Cypher: Avoid cycle in multiple relationship paths - neo4j

Question

My website has pages dedicated to some events, represented by nodes in neo4j. Those events possess sub-events which are relationships under neo4j, and which correspond to links on the source page to the target page. I currently have a search engine that highlights the links to the searched events, but it is flawed by cycles in the data model. Indeed it is highlighting all the links that contain cyclic references to the current page if this page contains any link to the searched event.
The aim is therefore to have a query that is able to flag the nodes and relationships which are related to the searched events, without flagging a path only because of cyclic relationships.
I've created a small dataset representative of the issue that you can build using this query:
CREATE
(r:Event:Searched {name:'R', tag:1}),
(d:Event:Searched {name:'D', tag:1}),
(o:Event {name:'O'}),
(a:Event {name:'A'}),
(b:Event {name:'B'}),
(c:Event {name:'C'}),
(e:Event {name:'E'}),
(o)-[:hasEvent]->(a),
(o)-[:hasEvent]->(e),
(o)-[:hasEvent]->(r),
(o)-[:hasEvent]->(c),
(a)-[:hasEvent]->(b),
(b)-[:hasEvent]->(o),
(c)-[:hasEvent]->(d)
Which produces the following graph:
My aim is to have a query that only fetches nodes C and O, but not A or B, as the only reason they are flagged is that O is already flagged:
.
My current query that I need to fix is the following:
MATCH path=(upper:Event)-[:hasEvent*]->(source:Event:Searched)
RETURN upper
I hope you can help me, I couldn't manage to make similar questions' answers work on my specific case.
Ideally, the solution shouldn't be too computing-intensive, as my real model is quite big (2.300.000 nodes and 9.500.000 relationships), and the current indexing in the search engine is already quite slow.
Thank you in advance for your help

Answer 1

You could try out apoc.path.expandConfig procedure. It has a uniqueness property, which you can be configured that a node cannot be traversed more than once.
MATCH (n:Event)
CALL apoc.path.expandConfig(n, {
relationshipFilter: "hasEvent>",
labelFilter: "/Searched",
uniqueness: "NODE_GLOBAL"
}) YIELD path
RETURN [n IN nodes(path) WHERE NOT n:Searched] AS upper
However, this query will still return A and B, because with MATCH (n:Event) you start looking from every node (regardless of the relationship between O and A). But if I understood you correctly, you don't wan't to start from all nodes, but from a specific one ("pages dedicated to some events"). So you might want to start with, e.g., MATCH (n:Event {name: "O"}) that will return only O and C.