I pulled the Graph Data Science package and organized it to create my custom functions. I developed my function and tested it. The algorithm test and procedure test(Call from cypher - I mean like algo and proc packages in GDS) both passed. Now I want to import my library into Neo4j.
In open-packaging module I excluded other packages and now shadowCopy command only produces a jar with only classes that I created.
The problem is that, when I examine debug.log during Neo4j initialization, I see an error saying 'Failed to load com.semanticspace.shortestpath.DijkstraMultiplePairs from plugin jar /var/lib/neo4j/plugins/semantic-space-gds-2.0.4.jar: org/neo4j/gds/shaded/com/carrotsearch/hppc/BitSet' (There is an import to this class in my class)
There are a lot of errors like that : having problem loading packages starting with 'org/neo4j/gds/shaded/'. So I modified the project to produce classes without 'org/neo4j/gds/shaded/' prefix, then again it 'fails to load'.
This class is also used in GDS Dijkstra method so I expect it resides in GDS jar at least. There are many others like that, maybe there is another prefix for these classes.
And when I call the function via Cypher it says 'Unable to inject component to field taskRegistryFactory, please ensure it is public and non-final: null'. It's probably because neo4j failed to load this class.
Can you tell me how should I import this kind of classes in my custom GDS functions?
UPDATE : So I figured that out that Neo4j expects from a package imported from within a package to reside in same jar. So I made it possible and now there is no 'Failed to error' is shown in the debug.log relating to my package.
Although in Cypher I still see the same error. And with the second call onwards, it's this : 'Unable to inject component to field taskRegistryFactory, please ensure it is public and non-final: Could not initialize class org.neo4j.gds.compat.Neo4jProxy'
I made the class non-final and it is public (Although the GDS functions are final). Still the same error.
Thanks.