Knowledge Base

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 on match results 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)
                   RETURN p, m LIMIT 3

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

Here are some solutions to apply a limit to match results per-row

Use LIMIT within a subquery in 4.1+

Neo4j 4.1 introduced correlated subqueries, letting us perform a subquery using variables present mid-query.

Since subqueries execute per row, we can perform the MATCH and apply the LIMIT within the subquery, giving us the easiest means of limiting match results per row.

This requires the use of WITH as the first clause within the subquery CALL block, for the purpose of importing variables to the subquery.

MATCH (:Movie{title:'The Matrix'})<-[:ACTED_IN]-(p:Person)
CALL {
    WITH p
    MATCH (p)-[:ACTED_IN]->(m)
    RETURN m
    LIMIT 3
}

RETURN p, m

This will correctly return all actors in the matrix and up to 3 of the movies they have acted in.

If we wanted each actor only once, with the collection of up to 3 movies per actor, we could collect those movies within the subquery after the limit:

MATCH (:Movie{title:'The Matrix'})<-[:ACTED_IN]-(p:Person)
CALL {
    WITH p
    MATCH (p)-[:ACTED_IN]->(m)
    WITH m
    LIMIT 3
    RETURN collect(m) as movies
}

RETURN p, movies

Be aware that this WITH import has some special restrictions that do not normally apply to WITH usage:

  1. You may only include variables from the outer query and no others.

  2. You cannot perform calculations, aggregations, or introduction of new variables in the initial WITH.

  3. You cannot alias any variables within this initial WITH.

  4. You cannot follow the initial WITH with a WHERE clause for filtering.

If you try any of these, you will be met with some kind of error, such as:

Importing WITH should consist only of simple references to outside variables. Aliasing or expressions are not supported.

or more cryptically, if you try to use a WHERE clause after the initial WITH

Variable `x` not defined

(where the variable is the first one present in the WITH clause)

You can get around all of these restrictions by simply introducing an additional WITH clause after the importing WITH, like so:

MATCH (:Movie{title:'The Matrix'})<-[:ACTED_IN]-(p:Person)
CALL {
    WITH p
    WITH p as actor
    MATCH (actor)-[:ACTED_IN]->(m)
    RETURN m
    LIMIT 3
}

RETURN p, m

This shows how we can alias (or filter, if we wanted) the imported variables, but not on the initial import WITH itself.

For 4.0.x and earlier

For earlier versions, native correlated subqueries are not available, so other workarounds must be used.

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 apoc.cypher.run() 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 apoc.cypher.run() 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)
CALL apoc.cypher.run('
 WITH {p} AS p
 MATCH (p)-[:ACTED_IN]->(m)
 RETURN m LIMIT 3',
 {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 3.1.3.6 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