Neo4j - Graph Data Science library - How to cypher-query against a named graph in Graph Catalog?

Question

Creating a named graph from the main Neo4j Graph is documented . Beside, one also knows how to list, drop, check if a named graph already exists, eg CALL gds.graph.exists('my-store-graph') YIELD exists;

However, I wonder if there is any method for cypher-query against the just created named graph ?

One workaround is to push this named graph into an offline/empty Neo4j Graph, ie CALL gds.beta.graph.export('my-graph', { dbName: 'mydatabase' }) . However, this method is less convenient because we often want to check if the named graph is projected correcly before applying, eg PageRank on it. And the projection can be a trial-and-error cycle.

Answer 1

There is currently no other way of querying the named graph other than the workaround you already found.

However, there are additional functions, eg gds.util.nodeProperty that allow you to access a node property in the named graph without writing it back to Neo4j. An example for querying a score property could look like this:

CALL gds.graph.create('my-graph', 'User', 'LINK');
CALL gds.pageRank.mutate('my-graph', { mutateProperty: 'score' });
MATCH (user:User)
WHERE user.name = 'Alice'
RETURN
    user.name AS name,
    gds.util.nodeProperty('my-graph', id(user), 'score') AS score

Could you maybe elaborate why your projections are "trial-and-error" cycles. Maybe an option is to run your validation queries on the subgraph you want to project?