Quick Start

Ultipa GQLDB supports stored procedures for encapsulating complex graph logic, enabling reusable queries, and implementing high-performance graph algorithms.

1. Create a Procedure

GQL
CREATE PROCEDURE greet(name: STRING)
RETURNS (message: STRING)
AS {
    RETURN 'Hello ' || $name AS message
}

2. Call It

GQL
CALL greet('World') YIELD message
-- Returns: "Hello World"

3. Drop It

GQL
DROP PROCEDURE greet

About Stored Procedures

Stored procedures in Ultipa GQLDB are an extension to the GQL standard (ISO/IEC 39075). The GQL standard does not yet define syntax for creating or dropping stored procedures - CREATE PROCEDURE, DROP PROCEDURE, SHOW PROCEDURES, and the procedure body language are all Ultipa-specific extensions.

The procedure body (AS { ... }) is written in a procedural language that builds on GQL syntax. It supports GQL statements like MATCH, INSERT, SET, DELETE, RETURN, and all standard GQL functions. On top of that, it adds:

• Control flow: IF / ELSE IF / ELSE, FOR, WHILE, BREAK, CONTINUE
• Error handling: TRY / CATCH / THROW
• Parallel execution: PARALLEL FOR with WORKERS
• Variables: LET for declaration and assignment
• Debugging: PRINT for output to stderr
• Temporary node properties (in-memory only, not persisted)
• High-performance data structures: slice properties, neighbor aggregation functions, topology functions

Compute Engine (Optional)

Stored procedures work without the compute engine. Control flow (IF, FOR, WHILE, TRY/CATCH), data operations (LET, MATCH, INSERT, SET, DELETE), standard GQL functions, set/map operations, and type conversions are all available by default.

For graph algorithm workloads, enable the compute engine per graph:

GQL
ALTER GRAPH <graph_name> SET COMPUTE ENABLED

This is required for:

• Topology functions: OUT_DEGREE, IN_DEGREE, NODE_COUNT
• Fused neighbor operations: SUM_OUT_NEIGHBOR_PROP, IN_NEIGHBOR_SUM, MIN_BOTH_NEIGHBOR_PROP, etc.
• Slice properties: INIT_SLICE_PROP, GET_SLICE_PROP, SET_SLICE_PROP, PARALLEL FOR with SCAN()
• Parallel reductions: SUM_SLICE_PROP, MAX_SLICE_PROP, etc.
• Common neighbor functions: COMMON_NEIGHBORS, JACCARD_SIMILARITY, ADAMIC_ADAR
• Batch operations: BATCH_MAP_TO_SLICE, BATCH_PERSIST_SLICE, etc.

Without the compute engine, these functions return 0 or default values.