Constraints

Abstract

Graph-shaped data differs from structured data mainly because of the lack of an underlying schema and metadata. Graph datasets typically blend values with metadata information without a clear distinction among them. An important class of metadata is given by integrity constraints and dependencies, whose goal is to impose the adherence to a specified structure. Constraints and dependencies may serve the need of imposing a limited schema, solving inconsistencies among different parts of the graphs in order to fulfill data cleaning tasks or, in the case of inter-source constraints, aiding the process of exchanging data among different graph data sources. Graphs arising in many applications exhibit inconsistencies and conflicts that are inherent to the processes of graph generation, graph reasoning, and graph data integration and fusion. Indeed, graphs are often obtained by direct translation from other formats and sets of graphs are integrated into a target graph without paying attention to possible conflicts that may arise in the process at hand. Due to these reasons, the data quality of graph datasets needs to be defined in a principled manner in order to be able to detect conflicts and possibly repair them. Graph-to-graph dependencies are also used to guide the process of translating a graph data source into a target data source. These dependencies rely on graph queries to extract data from a source graph and cast it into the format given by a target graph.