SHF: Small: Natural GUI-Based Testing of Mobile Apps via Mining Software Repositories

SHF：小型：通过挖掘软件存储库对移动应用程序进行基于 GUI 的自然测试

• 批准号: 1815186
• 负责人: Denys Poshyvanyk
• 金额: $ 45万
• 依托单位: College of William and Mary
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2018
• 资助国家: 美国
• 起止时间: 2018-10-01 至 2022-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1815186&HistoricalAwards=false
• 关键词: SHF; Small; Natural; GUI; Based

Mobile devices have become an integral, ubiquitous part of modern society. The popularity of smartphones and tablets is largely due to the success of mobile software, colloquially referred to as "apps", that enable users to carry out a wide range of computing tasks in an intuitive and convenient manner. The burgeoning mobile app market is fueled by rapidly evolving performant hardware and software platforms that support increasingly complex functionality. In order for apps to achieve success in marketplaces such as Apple's App Store or Google Play, it is imperative that they function as intended and thus must be well tested. However, the unique aspects of mobile apps that make them popular, such as their touch-based interfaces, rapidly evolving platforms, and contextual features such as sensors, make them difficult to test effectively and efficiently. Additionally, as the marketplace for mobile apps matures, developers must ensure that their apps function well across a myriad of devices while addressing feedback from an increasingly large user base through app store reviews. These challenges illustrate that mobile developers require practical automated support to ensure that their apps are adequately tested. This research project aims to design, and thoroughly validate an automated testing approach for mobile apps that overcomes the challenges listed above. In turn, it is anticipated that the techniques enabled by this research will contribute to better-tested, higher quality mobile applications, benefiting both our society that increasingly depends on smartphone apps and the developers and teams that create them. To solve these fundamental challenges, this project aims to develop an automated testing framework that combines novel statistical representations of mobile apps and information gleaned via mining software repositories techniques to efficiently generate practical, effective test scenarios. More specifically, a novel testing framework, coined as T+, will be developed. T+ is rooted in a probabilistic model-based representation of mobile apps. This model will enable a transformative automated approach for generating feasible test cases that are decoupled from low level events, can be executed on different devices, and support multiple testing goals and adequacy criteria. Additionally, this research work will define and develop monitoring mechanisms for identifying change- and fault- prone APIs in underlying platform and third-party libraries, as well as informative reviews. Incorporation of this information into the statistical model of T+ will allow for the generation and prioritization of test cases covering these APIs and reviews. Broader impacts of this work will reside in (1) improving the state of the practice in testing mobile apps, where difficulties are faced in ensuring that apps are adequately tested with respect to changing platforms, APIs, reviews, and numerous devices; (2) demonstrating improved testing practices with industry partners, which will be documented as best practices for other development organizations and test centers to adopt; (3) developing educational course content and piloting it in the classroom as part of this research project; and (4) actively involving underrepresented categories of students in this research program.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

移动设备已成为现代社会不可或缺的、无处不在的一部分。智能手机和平板电脑的普及很大程度上归功于移动软件（俗称“应用程序”）的成功，这些软件使用户能够以直观、方便的方式执行各种计算任务。快速发展的高性能硬件和软件平台支持日益复杂的功能，推动了蓬勃发展的移动应用市场。为了让应用程序在 Apple App Store 或 Google Play 等市场取得成功，它们必须按预期运行，因此必须经过充分测试。然而，移动应用程序的独特之处使其广受欢迎，例如基于触摸的界面、快速发展的平台以及传感器等上下文功能，这使得它们难以有效且高效地进行测试。此外，随着移动应用程序市场的成熟，开发人员必须确保他们的应用程序能够在多种设备上正常运行，同时通过应用程序商店评论解决日益庞大的用户群的反馈。这些挑战表明，移动开发人员需要实用的自动化支持，以确保他们的应用程序得到充分的测试。该研究项目旨在设计并彻底验证移动应用程序的自动化测试方法，以克服上述挑战。反过来，预计这项研究支持的技术将有助于开发出经过更好测试、质量更高的移动应用程序，从而使我们日益依赖智能手机应用程序的社会以及创建这些应用程序的开发人员和团队受益。 为了解决这些基本挑战，该项目旨在开发一个自动化测试框架，该框架结合了移动应用程序的新颖统计表示和通过挖掘软件存储库技术收集的信息，以有效生成实用、有效的测试场景。更具体地说，将开发一种新颖的测试框架，称为 T+。 T+ 植根于基于概率模型的移动应用程序表示。该模型将实现一种变革性的自动化方法，用于生成与低级别事件分离的可行测试用例，可以在不同设备上执行，并支持多个测试目标和充分性标准。此外，这项研究工作将定义和开发监控机制，用于识别底层平台和第三方库中容易变化和容易出错的 API，以及信息丰富的评论。将此信息合并到 T+ 的统计模型中将允许生成涵盖这些 API 和评论的测试用例并确定其优先级。这项工作的更广泛影响将在于 (1) 改善移动应用程序测试的实践状况，在确保应用程序针对不断变化的平台、API、评论和众多设备进行充分测试方面面临着困难； (2) 与行业合作伙伴一起展示改进的测试实践，这些实践将被记录为最佳实践，供其他开发组织和测试中心采用； (3) 作为本研究项目的一部分，开发教育课程内容并在课堂上进行试点； (4) 积极让代表性不足的学生参与该研究计划。该奖项反映了 NSF 的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Machine Learning-Based Prototyping of Graphical User Interfaces for Mobile Apps

基于机器学习的移动应用程序图形用户界面原型设计

• DOI: 10.1109/tse.2018.2844788
• 发表时间: 2018-02-07
• 期刊: IEEE Transactions on Software Engineering
• 影响因子: 7.4
• 作者: Kevin Moran;Carlos Bernal;Michael Curcio;R. Bonett;D. Poshyvanyk
• 通讯作者: D. Poshyvanyk

An Empirical Study on Learning Bug-Fixing Patches in the Wild via Neural Machine Translation

通过神经机器翻译在野外学习错误修复补丁的实证研究

• DOI: 10.1145/3340544
• 发表时间: 2018-12-20
• 期刊: ACM Transactions on Software Engineering and Methodology (TOSEM)
• 作者: Michele Tufano;Cody Watson;G. Bavota;M. D. Penta;Martin White;D. Poshyvanyk
• 通讯作者: D. Poshyvanyk