Read/Set conflict when trying to "extend" a relationship

Neo4j Graph Platform Cypher
,
1

Using neo4j 5.20.0 on Ubuntu, and just the cypher-shell

In my EXPLAIN/PROFILE on a LOAD CSV query, I am seeing an Eager operation with the details

read/set conflict for relationship type: PORTRAYS (Operator: 4 vs 11)

First off, the single google result for "read/set conflict neo4j" points to Operators in detail in which it only appears in the output of example commands, but never explained. I can deduce what it means: that a node or relationship is simultaneously trying to be read and written for different rows, and if done lazily then you could end up with trouble; thus the query eagerly loads all the incoming data. However, I have no idea how to interpret "Operator: 4 vs 11" or if it's even useful information.

I have resolved a number of these in my load queries by simply separating them into two queries, so each one executes a single MERGE instead of two MERGEs in the same query. However, in this final example there is no way for me to split them up. I am looking for a relationship A-->B, then for a relationship B-->C, then extending it transitively to A-->C with the same properties. The end state is two relationships, so apoc.refactor.to (which moves a relationship) is not what I'm looking for.

This is what I am using (adapted to the movie scenario):

LOAD CSV WITH HEADERS FROM 'file:///input.csv' AS row
WITH row
MATCH (p:Person {personID: row.personID})-[:AKA]->(a:Alias)
MATCH (c:Character {characterID: row.characterID})<-[r1:PORTRAYS]-(a)
MERGE (p)-[r2:PORTRAYS]->(c)
    SET r2 = properties(r1)

For example, you know that "Ludacris" (alias) portrays "Tej Parker" (character) in the Fast & Furious movies. You also know that Ludacris's real name is Christopher Bridges (person). Thus, (Christopher Bridges)-[:PORTRAYS]->(Tej Parker) with the exact same relationship properties as (Ludacris)-[:PORTRAYS]->(Tej Parker).

How can I remove this eager operation? Or in other words, how can I clone a relationship?

2

Are you having performance issues?

Why do you want to create the redundant relationship? You get this information from the original relationships. Having a duplicate, means that you need to maintain parity between the two relationships. What happens when a property of 'r1' changes? Does this necessitate updating 'r2'? What about if the original relationship 'r1' is no longer valid and it is delete? You will need to delete 'r2' to keep you data model accurate.

3

You're right, I am not having performance issues with the database itself, this was merely a thought to simplify queries as I'm hand-writing cypher all the time. The path is there, it's just extra typing and extra thinking to remember the relationship & node labels that lie between A and C. This difficulty is born out of having used cypher-shell exclusively over ssh for a while (instead of Neo4j browser or Bloom, which suggest labels nicely).

Parity would be handled by scheduled scripts which are idempotent. And since I planned on running these load scripts often, I was on a mission to remove every EAGER operation I saw. In the end, I could live without this relationship or with the EAGER operation, but you've given me something to think about. Thank you.