Early Path Traversal Pruning
summary
Variable-length traversals from a bound source to unbound targets now prune dead branches during traversal when a running accumulator fails a condition, instead of traversing everything and post-filtering. The new all reduce function embeds the check into the walk (requires Cypher 25 and ships in Neo4j 2025.8.0), cutting DB hits in examples from >100 to 42 and from 91 to 26 for sum-of-cost and increasing-cost conditions. Useful for UI exploratory queries that constrain paths by accumulated values.
AI-written summary · about
related talks
27mSpeed and Precision: Mastering Graph Traversal With Quantified Path Patterns
21mBonsai: Compiling Queries to Pruned Tree Traversals
31mIndexing the Extended Dyck-CFL Reachability for Context-Sensitive Program Analysis
38mNeo4j for Solving Graph-Based Problems
28mMy Cypher Query Takes Too Long. What Can I Do ?
53mBreaking Away from the Empire: Avoiding the Evil Clutches of If-Then Statements
41mOptimizing Relational Programs I
30mPerformant GraphRAG Improving Neo4j to Drive Autonomous AI ...