Module operator 

Physical operators for query execution using the Volcano iterator model.

Volcano Iterator Model (ADR-007)

The Volcano model (Goetz Graefe, 1990s) is the dominant query execution paradigm in relational and graph databases. Each operator implements a next() method that returns one record at a time, pulling from child operators on demand. This creates a pipeline where data flows from leaf operators (scans) up through filters, joins, and projections to the root operator that produces final results.

Physical Operators

Operator	Description
NodeScanOperator	Scans all nodes matching a label (like a table scan in SQL)
IndexScanOperator	Uses a B-tree index to find nodes matching a predicate
FilterOperator	Evaluates a WHERE predicate, discarding non-matching records
ExpandOperator	Traverses edges from bound nodes to discover neighbors (graph-native; no SQL equivalent without expensive JOINs)
ExpandIntoOperator	Checks if an edge exists between two already-bound nodes (a semi-join)
ProjectOperator	Evaluates RETURN expressions, materializing NodeRef → Node for output
LimitOperator / SkipOperator	LIMIT and SKIP clauses
SortOperator	ORDER BY with multi-key comparison
AggregateOperator	GROUP BY + aggregation functions (count, sum, avg, min, max, collect)
JoinOperator	Hash join for multi-pattern MATCH queries
LeftOuterJoinOperator	For OPTIONAL MATCH (preserves unmatched left rows with NULLs)
CartesianProductOperator	Cross product for disconnected patterns
UnwindOperator	Expands arrays into individual rows
MergeOperator	MERGE (upsert): CREATE if not exists, otherwise match
ShortestPathOperator	BFS/Dijkstra for shortestPath() function
VectorSearchOperator	HNSW approximate nearest neighbor search
AlgorithmOperator	Runs graph algorithms (PageRank, WCC, SCC, etc.)
DDL operators	CreateIndex, DropIndex, CreateConstraint, ShowIndexes, etc.

Expression Evaluation

The eval_expression() function recursively evaluates AST expressions against a record. It handles property access (n.name), arithmetic (a + b), comparisons (a > b), boolean logic (AND/OR/NOT), function calls (toUpper(), count()), CASE expressions, list operations, and more.

Type Coercion and NULL Propagation

Integer/Float automatic promotion (widening), String concatenation via +, and NULL propagation following three-valued logic: any operation involving NULL returns NULL, except IS NULL / IS NOT NULL.

Late Materialization

Operators work with Value::NodeRef(id) instead of full Value::Node(id, clone). Property access goes through resolve_property(), which looks up the property from the GraphStore on demand. Full materialization only happens at ProjectOperator when the query returns a node variable. See ADR-012.

Metaheuristic Optimization Solvers

AlgorithmOperator integrates 16 solvers from graphmind-optimization (Jaya, Rao, TLBO, Firefly, Cuckoo, GWO, GA, SA, Bat, ABC, GSA, NSGA2, MOTLBO, HS, FPA) for solving continuous optimization problems within graph queries.

Rust Patterns

• Box<dyn PhysicalOperator> — dynamic dispatch via trait objects for operator trees
• &GraphStore — lifetime-bounded borrow of the graph during execution
• HashMap — build phase of hash joins in JoinOperator
• BTreeSet — sorted unique results where ordering matters

Structs

AggregateFunction

Aggregation function definition

AggregateOperator

Aggregate operator: GROUP BY + Aggregations

AlgorithmOperator

Algorithm operator: CALL algo.pageRank(…)

CartesianProductOperator

Cartesian product operator: MATCH (a:X), (b:Y) Produces all combinations of records from left and right inputs

CompositeCreateIndexOperator

Composite create index operator: CREATE INDEX ON :Label(prop1, prop2, …)

CreateConstraintOperator

Create unique constraint operator

CreateEdgeOperator

Create edge operator: CREATE (a)-[:KNOWS]->(b)

CreateIndexOperator

Create property index operator: CREATE INDEX ON :Person(id)

CreateNodeOperator

Create node operator: CREATE (n:Person {name: “Alice”})

CreateNodesAndEdgesOperator

Combined operator for CREATE patterns with both nodes and edges Example: CREATE (a:Person)-[:KNOWS]->(b:Person) This operator first creates all nodes, then creates edges between them

CreateVectorIndexOperator

Create vector index operator: CREATE VECTOR INDEX …

DeleteOperator

Delete operator: DELETE n or DETACH DELETE n

DropIndexOperator

Drop index operator: DROP INDEX ON :Label(property)

ExpandIntoOperator

ExpandInto operator: checks whether an edge exists between two already-bound endpoints.

