The following query can't run on a dataset with ~2M nodes. What should i do to make it run faster?
MATCH (cc:ConComp)-[r1:IN_CONCOMP]-(p1:Person)-[r2:SAME_CLUSTER]-(p2:Person)
WHERE cc.cluster_type = "household"
MERGE (cluster:Cluster {CLUSTER_TMP_ID:cc.CONCOMP_ID '|' r2.root_id, cluster_type:cc.cluster_type })
MERGE (cluster)-[r3:IN_CLUSTER]-(p1)
CodePudding user response:
A number of suggestions:
- adding directions to your relationships will decrease the number of paths in the MATCH
- make sure that you have indexes on all properties that you MERGE on
- in the second MERGE , also add direction.
CodePudding user response:
I finally found a solution by using the following query (and by indexing cc.cluster_type and cc.CONCOMP_ID):
CALL apoc.periodic.iterate('MATCH (cc:ConComp)<-[r1:IN_CONCOMP]-(p1:Person)-[r2:SAME_CLUSTER]-(p2:Person) WHERE cc.cluster_type = "household" WITH DISTINCT cc.CONCOMP_ID "|" r2.root_id as id_name, cc.cluster_type as cluster_type_name, p1 RETURN id_name, cluster_type_name, p1', '
MERGE (cluster:Cluster {CLUSTER_TMP_ID: id_name, cluster_type: cluster_type_name})
MERGE (cluster)-[r3:IN_CLUSTER]->(p1)', {batchSize:10000, parallel:false})
I precise that I had previously ran my initial question query with apoc.periodic.iterate without success.