Automated translation of Android context-dependent gestures to visual GUI test instructions

Abstract

Layout-based (2nd Generation) and Visual (3rd Generation) GUI testing are two very common approaches for mobile application testing. The two techniques expose complementary advantages and drawbacks, and the literature on GUI Testing has highlighted the benefits of an approach based on a translation from one generation to the other. The objective of this work is to provide an improvement to our prototype tool, TOGGLE, designed to translate 2nd Generation test suites, written with the Espresso framework, to 3rd Generation ones that can be run by the EyeAutomate and Sikuli tool. We extended TOGGLE by adding (1) support for context-based gestures, performed through the scrollTo and onData commands, and (2) support for the combination of Layout-based locators with logical operators. We evaluated the new version of the tool on five different experimental subjects. For each of the applications, 30 test cases were developed and automatically translated with TOGGLE+. We observed an increase of 68% of translatable test cases when transitioning from the previous prototype to the current version of the tool. The generated Visual test cases also proved to have high robustness, with flakiness of just 2% (i.e., 98% correct executions).