Limiting MATCH results per row

Since LIMIT applies to the total number of rows of the query, it can't be used in cases when matching from multiple nodes where the limit must be per row.

Take an example case using the Movies database.

If you needed a query to get all the actors from The Matrix, and for each actor, get 3 movies that actor has acted in, a first (incorrect) attempt might look like this:

MATCH (:Movie{title:'The Matrix'})<-[:ACTED_IN]-(p:Person)
MATCH (p)-[:ACTED_IN]->(m)

The above query won't return the desired results. The LIMIT will instead make the query return only 3 rows total.

Some solutions to apply a limit to match results per-row

Take the interested slice of a collection

One common solution is to collect() and take the interested slice:

MATCH (:Movie{title:'The Matrix'})<-[:ACTED_IN]-(p:Person)
MATCH (p)-[:ACTED_IN]->(m)
RETURN p, collect(m)[..3] AS movies

In Neo4j 3.1.x and newer you can use pattern comprehension as a shorthand approach:

MATCH (:Movie{title:'The Matrix'})<-[:ACTED_IN]-(p:Person)
RETURN p, [(p)-[:ACTED_IN]->(m) | m][..3] as movies

If only one element in the collection is needed, the head() function can be used to get the first element from the pattern comprehension:

MATCH (:Movie{title:'The Matrix'})<-[:ACTED_IN]-(p:Person)
RETURN p, head([(p)-[:ACTED_IN]->(m) | m]) as movie

While this works when there are few relationships per node, it may become infeasible on supernodes with larger numbers of relationships, as it must expand all :ACTED_IN relationships before collecting.

Use to execute a limited subquery

Neo4j doesn't currently offer native subquery support aside from pattern comprehension, but even those don't support LIMIT.

However, in Neo4j 3.0.x and newer, using APOC Procedures, you can use to execute a subquery with a LIMIT, which performs the way we want since it executes per-row.

MATCH (:Movie{title:'The Matrix'})<-[:ACTED_IN]-(p:Person)
 WITH {p} AS p
 MATCH (p)-[:ACTED_IN]->(m)
 {p:p}) YIELD value
RETURN p, value.m AS movie

This approach is efficient since by using LIMIT we don't have to pay the cost of expanding all :ACTED_IN relationships, we only need to expand 3 per row.

Use APOC path expander using end node or termination filters and limit param

With Neo4j 3.1.3 and higher, and APOC Procedures and higher, you can use use new path expander features to limit expansion to certain nodes.

The limit param is only usable with path expander procedures that take a config map, and only when using the end node (>) or termination label filters (/):

  • apoc.path.expandConfig()
  • apoc.path.subgraphNodes()
  • apoc.path.subgraphAll()
  • apoc.path.spanningTree()

Using this approach, the query becomes:

MATCH (:Movie{title:'The Matrix'})<-[:ACTED_IN]-(p:Person)
CALL apoc.path.subgraphNodes(p, {relationshipFilter:'ACTED_IN>', labelFilter:'/Movie', limit:3}) YIELD node
RETURN p, node as movie