bigquery-graph

Use the GRAPH prefix with your graph name followed by MATCH patterns with node and edge labels, then RETURN the results. BigQuery supports only the GoogleSQL GQL standard, never Cypher, and path variables should be assigned for path traversal queries.

Use the CREATE PROPERTY GRAPH DDL statement with NODE TABLES and EDGE TABLES blocks mapping to existing BigQuery tables. Define KEY columns, SOURCE KEY and DESTINATION KEY references for edges, and scope properties explicitly rather than using PROPERTIES ALL COLUMNS.

GRAPH_TABLE integrates GQL pattern matching with standard SQL for aggregations and relational joins. GRAPH_EXPAND flattens a semantic graph into a single table view for SQL analysis, requiring AGG() for measure columns and a tree-structured graph topology.

Measure columns cannot be selected directly and must be wrapped in the AGG() function when querying via GRAPH_EXPAND. Measures are also incompatible with native GQL queries like GRAPH_TABLE, which only operate on standard dimension properties.

GRAPH_EXPAND requires the graph to form a valid tree structure. Cyclic graphs, self-loops, disconnected components with multiple roots, and convergent diamond paths are not supported and must be queried with standard GQL instead.

Use GRAPH_TABLE only when you need SQL capabilities beyond graph pattern matching, such as aggregations with GROUP BY, joins with relational tables, or advanced filtering. For pure graph traversal, standalone GQL with RETURN is preferred.