如何在Cypher的WHERE子句中将关系路径拆分为两个关系？

拆分WHERE子句后的Cypher查询 #

要将原WHERE子句中的路径条件拆分为带共享临时变量的形式，可利用EXISTS子查询实现，确保逻辑与原查询完全一致。拆分后的完整查询如下：

MATCH 
(tom:`Person` {name: 'Tom Hanks'})-[:`ACTED_IN`]->(m),
(m)<-[:`ACTED_IN`]-(coActors),
(coActors)-[:`ACTED_IN`]->(m2),
(m2)<-[:`ACTED_IN`]-(cocoActors)
WHERE 
NOT EXISTS {
    MATCH (tom)-[:`ACTED_IN`]->(m3),
          (m3)<-[:`ACTED_IN`]-(cocoActors)
} AND tom <> cocoActors
RETURN cocoActors.name;

这里将原路径(tom)-[:ACTED_IN]->()<-[:ACTED_IN]-(cocoActors)拆分为两个独立的三元组，通过共享变量m3关联，并用EXISTS检查该路径是否存在后取反。

Cypher Java DSL实现示例 #

使用Neo4j官方Cypher Java DSL实现上述拆分后的查询，代码如下：

import org.neo4j.cypherdsl.core.Cypher;
import org.neo4j.cypherdsl.core.Node;
import org.neo4j.cypherdsl.core.RelationshipPattern;

public class CoActorQueryExample {
    public static void main(String[] args) {
        // 定义所有节点
        Node tom = Cypher.node("Person").withProperties("name", Cypher.literalOf("Tom Hanks")).named("tom");
        Node m = Cypher.node().named("m");
        Node coActors = Cypher.node().named("coActors");
        Node m2 = Cypher.node().named("m2");
        Node cocoActors = Cypher.node().named("cocoActors");
        Node m3 = Cypher.node().named("m3");

        // 定义MATCH中的关系三元组
        RelationshipPattern tomToM = tom.relationshipTo(m, "ACTED_IN");
        RelationshipPattern coActorsToM = coActors.relationshipTo(m, "ACTED_IN").inverse();
        RelationshipPattern coActorsToM2 = coActors.relationshipTo(m2, "ACTED_IN");
        RelationshipPattern cocoActorsToM2 = cocoActors.relationshipTo(m2, "ACTED_IN").inverse();

        // 构建WHERE子句中的子查询
        var existsSubquery = Cypher.match(
                tom.relationshipTo(m3, "ACTED_IN"),
                cocoActors.relationshipTo(m3, "ACTED_IN").inverse()
        ).build();

        // 组装完整查询
        var cypherQuery = Cypher.match(tomToM, coActorsToM, coActorsToM2, cocoActorsToM2)
                .where(Cypher.not(Cypher.exists(existsSubquery)))
                .and(tom.isNotEqualTo(cocoActors))
                .returning(cocoActors.property("name"))
                .build();

        // 输出生成的Cypher语句
        System.out.println(cypherQuery.getCypher());
    }
}

关于Cypher长路径写法的疑问解答 #

• 长路径写法是可选的：Cypher支持合并式长路径（如(a)-[:REL]->(b)<-[:REL]-(c)）和拆分式三元组两种写法，二者可互相转换，无强制要求。
• 仅用带变量的三元组足够表达所有逻辑：任何复杂路径查询都可拆分为多个独立的节点-关系-节点三元组，通过共享变量关联各部分。拆分写法在分步表达复杂逻辑、或与SPARQL写法对齐时，会更清晰易读。

内容的提问来源于stack exchange，提问作者Takis