Advanced Cypher

As queries grow, you need nested units of work: per-movie aggregations inside a larger report, or a write subquery that must not leak intermediate variables. The CALL { ... } subquery form executes an inner query and returns its projected rows to the outer scope. This is more structured than stacking many WITH clauses alone, especially when each outer row should drive an independent inner computation.

Scope and import rules are version-sensitive. In modern Cypher, outer variables are not freely visible inside the subquery unless you import them—commonly via a scope clause such as CALL (movie) { ... } or the documented import form for your Neo4j version. Always check the Cypher Manual’s “CALL subquery” page for the exact import syntax on your server. Treating imports as explicit documentation of data flow prevents subtle bugs when refactoring.

Subqueries can be read-only or contain writes. Combining writes with union-like fan-out requires care around transactional boundaries; prefer clear transaction functions in the driver (Lesson 10) for multi-step application workflows.

MATCH (m:Movie)
WHERE m.released >= 2000
CALL (m) {
  MATCH (m)<-[:ACTED_IN]-(p:Person)
  WITH p
  ORDER BY p.name
  RETURN count(p) AS castSize,
         collect(p.name)[0..3] AS sampleCast
}
RETURN m.title AS title, castSize, sampleCast
ORDER BY castSize DESC

If your Neo4j version does not yet support the parenthesized import form, use the documented alternative (historically an importing WITH at the start of the subquery). The educational point is the same: make outer-to-inner data flow explicit.

UNION concatenates compatible result columns from multiple query branches and removes duplicate rows. UNION ALL concatenates without deduplication and is cheaper when duplicates are impossible or acceptable. Column names and types must align across branches—alias carefully.

Existence predicates answer “is there any match?” without returning the matched data. EXISTS { MATCH pattern } (or the pattern form documented for your version) is true when the inner pattern succeeds under current bindings. Prefer EXISTS over collecting and checking list size when you only need a boolean, because the engine can stop at the first match.

Classic co-actor or “movies with reviews” filters become clearer as EXISTS predicates than as optional matches with null checks, especially when the pattern is multi-hop.

List comprehensions compactly project and filter lists: [name IN names WHERE name STARTS WITH 'T' | toUpper(name)]. Combined with collect, they shape API payloads without extra UNWIND stages. They do not replace graph pattern matching; they refine list values already in scope.

Neo4j temporal types—Date, DateTime, LocalDateTime, Time, Duration—support precise time modeling beyond integer years. Construct values with functions such as date('1999-03-31') or datetime() for “now.” The Movie sample often stores release years as integers; when you evolve a model, migrate deliberately and index the temporal properties you filter on. Temporal arithmetic uses durations, for example date() + duration('P7D').

APOC (Awesome Procedures on Cypher) is a widely used plugin providing utilities for data integration, refactoring, and advanced helpers. It is not part of core Cypher: clusters may disable it, versions may diverge, and procedure allow-lists / security configs restrict what is callable. Prefer core Cypher for portable application logic. When you need APOC, pin versions, document the dependency, and gate features when procedures are missing.