more complex.However,interface testing for these applications are very difficult.The current mainstream technology is manually writing,record and playback technology and some other automatically GUI testing technologies.Manually writing is tedious,inefficient and fallible.Generating test cases with record and playback technology usually requires repetitive work,because only one test case can be recorded during each execution phase.Some automatic technologies focus more on the combination of GUI events.Because of the combination number is huge,these automatic technologies often result in redundant test cases which make them inefficient.In this thesis,we make some research on the existing android testing technologies,elaborate the system of android graphical user interface,mainstream testing model for android graphical user interface and testing framework for android applications.Based on these technologies,we introduce a sketch-guided technology and improve a testing system to generate test cases for mobile applications.Firstly,testers just need to draw some strokes on the screenshots.Then the testing system can recognize these strokes and translate them to a testing model.Finally,the testing system can generate a batch of test cases based on the testing model.Compared with record and playback technology,our approach can improve testing efficiency because a batch of test cases will be generated with one stroke.Compared with some automatic technologies,our approach can enable testers specify their test purpose,thus reduce redundant test cases under the guide of testers.To validate the effectiveness of this method,we improved a testing system firstly.It consists of three parts:an Android APP for sketching,a model-building system and a test-case generating system.Testers can choose screenshots and draw strokes with the Android APP,then the APP will recognize the strokes using a machine-learning algorithm and record these strokes in XML format.In this thesis,we design a testing model based on event-flow model,the second part will generate this model with the recording strokes.The last part will use the testing model to generate test cases which are based on Robotium testing framework.In this thesis,we collected some Android APPs to do the experiments and chose some automatic testing technologies to evaluate our approach.The results show that compared with existing automatic technologies,our approach is more powerful in reaching higher testing coverage,and can reach high test coverage within a short time.So we can claim that our approach can improve testing efficiency for testers.