bigquery-graph

Use BigQuery's GoogleSQL GQL syntax with MATCH patterns against a property graph, such as GRAPH project.dataset.graph MATCH (a:Account)-[t:Transfers]->(b:Account) RETURN a.id. Standalone GQL with RETURN is preferred; use GRAPH_TABLE only when you need SQL aggregation or relational joins.

Use the CREATE PROPERTY GRAPH DDL statement, defining NODE TABLES with keys, labels, and scoped properties, plus EDGE TABLES with SOURCE KEY and DESTINATION KEY references. Always assign safe aliases with AS and limit exposed columns via PROPERTIES (col1, col2) for performance.

No. BigQuery graph analytics exclusively supports the BigQuery GoogleSQL GQL standard, which implements the ISO GQL standard. Cypher queries are not supported and must never be generated for BigQuery property graphs.

Query semantic graphs using the GRAPH_EXPAND table-valued function with the full graph name string, never as a regular table. Measure columns marked with is_measure=TRUE must be wrapped in the AGG() function rather than standard aggregations like SUM or COUNT.

GRAPH_EXPAND requires the graph to form a valid tree structure. It does not support disconnected graphs with multiple roots, directed cycles, self-loops, or convergent diamond paths. Workarounds include separate GRAPH_EXPAND calls per component or standard GQL pattern matching.

EXISTS, IN, and LIKE subqueries throw errors when used inside a WHERE clause in BigQuery GQL. Use the FILTER clause instead, and always re-specify the graph name inside the subquery block, since BigQuery requires it unlike standard GoogleSQL.