Versatile Property Graph Transformations

Property graphs are key components of modern graph database systems as well as graph analytical systems. They support highly expressive data models consisting of multi-labeled nodes and edges, along with properties represented as key/value pairs. Property graphs serve as versatile data integration paradigms, enabling data in any format to be seamlessly transformed into this model. Moreover, they are at the core of an active standardization effort led by ISO/IEC, which aims to establish standardized declarative graph query languages such as GQL and SQL/PGQ. In addition to these standards for data manipulation languages, other languages have emerged for property graph schemas and constraints as part of future data definition languages. In this paper, we introduce a new declarative paradigm for expressing property graph transformations, supporting both graph data integration and data cleaning tasks. We discuss the properties of these transformations, along with algorithmic issues and considerations for efficiency and scalability. Furthermore, we showcase the utility of property graph transformations for causal analysis and elaborate on a research agenda aimed at designing analytical extensions of graph languages to support property graph transformations for advanced analytical workloads on heterogeneous data.