Neo4j 2.1：传递节点ID与UNWIND

在发布Neo4j 2.1时，我们将具有UNWIND子句，该子句使处理事物集合更加容易。

在我的有关创建邻接矩阵的博客文章中，我们想要显示按字母顺序排列的前5个见面组中有多少人，然后检查其他各组中有多少人。

没有UNWIND子句，我们必须这样做：

MATCH (g:Group)
WITH g
ORDER BY g.name
LIMIT 5WITH COLLECT(id(g)) AS groupsMATCH (g1) WHERE id(g1) IN groups
MATCH (g2) WHERE id(g2) IN groupsOPTIONAL MATCH path = (g1)<-[:MEMBER_OF]-()-[:MEMBER_OF]->(g2)RETURN g1.name, g2.name, CASE WHEN path is null THEN 0 ELSE COUNT(path) END AS overlap

在这里，我们获得了前5个组，将其ID放入集合中，然后通过背对背具有节点ID查找的MATCH来创建组的笛卡尔积。

如果不是在“组”中传递节点ID，而是传递节点，然后在MATCH步骤中使用节点ID，那么我们将进行完整的节点扫描，随着存储的增长，它变得非常慢。

例如，此版本将非常慢：

MATCH (g:Group)
WITH g
ORDER BY g.name
LIMIT 5WITH COLLECT(g) AS groupsMATCH (g1) WHERE g1 IN groups
MATCH (g2) WHERE g2 IN groupsOPTIONAL MATCH path = (g1)<-[:MEMBER_OF]-()-[:MEMBER_OF]->(g2)RETURN g1.name, g2.name, CASE WHEN path is null THEN 0 ELSE COUNT(path) END AS overlap

这是原始查询的输出：

+-------------------------------------------------------------------------------------------------------------+
| g1.name                                         | g2.name                                         | overlap |
+-------------------------------------------------------------------------------------------------------------+
| "Big Data Developers in London"                 | "Big Data / Data Science / Data Analytics Jobs" | 17      |
| "Big Data Jobs in London"                       | "Big Data London"                               | 190     |
| "Big Data London"                               | "Big Data Developers in London"                 | 244     |
| "Cassandra London"                              | "Big Data / Data Science / Data Analytics Jobs" | 16      |
| "Big Data Jobs in London"                       | "Big Data Developers in London"                 | 52      |
| "Cassandra London"                              | "Cassandra London"                              | 0       |
| "Big Data London"                               | "Big Data / Data Science / Data Analytics Jobs" | 36      |
| "Big Data London"                               | "Cassandra London"                              | 422     |
| "Big Data Jobs in London"                       | "Big Data Jobs in London"                       | 0       |
| "Big Data / Data Science / Data Analytics Jobs" | "Big Data / Data Science / Data Analytics Jobs" | 0       |
| "Big Data Jobs in London"                       | "Cassandra London"                              | 74      |
| "Big Data Developers in London"                 | "Big Data London"                               | 244     |
| "Cassandra London"                              | "Big Data Jobs in London"                       | 74      |
| "Cassandra London"                              | "Big Data London"                               | 422     |
| "Big Data / Data Science / Data Analytics Jobs" | "Big Data London"                               | 36      |
| "Big Data Jobs in London"                       | "Big Data / Data Science / Data Analytics Jobs" | 20      |
| "Big Data Developers in London"                 | "Big Data Jobs in London"                       | 52      |
| "Cassandra London"                              | "Big Data Developers in London"                 | 69      |
| "Big Data / Data Science / Data Analytics Jobs" | "Big Data Jobs in London"                       | 20      |
| "Big Data Developers in London"                 | "Big Data Developers in London"                 | 0       |
| "Big Data Developers in London"                 | "Cassandra London"                              | 69      |
| "Big Data / Data Science / Data Analytics Jobs" | "Big Data Developers in London"                 | 17      |
| "Big Data London"                               | "Big Data Jobs in London"                       | 190     |
| "Big Data / Data Science / Data Analytics Jobs" | "Cassandra London"                              | 16      |
| "Big Data London"                               | "Big Data London"                               | 0       |
+-------------------------------------------------------------------------------------------------------------+
25 rows

如果使用UNWIND，则不再需要传递节点ID，而是可以将节点收集到一个集合中，然后将其分解为笛卡尔积：

MATCH (g:Group)
WITH g
ORDER BY g.name
LIMIT 5WITH COLLECT(g) AS groupsUNWIND groups AS g1
UNWIND groups AS g2OPTIONAL MATCH path = (g1)<-[:MEMBER_OF]-()-[:MEMBER_OF]->(g2)RETURN g1.name, g2.name, CASE WHEN path is null THEN 0 ELSE COUNT(path) END AS overlap

没有明显更少的代码，但是我认为使用UNWIND可以使查询的意图更加清晰。

我很期待看到一旦2.1是GA，UNWIND员工就会发挥创新的作用。