Understanding Postgres 19 Property Graphs

8 hours ago

Postgres 19 Property Graphs allow declaring existing tables as nodes and edges without moving data, with queries compiled to relational joins.
A property graph consists of vertices and edges, defined via CREATE PROPERTY GRAPH as an overlay on foreign-key structures, queried using MATCH patterns.
Dimension tables map to vertices, pure join tables to edges, while fact tables with multiple foreign keys often become hub vertices with separate edge tables.
Key columns are not automatically queryable properties; they must be listed in PROPERTIES. Variable-length paths are not supported, requiring recursive CTEs for deep traversals.
A single table can serve as both a vertex and edge simultaneously by aliasing it in VERTEX TABLES and EDGE TABLES, avoiding duplicate storage.
Property graphs excel at fixed-shape pattern matching (e.g., finding connected entities) but lack support for variable-length paths or graph algorithms like PageRank.
The feature integrates with Postgres' optimizer and tools (e.g., \dG for introspection), making it suitable for graph-shaped queries on existing schemas without external graph databases.

Hasty Briefsbeta