opencypher.org opencypher.org | opencypher@googlegroups.com - PowerPoint PPT Presentation

opencypher.org opencypher.org | opencypher@googlegroups.com

• • • • • • opencypher.org opencypher.org | opencypher@googlegroups.com

START john=node:Person(name="John") MATCH john-[:KNOWS]-friend-[:KNOWS]-foaf RETURN foaf opencypher.org opencypher.org | opencypher@googlegroups.com

START john=node:Person(name="John") MATCH (john)-[:KNOWS]-(friend)-[:KNOWS]-(foaf) WHERE NOT (john)-[:KNOWS]-(foaf) AND NOT (john)-[:RECOMMENDATION]->(foaf) CREATE (john)-[:RECOMMENDATION]->(foaf) RETURN foaf opencypher.org opencypher.org | opencypher@googlegroups.com

MATCH (john:Person{name:"John"}), (john)-[:KNOWS]-(friend)-[:KNOWS]-(foaf) WHERE NOT (john)-[:KNOWS]-(foaf) MERGE (john)-[:RECOMMENDATION]->(foaf) opencypher.org opencypher.org | opencypher@googlegroups.com

Query => AST => AST’ AST’ => Query Graph => Logical Plan L Plan => Instruction => CodeGen => JVM Class opencypher.org opencypher.org | opencypher@googlegroups.com

(query:String)-[:PARSED_TO]->(:AST {pos: $pos}) opencypher.org opencypher.org | opencypher@googlegroups.com

SemanticState => SemanticState opencypher.org opencypher.org | opencypher@googlegroups.com

The AST is optimised and canonicalized Custom functional rewriting code opencypher.org opencypher.org | opencypher@googlegroups.com

opencypher.org opencypher.org | opencypher@googlegroups.com

Output: Logical Plans similar to relational query plans Expand/VarExpand are the main differences to normal relational algebra opencypher.org opencypher.org | opencypher@googlegroups.com

Similar to the System R - Dynamic Programming Quick Reminder, the Selinger Algorithm: 1. Enumerate all access paths for single relation 2. Consider all ways to join two relations 3. Consider all ways to join three relations, etc [Access path selection in a relational database management system, Selinger et al] opencypher.org opencypher.org | opencypher@googlegroups.com

Neo4j Cost Planner 1. Enumerate all access paths for pattern nodes 2. Consider all ways to expand pattern relationships 3. Consider all ways to solve two pattern relationships, either by expand or join, etc for three, four patter relationships [Iterative dynamic programming: a new class of query optimization algorithms, Kossman, Stocker] opencypher.org opencypher.org | opencypher@googlegroups.com

Pattern: (a)-[r1]->(b)<-[r2]-(c) e -[r2]-> (c) ⋈ Solutions: e (b)<-[r2]- e -[r1]-> (b) e -[r1]-> (b) |a| |a| |c| opencypher.org opencypher.org | opencypher@googlegroups.com

opencypher.org opencypher.org | opencypher@googlegroups.com

opencypher.org opencypher.org | opencypher@googlegroups.com

IDPSolver SingleComponentPlanner CardinalityCostModel opencypher.org opencypher.org | opencypher@googlegroups.com

• • accept(...) L Plan => Instruction => CodeGen => JVM Class [Efficiently Compiling Efficient Query Plans for Modern Hardware, Thomas Neumann] opencypher.org opencypher.org | opencypher@googlegroups.com

Instruction Cypher While(Variable("v1") -> ScanAllNodes()){ Selection( MATCH (n) If("v1=100", WHERE n.x = 100 AcceptVisitor( Map("n"-> NodeProjection(Variable("v1")))))) RETURN n } ) opencypher.org opencypher.org | opencypher@googlegroups.com

AllNodesScan Filter Project opencypher.org opencypher.org | opencypher@googlegroups.com

AllNodesScan Filter Project opencypher.org opencypher.org | opencypher@googlegroups.com

opencypher.org opencypher.org | opencypher@googlegroups.com