Cypher Query Language

Cypher is a declarative language, meaning you describe what data you want, not how to retrieve it. You define patterns using an intuitive ASCII-art syntax.

• Example: (p:Person)-[:LIVES_IN]->(c:City) finds all Person nodes connected to City nodes via a LIVES_IN relationship.
• The engine's query optimizer determines the most efficient execution plan, abstracting traversal logic from the developer.

Cypher is designed for the property graph model, where both nodes (entities) and relationships (connections) can have properties (key-value pairs).

• Nodes are represented in parentheses: (:Person {name: 'Alice', age: 30}).
• Relationships use brackets and arrows: -[:WORKS_AT {since: 2020}]->.
• This model allows for a rich, schema-flexible representation of connected data, closely mirroring real-world domains.

Cypher uses a structured set of clauses to compose queries, similar to SQL.

• MATCH: Locates patterns in the graph.
• WHERE: Filters results with predicates.
• RETURN: Specifies what data to output.
• CREATE / MERGE: Inserts new data or ensures it exists.
• SET / REMOVE: Updates or deletes properties.
• DELETE: Removes nodes and relationships.
• WITH: Pipes results between query parts, enabling complex multi-step operations.

A core capability is querying for paths—sequences of nodes and relationships. Cypher excels at variable-length traversals to discover connections of unknown depth.

• Use * for variable hops: (:Person)-[:KNOWS*1..5]->(:Person) finds chains of KNOWS relationships between 1 and 5 hops long.
• The shortestPath() function finds the optimal path between two nodes.
• This is essential for recommendation systems, fraud detection (finding rings), and network analysis.

Cypher supports powerful aggregation functions and collection operations for analytical workloads directly on the graph structure.

• Use COUNT(), SUM(), COLLECT() in a RETURN clause with GROUP BY.
• Built-in graph algorithms like PageRank and Louvain community detection can be invoked via Cypher procedures (e.g., CALL gds.pageRank.stream).
• This enables business intelligence and graph data science without exporting data to external systems.

Cypher queries are optimized for index-free adjacency, a native graph storage paradigm where nodes hold direct pointers to connected relationships. This enables constant-time traversals regardless of total graph size.

• Query planners use statistical metadata to select optimal start nodes and traversal directions.
• Composite indexes and full-text indexes accelerate property lookups.
• Query caching and parameterization ($param) improve throughput for repeated query patterns.

Cypher is used to power graph-based analytics, uncovering insights through relationship-centric queries that are cumbersome in SQL.

• Fraud Detection: Find complex rings of connected entities (e.g., (p1:Person)-[:TRANSFERRED_TO*3..5]->(p2:Person)).
• Recommendation Engines: Query for paths between users and products ("People who bought X also bought Y").
• Supply Chain Analysis: Traverse supplier networks to identify single points of failure or optimize routes.
• Social Network Analysis: Calculate centrality metrics (PageRank, betweenness) using built-in or user-defined procedures.

Within Enterprise Knowledge Graphs, Cypher is used to query interconnected entities and their properties, serving as a foundational layer for deterministic reasoning.

• Semantic Search: Traverse ontology-based relationships to answer complex questions (e.g., "Which projects use deprecated libraries?").
• Data Lineage & Provenance: Trace the origin and transformation of data assets through explicit relationship chains.
• Graph-Based RAG: Retrieve precise subgraphs as context for LLMs, providing factual grounding and eliminating hallucinations. Cypher queries fetch connected facts, not just isolated chunks.

Cypher is executed from within application code via official drivers, making it the interface between business logic and the graph data layer.

• Official Drivers: Use Neo4j drivers for Java, JavaScript, Python, .NET, and Go to send parameterized Cypher queries.
• Parameterized Queries: Prevent injection attacks and enable query reuse (e.g., MATCH (p:Person {name: $name})).
• Transaction Management: Group multiple Cypher operations within ACID transactions to ensure data integrity.
• Microservice Integration: A dedicated graph microservice often exposes endpoints that execute specific Cypher queries to serve graph data to other services.

Data scientists use Cypher to extract connected datasets for feature engineering and to operationalize graph-native ML models.

• Feature Extraction: Query subgraphs around entities to create predictive features (e.g., a customer's network influence).
• Graph Algorithm Execution: Call Graph Algorithm Library functions (like Louvain community detection) via Cypher procedures (CALL gds.louvain.stream).
• Graph Embedding Generation: Use Cypher to project a graph and feed it into a Graph Neural Network (GNN) Service.
• Link Prediction: Write Cypher queries to find candidate pairs for relationship prediction models.