CODEWITHSUNDEEP
neo4jcypher
Tully
Tully

Reputation: 3

neo4j: file import, relationship match, create new relationship

I know how this would be accomplished in SQL but having difficulty wrapping my brain around how to do this in cypher..

Basically working on a master data setup where a user has a master_id (node) and need to use an existing relationship property to determine the master_id in order to create a new relationship between the master_id node and a location node.

Currently have master users created as nodes that contain a master_id property. A relationship is created between the master user and a brand, and the relationship has a property of brand_user_id.

I now have another file I need to import that contains data at the brand_user level, but need to create the relationship between the master_id and a location node. In order to do this because the file does not contain the master_id property, I am attempting to use the new file to lookup master_id's based on the existing relationship with the brand, then use that master_id to create the new relationship with the location.

Have this relationship:

(m:Master{master_id:12345})-[:IS_BRAND_USER{brand_user_id:9876}]->(b:Brand{name:"Acme"})

Have this file:

brand_user_id,location_id
9876,6

Need this relationship:

(m:Master{master_id:12345})-[:HAS_LOCATION]->(l:Location{id:6})

My approach:

LOAD CSV WITH HEADERS FROM "file:///brand_user_ids.csv" as buid
MATCH (m:Master)-[r:IS_BRAND_USER{brand_user_id:buid.id}]->(b:Brand)
WITH m, buid.location_id AS location_id
MATCH (l:Location {id: location_id})
CREATE (m)-[:HAS_LOCATION {source: 'abcdef'}]->(l)

Seems to run for an extremely long time and not seeing any real progress after an hour so I'm wondering if this is fundamentally the correct approach or not, or if I have inadvertently created some horrific cross join equivalent.

Upvotes: 0

Views: 140

Answers (1)

Tom Geudens
Tom Geudens

Reputation: 2656

The problem is that you are trying to enter a graph on a relationship. And that always requires lots of "scanning the graph".

Now, I'm not a specialist in your domain, but you might be missing a type of nodes here ... BrandUser. And there could be several reasons for that :

  1. Based on your data a Master can have many BrandUser id's. Potentially even more than one per Brand ? Do you have other properties that make sense on the BrandUser level ?
  2. That Location data is strange. Wouldn't you agree it's actually the BrandUser that has a location and that a Master may have many locations ?

The most important reason is however ... if you're going to enter the graph on the brand_user_id all the time (and judging from the location example that may be the case) ... you've got the reason to turn it into a node right there.

So ... it's a modeling issue really.

Hope this helps.

Regards, Tom

Upvotes: 1

Related Questions