Anmerkungen:
Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
Beschreibung:
The variability of highly configurable systems introduces new challenges and requires new and flexible testing strategies to ensure their quality. Product sampling proved to be one of the most promising strategies to reduce the testing effort by computing a set of products (i.e., a sample) that represents the whole system in the testing phase. A scientific survey about classifying product sampling for software product lines identified more than 48 algorithms to compute samples. We extended the survey with 16 new or missed publications and classified them accordingly to provide a more complete overview. However, the large amount of available sampling algorithms make the user’s process to select an appropriate one for their project more complex. Further, the algorithms focus on different objectives (e.g., the runtime of the sampling process or the size of the resulting sample) and there is no complete comparison between all of them. Coupled with the problem of performing redundant evaluations each time a new sampling algorithm is introduced motivated us to design a framework that automatically compares sampling algorithms. Moreover, users often lack the required expert knowledge to understand a complete comparison. This motivated us to provide a strategy to compute recommendations of sampling algorithms that consider the user’s requirements. We performed an empirical evaluation on four sampling algorithms with more than 160 real-world systems, including industrial-sized models from the financial services and automotive domain. Based on the data generated by our framework, we concluded that the modern sampling algorithm YASA achieves the best results for multiple objectives. Furthermore, we concluded that our strategy to compute recommendations, named Weighted Rank-Based Score (WRBS), produce correct and precise results. ; Die Variabilität hochgradig konfigurierbarer Systeme bringt neue Herausforderungen mit sich und erfordert neue und flexible Teststrategien zur Sicherung ihrer Qualität. Als eine der vielversprechendsten ...