如何在Neo4j中检索各图组件对应的NodeA节点？

我在Neo4j中有一个包含两个组件的图结构，目标是检索每个组件中的NodeA类型节点，预期结果为[4, 3, 2, 1]和[6, 5]。尝试使用Neo4j的gcc和wcc功能未成功，自行编写的Cypher查询也未得到预期结果。

可通过以下Cypher语句模拟该图结构：

CREATE (a:NodeA {name: "1"});
CREATE (a:NodeA {name: "2"});
CREATE (a:NodeA {name: "3"});
CREATE (a:NodeA {name: "4"});
CREATE (a:NodeA {name: "5"});
CREATE (a:NodeA {name: "6"});
CREATE (b:NodeB {anotherName: "11"});
CREATE (b:NodeB {anotherName: "12"});
CREATE (b:NodeB {anotherName: "13"});
CREATE (b:NodeB {anotherName: "14"});
MATCH (a:NodeA {name: "1"})
MATCH(b:NodeB {anotherName: "11"})
CREATE (a) - [:CONTAINS] -> (b);
MATCH (a:NodeA {name: "2"})
MATCH(b:NodeB {anotherName: "11"})
CREATE (a) - [:CONTAINS] -> (b);
MATCH (a:NodeA {name: "3"})
MATCH(b:NodeB {anotherName: "11"})
CREATE (a) - [:CONTAINS] -> (b);
MATCH (a:NodeA {name: "3"})
MATCH(b:NodeB {anotherName: "12"})
CREATE (a) - [:CONTAINS] -> (b);
MATCH (a:NodeA {name: "4"})
MATCH(b:NodeB {anotherName: "12"})
CREATE (a) - [:CONTAINS] -> (b);
MATCH (a:NodeA {name: "5"})
MATCH(b:NodeB {anotherName: "13"})
CREATE (a) - [:CONTAINS] -> (b);
MATCH (a:NodeA {name: "5"})
MATCH(b:NodeB {anotherName: "14"})
CREATE (a) - [:CONTAINS] -> (b);
MATCH (a:NodeA {name: "6"})
MATCH(b:NodeB {anotherName: "14"})
CREATE (a) - [:CONTAINS] -> (b);

解决方案 #

基于GDS的WCC查询方案 #

默认WCC未生效的核心原因是未将有向关系转为无向处理（连通性依赖NodeB中转，需要让A和B的连接被视为双向），且未筛选组件内的NodeA节点。以下查询将整个图的关系设为无向，计算连通组件后提取NodeA并按名称降序排列：

CALL gds.wcc.stream({
  nodeProjection: ['NodeA', 'NodeB'],
  relationshipProjection: {
    CONTAINS: {
      type: 'CONTAINS',
      orientation: 'UNDIRECTED'
    }
  }
})
YIELD nodeId, componentId
MATCH (n) WHERE id(n) = nodeId
WITH componentId, collect(CASE WHEN n:NodeA THEN n.name ELSE NULL END) AS tempNames
WITH componentId, [name IN tempNames WHERE name IS NOT NULL ORDER BY name DESC] AS nodeANames
WHERE size(nodeANames) > 0
RETURN nodeANames

纯Cypher遍历方案（无需GDS库） #

如果无法使用图数据科学库，可通过路径遍历标记并提取连通组件内的NodeA：

// 初始化处理标记
MATCH (a:NodeA) SET a.processed = false;

// 遍历每个未处理的NodeA，获取其所在组件的所有NodeA
MATCH (start:NodeA) WHERE NOT start.processed
MATCH path = (start)-[:CONTAINS*..2]-(:NodeA)
WITH start, collect(DISTINCT [n IN nodes(path) WHERE n:NodeA | n.name]) AS componentNames
UNWIND componentNames AS names
WITH start, collect(DISTINCT names) AS uniqueNames
// 标记组件内节点为已处理，避免重复计算
MATCH (a:NodeA) WHERE a.name IN uniqueNames[0] SET a.processed = true
RETURN uniqueNames[0] ORDER BY uniqueNames[0] DESC;

// 清理临时标记
MATCH (a:NodeA) REMOVE a.processed;

以上两种方案均可得到预期的[4, 3, 2, 1]和[6, 5]结果。

内容的提问来源于stack exchange，提问作者Erdem Tuna