G-CORE: A Core for Future Graph Query Languages

Reviewer: Jaroslav Pokorny

With a description of the G-CORE language, designed by the Linked Data Benchmark Council (LDBC) Graph Query Language Task Force, this paper presents a standardization proposal of a graph query language for property graphs. In section 1, the authors focus on three main challenges to existing graph query languages: "composability," "paths as first-class citizens," and "capture the core of available languages." By composability the authors mean that "graphs are the input and the output of queries." Their approach addresses the third challenge by taking "the successful functionalities of current [graph query] languages ... to develop the next generation of graph languages. Section 2 discusses the second challenge and defines an extended path property graph (PPG) model. PPGs allow for multi-valued properties and stored paths. Paths have an identity and can also have labels. Queries on paths are also enabled. The core section (3) "demonstrate[s] and explain[s] the main features of the G-CORE language," for example, returning a graph for every query, matching and filtering, multi-graph queries and joins, dealing with multi-valued properties, constructions that respect identities, graph aggregations, treatment of path, existential subqueries, views and optional matching, weighted shortest paths, and the use of graph patterns. All constructs are explained in examples. In section 4, the authors "provide a formal definition of the syntax and semantics of [G-CORE]." A complexity analysis is also considered here. The authors prove that "evaluating [a query, q ] over an input PPG G can be computed in polynomial time," that is, G-CORE is tractable. In section 5, the authors "show how G-CORE is extended to handle tabular data." This approach is in accordance with today's trends to integrate heterogeneous data and to develop polyglot databases. Section 6 presents related works, that is, the most important industrial graph database products represented by their graph query languages, Gremlin, Cypher, and PGQL. The authors "describe the main differences among G-CORE [and these languages]." Finally, in the conclusion, one more use is emphasized: G-CORE could be used as a base for integrating many graph-oriented data models and approaches to querying graphs. The notions defined in the paper are specified in the usual denotational way, which provides the needed clarity and preciseness. A lot of representative examples increase the paper's readability. Without doubt, the paper offers interesting, valuable, and useful information for those interested in graph query languages.