Neo4j Cypher负载CSV在唯一约束下的失效

Question

我在使用Cypherload csv命令将大量数据导入Neo4j实例时遇到了问题。我试图加载大约253K个用户记录，每个记录都有一个唯一的user_id。我的第一步是在tje标签上添加一个唯一的约束，以确保用户只运行一次

CREATE CONSTRAINT ON (b:User) ASSERT b.user_id IS UNIQUE;

然后，我尝试用定期提交运行加载CSV来拉入这些数据。

此查询失败，因此在设置

USING PERIODIC COMMIT 1000
load csv with headers from "file:///home/data/uk_users.csv" as line
match (t:Territory{territory:"uk"})
merge (p:User {user_id:toInt(line.user_id)})-[:REGISTERED_TO]->(t)
set p.created=toInt(line.created), p.completed=toInt(line.completed);

USING PERIODIC COMMIT 1000
load csv with headers from "file:///home/data/uk_buddies.csv" as line
match (t:Territory{territory:"uk"})
merge (p:User {user_id:toInt(line.user_id), created:toInt(line.created), completed:toInt(line.completed)})-[:REGISTERED_TO]->(t);

LoadCsvStatusWrapCypherException: Node 9752 already exists with label Person and property "hpcm_uk_buddy_id"=[2446] (Failure when processing URL 'file:/home/data/uk_buddies.csv' on line 253316 (which is the last row in the file). Possibly the last row committed during import is line 253299. Note that this information might not be accurate.)

neo4j-sh (?)$ match (n) return count(n);
+----------+
| count(n) |
+----------+
| 5446     |
+----------+
1 row
768 ms

如果这个ID在CSV文件中只有大约5000行，我可以理解节点的数量是不正确的。但是有什么技术或命令可以用来成功地导入吗？

Answer 1

我认为，merge是一个常见错误的牺牲品。相对于cypher查询，严肃地说，这就像我关于cypher常见问题的10大常见问题解答一样。你这么做了：

USING PERIODIC COMMIT 1000
load csv with headers from "file:///home/data/uk_buddies.csv" as line
match (t:Territory{territory:"uk"})
merge (p:User {user_id:toInt(line.user_id), created:toInt(line.created), completed:toInt(line.completed)})-[:REGISTERED_TO]->(t);

合并的工作方式是，上一次合并与整个关系匹配，而不仅仅是与用户节点匹配。因此，您可能创建了不该创建的重复用户。在运行此合并时，即使具有这些确切属性的用户已经存在，但与T节点的关系也不存在，因此它试图创建一个具有这些属性的新用户节点，以连接到T，这不是您所希望的。

解决方案是分别合并用户，然后分别合并关系路径，如下所示：

USING PERIODIC COMMIT 1000
load csv with headers from "file:///home/data/uk_buddies.csv" as line
match (t:Territory{territory:"uk"})
merge (p:User {user_id:toInt(line.user_id), created:toInt(line.created), completed:toInt(line.completed)})
merge (p)-[:REGISTERED_TO]->(t);