TuGraph-DB Query Engine: Graph Query Language Overview, Engine Features, and Architectural Evolution

The presentation introduces the TuGraph-DB database query engine, focusing on three main parts: an introduction to graph query languages, a detailed look at the query engine, and the architecture with its evolution plan.

Graph query language development is divided into three stages: the early graph database era without a dedicated language, the emergence of languages such as Gremlin and Cypher (which became declarative in 2012), and the current iteration phase where GQL is being standardized internationally and adopted by TuGraph since 2023.

Declarative languages (e.g., Cypher, PGQL, G‑CORE) resemble SQL and rely heavily on query optimization, while imperative languages (e.g., GSQL, Gremlin) are more expressive but require higher user expertise.

The TuGraph 4.0 query engine implements GQL support (currently limited to SNB and FINBENCH short queries) and plans to expand to DDL and other features. It also introduces a new GEAX optimization engine to provide a more generic and community‑friendly optimization layer.

Benchmarking with SNB and FINBENCH is discussed, emphasizing that performance, expressiveness, robustness, and optimization are all important evaluation dimensions for a query language.

Architecturally, TuGraph follows a pipeline of parsing to AST, validation, planning, and execution. Challenges include supporting multiple query languages (Cypher and GQL) and integrating various storage engines, prompting a redesign inspired by Apache Calcite.

The upcoming architecture adds a GeaX front‑end that abstracts graph syntax, provides pluggable logical operators, and separates query language processing from execution, enabling easy integration of GQL with other graph engines.

Future work (3‑6 months) includes completing the GeaX front‑end, exposing a playground for users to view logical execution plans, and further refining the optimizer to handle complex queries more effectively.