Neighbor Functions

The Neighborhood search procedures enable quick discovery of surrounding nodes based on a specific relationship type and number of hops.

Available Procedures

The table below describes the available procedures:

Qualified Name Type Release

apoc.neighbors.athop

- returns distinct nodes of the given relationships in the pattern at a distance, can use '>' or '<' for all outgoing or incoming relationships

Procedure

APOC Core

apoc.neighbors.byhop

- returns distinct nodes of the given relationships in the pattern at each distance, can use '>' or '<' for all outgoing or incoming relationships

Procedure

APOC Core

apoc.neighbors.tohop

- returns distinct nodes of the given relationships in the pattern up to a certain distance, can use '>' or '<' for all outgoing or incoming relationships

Procedure

APOC Core

Relationship Filters

The 2nd parameter in each of the neighborhood search procedures is a relationship filter. A relationship filter is a | separated list of relationship types, using the following syntax:

Syntax: [<]RELATIONSHIP_TYPE1[>]|[<]RELATIONSHIP_TYPE2[>]|…​

input type direction

LIKES>

LIKES

OUTGOING

<FOLLOWS

FOLLOWS

INCOMING

KNOWS

KNOWS

BOTH

>

any type

OUTGOING

<

any type

INCOMING

Relationship filters are white space sensitive, so check for trailing white spaces (and then remove them!) if you’re not seeing expected results.

Examples

The examples in this section are based on the following sample graph:

MERGE (mark:Person {name: "Mark"})
MERGE (praveena:Person {name: "Praveena"})
MERGE (joe:Person {name: "Joe"})
MERGE (lju:Person {name: "Lju"})
MERGE (michael:Person {name: "Michael"})
MERGE (emil:Person {name: "Emil"})
MERGE (ryan:Person {name: "Ryan"})

MERGE (ryan)-[:FOLLOWS]->(joe)
MERGE (joe)-[:FOLLOWS]->(mark)
MERGE (mark)-[:FOLLOWS]->(emil)
MERGE (michael)-[:KNOWS]-(emil)
MERGE (michael)-[:KNOWS]-(lju)
MERGE (michael)-[:KNOWS]-(praveena)
MERGE (emil)-[:FOLLOWS]->(joe)
MERGE (praveena)-[:FOLLOWS]->(joe)

The KNOWS relationship type is considered to be bidirectional, where if Michael knows Emil, we can imply that Emil knows Michael. When using the KNOWS relationship we will ignore the direction.

The FOLLOWS relationship has a direction, so we will specify a direction when we use it.

Find neighbors at specified hop count

The apoc.neighbors.athop procedures compute a node’s neighborhood at a specific hop count.

The following returns the people that Emil KNOWS at 1 hop
MATCH (p:Person {name: "Emil"})
CALL apoc.neighbors.athop(p, "KNOWS", 1)
YIELD node
RETURN node
Table 1. Results
node

(:Person {name: "Michael"})

Emil only has a direct KNOWS relationship to Michael, so Michael is the only node returned by this query.

The following returns the people that Emil KNOWS at 2 hops
MATCH (p:Person {name: "Emil"})
CALL apoc.neighbors.athop(p, "KNOWS", 2)
YIELD node
RETURN node
Table 2. Results
node

(:Person {name: "Praveena"})

(:Person {name: "Lju"})

Michael also KNOWS Praveena and Lju, and since Emil doesn’t KNOW either of those directly, he only KNOWS them at a hop distance of 2. If we aren’t interested in knowing which nodes are in our neighborhood, but just want a count of the number, we can do that as well.

The following returns the number of people that Emil KNOWS at 2 hops
MATCH (p:Person {name: "Emil"})
CALL apoc.neighbors.athop.count(p, "KNOWS", 2)
YIELD value
RETURN value
Table 3. Results
value

2

As expected we get a count of 2, those people being Praveena and Lju!

Find neighbors at specified hop counts

The apoc.neighbors.byhop procedures compute a node’s neighborhood at multiple hop counts.

The following returns the people that Emil KNOWS up to 2 hops
MATCH (p:Person {name: "Emil"})
CALL apoc.neighbors.byhop(p, "KNOWS", 2)
YIELD nodes
RETURN nodes
Table 4. Results
nodes

[(:Person {name: "Michael"})]

[(:Person {name: "Praveena"}), (:Person {name: "Lju"})]

From these results we can see that at level 1 Emil KNOWS Michael, and at level 2 Emil KNOWS Lju and Praveena. The following graph patterns describe how Emil knows the different people:

Level 1

  • (emil)-[:KNOWS]-(michael)

Level 2

  • (emil)-[:KNOWS]-(michael)-[:KNOWS]-(lju)

  • (emil)-[:KNOWS]-(michael)-[:KNOWS]-(praveena)

We can also use multiple relationship types when searching the neighborhood.

