Traverse from your application data

Download the complete SQL script, including dataset setup for a fresh database.

Find how much of the network is reachable from a person, then return their neighbours as ordinary application rows. Run the setup once, then execute this script in the same database.

Traverse and join

-- Node ids are the original people.id values. Look up a start dynamically.
SELECT depth, count(*) AS people_reached
FROM warren.bfs('example_social', ARRAY(
    SELECT id FROM warren_examples.people WHERE name = 'person-1'
), 10)
GROUP BY depth ORDER BY depth;

-- Return business rows, including a predecessor for each shortest route.
SELECT p.id, p.name, b.depth, previous.name AS reached_via
FROM warren.bfs('example_social', ARRAY(
    SELECT id FROM warren_examples.people WHERE name = 'person-1'
), 3) b
JOIN warren_examples.people p ON p.id = b.node_id
LEFT JOIN warren_examples.people previous ON previous.id = b.via
ORDER BY b.depth, p.id;

SELECT warren.shortest_path('example_social', 1::bigint, 9999::bigint) AS path;

The start array comes from a name lookup. In an application, use a bound parameter for the name, email, or business identifier. The returned ids already belong to your table; there is no internal-id translation step.

The first query visits each person once at their shortest distance from the start. For person-1, its depth distribution is:

Depth	People
0	1
1	5
2	25
3	124
4	599
5	2,425
6	4,912
7	1,825
8	80
9	4

All 10,000 people are reached within nine hops. The second query returns names and each node’s predecessor, limited to three hops. The last query returns a shortest path as an array of original ids.

Change the starting set

The lookup can return several ids. BFS then measures the minimum distance from any start, with all starts at depth zero. A lookup returning no rows produces an empty start array and no traversal results.

Edges are followed outward. Use direction := 'in' to traverse against them or 'both' to ignore direction. A depth limit bounds hops, not the number of rows; max_results can additionally cap the number visited. See traversal for filters and truncation semantics.