ExpandOperator

Expand operator: -[:KNOWS]->

FilterOperator

Filter operator: WHERE n.age > 30

ForeachOperator

FOREACH operator: FOREACH (x IN list | SET x.prop = val)

IndexScanOperator

Index scan operator: MATCH (n:Person) WHERE n.id = 1

JoinOperator

Join operator: Joins two inputs on a shared variable

LeftOuterJoinOperator

Left outer join operator for OPTIONAL MATCH Iterates left records and probes right records by join variable. When no right match exists, emits the left record with NULL for right-only variables.

LimitOperator

Limit operator: LIMIT 10

MatchCreateEdgeOperator

Operator for MATCH…CREATE queries Example: MATCH (a:Trial {id: 'NCT001'}), (b:Condition {mesh_id: 'D001'}) CREATE (a)-[:STUDIES]->(b) This operator takes matched nodes and creates edges between them

MergeOperator

MERGE operator - upsert: match or create pattern

MockProcedureOperator

Mock procedure operator for TCK test procedures

NodeByIdOperator

NodeById operator: start from a specific set of node IDs.

NodeScanOperator

Node scan operator: MATCH (n:Person)

OperatorDescription

Description of an operator for EXPLAIN output

PerRowCreateOperator

Per-row CREATE operator: for each input record, creates nodes with properties resolved from the input record’s variable bindings. Used by UNWIND+CREATE.

PerRowMergeOperator

Per-row MERGE operator for the Apply pattern in multi-part queries. For each input row, performs a MERGE (find-or-create) on a node pattern, binding the result to the record for downstream operators.

ProjectOperator

Project operator: RETURN n.name, n.age

RemovePropertyOperator

Remove property operator: REMOVE n.name

SchemaVisualizationOperator

Schema visualization operator: CALL db.schema.visualization()

SetPropertyOperator

Set property operator: SET n.name = “Alice”

ShortestPathOperator

ShortestPathOperator - finds shortest path(s) between two nodes using BFS

ShowConstraintsOperator

Show constraints operator: SHOW CONSTRAINTS

ShowIndexesOperator

Show indexes operator: SHOW INDEXES

ShowLabelsOperator

Show labels operator: CALL db.labels()

ShowPropertyKeysOperator

Show property keys operator: CALL db.propertyKeys()

ShowRelationshipTypesOperator

Show relationship types operator: CALL db.relationshipTypes()

ShowVectorIndexesOperator

Show vector indexes operator: SHOW VECTOR INDEX[ES]

SingleRowOperator

Single-row operator: emits exactly one empty record. Used as the input for standalone RETURN/UNWIND queries (no MATCH/CREATE). Analogous to Oracle’s DUAL table or PostgreSQL’s implicit single row.

SkipOperator

Skip operator: SKIP n

SortOperator

Sort operator: ORDER BY n.age ASC

UnwindOperator

UNWIND operator - expands a list expression into individual rows

VarLengthExpandOperator

Variable-length path expand operator: MATCH (a)-[*1..5]->(b) Uses BFS to find all paths within the specified hop range.

VectorSearchOperator

Vector search operator: supports both CALL db.index.vector.queryNodes(…) and SEARCH clause within MATCH. Returns nodes ordered by similarity (highest first) with optional SCORE AS alias.

WithBarrierOperator

WITH projection barrier operator.

Enums

AggregateType

Aggregation type

Traits

PhysicalOperator

Physical operator trait - all operators implement this

Functions

set_query_deadline

Set the query deadline for the current thread (called by QueryExecutor)

Type Aliases

OperatorBox

Type alias for boxed operators