一句话：工业混合方法研究实证表明，以多模态 DSL 将 GUI 测试抽象到 ViewModel 层可降低测试维护开销。
该研究针对用通用语言编写 ViewModel 测试时样板代码多、断言链复杂、维护成本高的问题，评估多模态 DSL 抽象到 ViewModel 层对维护开销的影响。以 ViMoTest 为代表性实现，作者先通过行动研究与案例研究确认其工业适用性，再开展基于任务的工业专家评估，并用半结构化访谈与问卷三角验证。定量结果显示，相比 C++ 工程基线，ViMoTest 明显降低测试规约的工作量与规模；定性反馈表明可视化断言与稳健重构能力减少了测试结构演进中的认知负担。研究为多模态 DSL 提升 ViewModel 测试可维护性提供了实证证据。
As in many modern software engineering domains, the quality of an application's Graphical User Interface (GUI) is fundamentally verified by automated testing. One established approach to mitigate testing challenges is to leverage architectural patterns, such as Model-View-ViewModel (MVVM), to decouple presentation logic and test it in isolation. However, developing these ViewModel tests in general-purpose languages (GPLs) often forces engineers to write excessive boilerplate code for test setup and complex assertion chains. While multi-modal domain-specific languages (DSLs) show promise in mitigating these GPL-related issues, empirical evidence of their impact on test maintainability in practice is lacking. To address this gap, we present an industrial mixed-methods study evaluating how abstracting GUI tests to the ViewModel level via multi-modal DSLs impacts maintenance overhead in practice, using the ViMoTest method as our representative realization. After establishing industrial applicability via a pre-study comprising action research and a case study, we conducted a task-based evaluation with industry experts, triangulated with semi-structured interviews and a questionnaire. Quantitative results indicate that the ViMoTest method shows a strong tendency to reduce test specification effort and size compared to a C++ engineering baseline. Furthermore, qualitative expert feedback indicates that visual assertions and robust refactoring capabilities reduce cognitive load during structural test evolution. Overall, this study provides empirical evidence that multi-modal DSLs positively impact the maintainability of ViewModel tests.
@inproceedings{paperbot12274,
title = {Evaluating the Maintainability of ViewModel Tests using Multi-modal DSLs: An Industrial Mixed-Methods Study},
author = {Mario Fuksa and Sandro Speth and Steffen Becker},
booktitle = {ASE},
year = {2026},
doi = {10.1145/3832783.3834370}
}