Let’s say that as well as finding the people that Emil knows, we also want to find the people that follow him. We can specify a direction to the relationship types, by using < to indicate an incoming relationship, or > to indicate an outgoing relationship. So to find people that follow Emil, we’d use <FOLLOWS.

The following returns the people that Emil KNOWS and those that have FOLLOWS relationships to him, at up to 3 hops
MATCH (p:Person {name: "Emil"})
CALL apoc.neighbors.byhop(p, "KNOWS|<FOLLOWS", 3)
YIELD nodes
RETURN nodes
Table 5. Results
nodes

[(:Person {name: "Mark"}), (:Person {name: "Michael"})]

[(:Person {name: "Praveena"}), (:Person {name: "Joe"}), (:Person {name: "Lju"})]

[(:Person {name: "Ryan"})]

We’ve got some more results this time. Mark is in Emil’s level 1 neighborhood, Joe is in his level 2 neighborhood, and Ryan is in his level 3 neighborhood.

The following graph patterns describe how Emil knows the different people:

Level 1

  • (emil)-[:KNOWS]-(michael)

  • (mark)-[:FOLLOWS]→(emil)

Level 2

  • (emil)-[:KNOWS]-(michael)-[:KNOWS]-(lju)

  • (emil)-[:KNOWS]-(michael)-[:KNOWS]-(praveena)

  • (joe)-[:FOLLOWS]→(mark)-[:FOLLOWS]→(emil)

Level 3

  • (ryan)-[:FOLLOWS]→(joe)-[:FOLLOWS]→(mark)-[:FOLLOWS]→(emil)

And, as with the apoc.neighbors.athop procedure, we can also return just the neighborhood size at each hop.

The following returns the number of people that Emil KNOWS and the number that have FOLLOWS relationships to him, at up to 3 hops
MATCH (p:Person {name: "Emil"})
CALL apoc.neighbors.byhop.count(p, "KNOWS|<FOLLOWS", 3)
YIELD value
RETURN value
Table 6. Results
value

[2, 3, 1]

And as expected we have a count of 2 at level 1, 3 at level 2, and 1 at level 3.

We could even turn that list of numbers into a map with the key being the number of hops and the value the neighborhood size. The following query shows how to do this using the apoc.map.fromLists function:

MATCH (p:Person {name: "Emil"})
CALL apoc.neighbors.byhop.count(p, "KNOWS|<FOLLOWS", 3)
YIELD value
RETURN apoc.map.fromLists(
         [value in range(1, size(value)) | toString(value)],
         value) AS value
Table 7. Results
value

{1: 2, 2: 3, 3: 1}

Find neighbors up to specified hop count

The apoc.neighbors.tohop procedures compute a node’s neighborhood up to a specified hop count.

The following returns the people that Praveena FOLLOWS up to 1 hop
MATCH (p:Person {name: "Praveena"})
CALL apoc.neighbors.tohop(p, "FOLLOWS>", 1)
YIELD node
RETURN node
Table 8. Results
nodes

(:Person {name: "Joe"})

The only person that Praveena follows is Joe, so that’s the only node returned. What about if we include people at up to 2 hops?

The following returns the people that Praveena FOLLOWS up to 2 hops
MATCH (p:Person {name: "Praveena"})
CALL apoc.neighbors.tohop(p, "FOLLOWS>", 2)
YIELD node
RETURN node
Table 9. Results
nodes

(:Person {name: "Mark"})

(:Person {name: "Joe"})

Now Mark is returned as well. The following graph patterns describe how Emil knows the different people:

  • (praveena)-[:FOLLOWS]-(joe)

  • (praveena)-[:FOLLOWS]-(joe)-[:FOLLOWS]→(mark)

And if we just want a count of the number of people, we can use the count variant.

The following returns the number of people that Praveena FOLLOWS up to 2 hops
MATCH (p:Person {name: "Praveena"})
CALL apoc.neighbors.tohop.count(p, "FOLLOWS>", 2)
YIELD value
RETURN value
Table 10. Results
value

2