Empirical software engineering research relies on sampling strategies to enable large-scale analyses of code repositories, where the design and evaluation of these strategies are critical for the generalizability of research findings. We propose a Domain-Specific Language that models sampling strategies as workflows with composable operators and automatically generates statistical indicators to support representativeness reasoning at each stage. Through a case study of MSR papers, we demonstrate the DSL can effectively capture real-world sampling strategies from recent literature.

In this paper, we investigate the energy profiles of Continuous Integration (CI) tasks managed by Maven and Gradle. Build systems like Maven and Gradle, with their internal models, provide opportunities for fine-grained classification, which we leverage to study energy-intensive tasks and their associated profile.

Configurable systems become more complex, feature reduction can help focus on essential features while eliminating bloated and unnecessary ones. Although several studies have explored how feature interactions and runtime performance affect energy consumption, none, to our knowledge, have studied the impact of feature reduction. This paper fills this gap. In particular, we investigate how both on-demand and built-in feature reduction affect the energy consumption of configurable systems.

Annual Software Heritage Summit and Symposium where the first result of our work have been presented at Unesco by Stefano Zacchiroli.