Discovering Common Anti-patterns Present in Low-Code using Multi-Layered Graph-Based Pattern Mining

Abstract

In recent years Low-Code has seen a surge in popularity amongst companies to speed up their workflows. Yet, scientific work on Low-Code is still in its infancy. We set out to investigate the presence of anti-patterns within Low-Code applications. Given the typically less technically inclined nature of Low-Code developers, as well as the specific use cases of Low-Code in general, we expect that these anti-patterns differ from traditional programming languages. We apply a graph-based methodology to mine edit patterns across real-world commit data supplied to us by Mendix, one of the leading platforms in the Low-Code space. Additionally, we discuss the lack of current guidelines in the Low-Code field. While we are able to find common edit patterns using our approach, linking them to anti-patterns remains difficult in practice. We do establish that Low-Code in Mendix might lack reuse-ability and that the Low-Code often revolves around a few distinct tasks. However, there is a current lack of quality data available to properly assess the development practices of Low-Code developers and anti-patterns, increasing the availability of high-quality data is essential for further research in this area.