以常識對安卓應用程式測試之技術與工具

Abstract

由於行動裝置應用市場的高度競爭，軟體測試逐漸成為一項不可或缺的流程。作為介於程式開發者與使用者間的第三方測試者，在無法拿到程式源碼、正規或半正規的規格、或是開發者所撰寫的基礎測試腳本，要對程式進行有效率的測試是非常困難的。我們希望能提出系統化的方式，將人類對程式的理解應用在程式的測試上。 為此我們創造了一個叫作「常識」的模型以人類對畫面與動作理解的概念來對程式行為進行描述。為了統一並減少混淆，我們對各種類的應用程式定義了各自的統一辭彙來描述程式的畫面與動作。又為了有效率的以人類的概念對程式行為賦予意義，我們開發了一套圖型化使用者介面工具叫作「ＳｐｅｃＥｌｉｃｉｔｏｒ」來輔助測試者建立「常識」模型。並且我們提出了兩個依據「常識」模型的演算法，分別是針對測試案例的產生與測試案例的評估。在章節最後討論了我們對「常識」模型在應用的最大目標與其可能所需要的技術。

Software testing becoming an inevitable process due to the highly competitive market of the mobile applications. As a third party between an application developer and customers, testing an application without source codes, formal or semi-formal specification, or any testing scripts written by the developer is hard to be efficiency. We want to find a systematic methodology testing an application by the human understanding of the application. We create a structure named “Common Sense” to model an application behavior with human concepts, and define sets of normalized terms for many kinds of application to descript concepts of screens and actions. To efficiently extract human concepts into a common sense model, we develop a GUI tool called “SpecElicitor”. We also present algorithms generating and evaluating test cases by using common sense models. In the end, we discuss the ultimate objective and potential needs of techniques to achieve our goal.