Label limits for the query planner?

Neo4j Graph Platform Cypher
,
1

I have a Cypher query:

MATCH (x:lblA:lblB:lblC:lblD)-[:relA]->(:lblE)
WITH x as x
MATCH (x)-[:relB]->(y:lblY),
(y)-[:relC]->(z:lblZ)
RETURN count(DISTINCT z)

Where labelA represents approximately 250k nodes, and labels B, C and D are essentially subsets of those 250k.

The above query returns a count of 33, however, it takes about 20 minutes to run.
If I remove any one of labels lblA, lblB, lblC or lblD, the query returns in 200ms.

According to a profile of the queries, the difference between having 3 labels and 4 labels on x is that with 3 labels, Neo seems to apply those first and then do joins to y and z but with 4 labels, Neo seems to Cartesian join all my data, then apply the labels, resulting in about 2 billion DB hits.

Any ideas why Neo is doing this? Is there a way to stop it doing the Cartesian join if I have 4 labels on x?

Thanks,

Stephen

2

You probably want to change that to exists, so it doesn't blow up your cardinality?
If you can leave off lblE that helps too

WHERE exists { (x)-[:relA]->(:lblE) }
RETURN count(DISTINCT z)```