SHF: Medium: Automated Graphical User Interface Testing with Learning

具有丰富图形用户界面（gui）的智能手机和平板电脑正变得越来越流行。这些移动平台上已经有成千上万的专门应用程序，这些应用程序被称为应用程序，而且新发布的应用程序的数量还在不断增加。这些应用程序的复杂性通常在于用户界面，数据处理要么次要，要么委托给后端组件。在使用软件即服务体系结构的应用程序中也存在类似的情况，其中客户端组件主要由用户界面代码组成。测试这类应用程序主要涉及GUI测试。现有的用于测试这些接口的自动化技术要么需要接口的先验模型，因此很难使用，要么通过向应用程序发送随机用户事件来盲目操作，并且通常无法对应用程序进行令人满意的深度测试。这个项目研究了自动GUI测试技术，系统地探索应用程序的状态空间，而不需要先验定义的模型。这个项目背后的一个见解是，用户界面模型的自动构建和界面的测试是可以以互利的方式合作的任务。此外，贯穿本研究的指导原则是设计使用抽象和启发式操作的算法，这些算法足够简单，可以被不一定了解被测试应用程序内部的人类理解。这样的算法更容易理解，并纳入一个整体的测试过程，包括自动化技术，例如在这个项目中开发的技术，以及手动测试和指导。本项目开发的技术直接使这些应用程序的程序员受益，并间接使移动和web应用程序的众多用户受益。

Smartphones and tablets with rich graphical user interfaces (GUIs) arebecoming increasingly popular. Hundreds of thousands of specializedapplications, called apps, are already available for these mobileplatforms, and the number of newly released apps continues toincrease. The complexity of these apps lies often in the userinterface, with data processing either minor, or delegated to abackend component. A similar situation exists in applications usingthe software-as-a-service architecture, where the client-sidecomponent consists mostly of user interface code. Testing suchapplications predominantly involves GUI testing. Existing automatictechniques for testing these interfaces either require a priori modelsof the interface and are thus hard to use, or operate blindly bysending random user events to the application and are typically unableto test the application in satisfactory depth.This project investigates automatic GUI testing techniques thatsystematically explore the state space of an application withoutrequiring an a priori defined model. One insight behind this projectis that the automatic construction of a model of the user interfaceand the testing of the interface are tasks that can cooperate in amutually beneficial way. Furthermore, a guiding principle throughoutthis research is to design algorithms that operate with abstractionsand heuristics that are simple enough to be understood by humans whodo not necessarily understand the internals of the tested app. Suchalgorithms are easier to comprehend and to incorporate into awholistic test process that includes automated techniques, such as theones developed in this project, and manual testing and guidance. Thetechniques developed in this project benefit directly programmers forthese apps, and indirectly the numerous users of mobile and webapplications.