Create a native projection having nodes with a specific name

[abhishekjayant1]

I want to use similarity algorithm and for that, I am trying to create a native projection which has nodes having a specific name (for example, create a native graph containing nodes of label 'Person' where name of the person is 'Carl' and 'John'). Thanks in advance!

[tlibersat]

Hi @abhishekjayant1111, I’m not sure of what your question is
But you can create your nodes with Person as label and names as properties and then launch the algorithm on the native projection. Does it answer your question?

[abhishekjayant1]

Hi @tlibersat , thanks for your reply. This didn't answer my question. I want to create a native projection, which has nodes of label 'Person' and their names should be 'Carl' and 'John' ( I want to basically select a few nodes from 'Person' nodes on the basis of their name).

[Cobra]

Hello @abhishekjayant1111

You will have to use gds.graph.create.cypher:

CALL gds.graph.create.cypher(
    'my-cypher-graph',
    'MATCH (n:Person) WHERE n.name IN ["Carl", "John"] RETURN id(n) AS id',
    'MATCH (n)-->(m) RETURN id(n) AS source, id(m) AS target'
)

Regards,
Cobra

[abhishekjayant1]

Hi @Cobra , thanks for your reply. I used the above query I got an error saying:

"Failed to invoke procedure gds.graph.create.cypher: Caused by: java.lang.IllegalArgumentException: Failed to load a relationship because its source-node with id 732 is not part of the node query or projection. To ignore the relationship, set the configuration parameter relationshipValidation to false."
How can I set this property to false in the cypher projection?
Thanks a lot for all of your responses, I feel great to be a part of this wonderful community.

[Cobra]

You only want nodes in your projection, no relationships?

Here is the link to the Github issue.

CALL gds.graph.create.cypher(
    'my-cypher-graph',
    'MATCH (n:Person) WHERE n.name IN ["Carl", "John"] RETURN id(n) AS id',
    'MATCH (n)-->(m) RETURN id(n) AS source, id(m) AS target',
    {validateRelationships: False}
)

[abhishekjayant1]

Hi @Cobra , I do want relationships in my projection. I used validateRelationships and ended up having 0 relationships in my projection.

[Cobra]

Hello @alicia.frame1

I think there is an issue with the error message:

"Failed to invoke procedure gds.graph.create.cypher : Caused by: java.lang.IllegalArgumentException: Failed to load a relationship because its source-node with id 732 is not part of the node query or projection. To ignore the relationship, set the configuration parameter relationshipValidation to false."

But it should be validateRelationships instead of relationshipValidation.

I also think the documentation is not updated with this new configuration parameter

Regards,
Cobra

[Cobra]

That's why you should select all Person nodes in your graph. If you really want to select a few Person nodes, you must make sure they have relationships between them. For example, Carl and John must be directly connected together. The validateRelationships must be used when you only want nodes. So in your case, you should use this query:

CALL gds.graph.create.cypher(
    'my-cypher-graph',
    'MATCH (n:Person) RETURN id(n) AS id',
    'MATCH (n)-->(m) RETURN id(n) AS source, id(m) AS target'
)

After if you are only interested about John and Carl similarity results from the algortihm:

CALL gds.nodeSimilarity.stream('my-cypher-graph')
YIELD node1, node2, similarity
WITH gds.util.asNode(node1).name AS p1, gds.util.asNode(node2).name AS p2, similarity
WHERE p1 = "John" AND p2 = "Carl"
RETURN p1, p2, similarity

[alicia_frame1]

@Cobra - I've created a card and we'll fix the error message

[Cobra · Accepted solution]

Solution of the thread (done by private message).

First, I create the projection of the whole graph:

CALL gds.graph.create('graph', '*', '*')

Then I compute similarity on it:

CALL gds.nodeSimilarity.stream('graph')
YIELD node1, node2, similarity
WITH gds.util.asNode(node1) AS n1, gds.util.asNode(node2) AS n2, similarity
WHERE n1:Adverse_Event
AND n2:Adverse_Event 
AND exists((n1)-[:reports]->(:SideEffect {SideEffect: "death"})) 
AND exists((n2)-[:reports]->(:SideEffect {SideEffect: "death"})) 
RETURN n1, n2, similarity

I don't know the name of the property in the SideEffect node so I put SideEffect as property name.