Is there any workaround for "allShortestPaths(...) requires a pattern containing a single relationship"

Neo4j Graph Platform Cypher
1

I have the following structure

(r:Industry)->[:REFERS]-(data:IndustryData)-[:OCCURRED]->(m:Month)

(r:Region)<-[:REFERS]-(data:RegionData)-[:OCCURRED]<-(m:Month)

(s:Size)<-[:REFERS]-(data:SizeData)-[:OCCURRED]<-(m:Month)

So, all the entities share Month nodes.

Using Industry as an input I want to find all the paths to Region and Size.

I can do it for 1 relationship, e.g. Industry to Size:

MATCH (i:Industry)
WHERE i.value in ["7379", "73", "Information Technology"]
MATCH (s:Size)
WHERE s.value in ["Large2", "Large"]
MATCH path=allshortestPaths((i)-[*]->(s))
RETURN path

However, when I execute the next query

MATCH (i:Industry)
WHERE i.value in ["7379", "73", "Information Technology"]
MATCH (s:Size)
WHERE s.value in ["Large2", "Large"]
MATCH (r:Region)
WHERE r.value in ["USA", "North America"]
match path=allshortestPaths((r)<-[*]-(i)-[*]->(s))
return path
LIMIT 50

I receive the following error allShortestPaths(...) requires a pattern containing a single relationship

What alternatives do I have? Changing the structure also is an option for me.

2

Hi @miron4dev,

I'm not sure about your strategy with shortestPath. You may be getting the answers that you want without the right strategy to do so. In any case, try this query:

MATCH (i:Industry)
WHERE i.value in ["7379", "73", "Information Technology"]
with collect(i) as inter
MATCH (s:Size)
WHERE s.value in ["Large2", "Large"]
with inter, s
MATCH (r:Region)
WHERE r.value in ["USA", "North America"]
with inter, s, r
match path=allshortestPaths((r)-[*]-(s))
with r, s, inter, path, nodes(path) as ns
where any(n in ns WHERE n in inter)
return path
LIMIT 50

H

3

My idea was in using cost attribute as a weight for relationships, but it seems allshortestPaths function doesn't take it into account.

4

Hi @miron4dev,

allshortestPaths/shortestPath just takes into consideration the number of hops. Was the previous query useful? I'm trying to get more insight on your intentions :)

1 Like
5

You'll probably want to look at the Graph Data Science library and its shortest path implementation, that should take into account relationship weights.