neoid | Extend Ruby on Rails ActiveRecord | Database library

I have this somewhat huge query which is returning the correct result when I run it on SQL Server Management Studio:

So I imagine there's some documentation somewhere that discusses this, but I haven't found it. (I don't expect an answer to have benchmarks, just ~"in general, this is what happens")

It is my understanding that each label indexes its member's neoID, so adding a label to a node, and querying on that label, yields a performance benefit. But how does the performance compare if you have 10 labels on a node (vs 1 label)? What about 100 labels? Or 1000 labels? I imagine that adding more labels to a node comes with some sort of penalty? (other than the additional space the index takes up)

Say I have a db with :Book, :Person, :Organization, and :CommunityGroup nodes. Each has a uuid property.

• (:Book)-[:OWNER]->(:Person)
• (:Book)-[:OWNER]->(:Organization)

:Person or :Organization nodes might own a :Book, but :CommunityGroup nodes never do. If a client passed my app an uuid for a book owner, I could find that node a few different ways

1. Match the node with that uuid which had either the :Person or :Organization labels.
   • MATCH (:Book)-[:OWNER]->(n {uuid: $the_id}) WHERE n:Person OR n:Organization
2. Ask the client for the node type, and match using that uuid and type.
   • MATCH (:Person {uuid: $the_id})
3. Add an :Owner label to the :Person and :Organization nodes
   • MATCH (:Owner {uuid: $the_uuid})

Option 2 must be the fastest. 3 is the overall nicest to query on. 1 minimizes extra labels + client work.

Obviously this example is small, but what if I created extra labels (like the :Owner label) all over the place. Is there a performance penalty for that? I could easily see a single node acquiring 10 or 20 labels, maybe more, depending on how liberal I was in applying them. The :Owner label is technically unnecessary, as I have the labels needed to find the node already...

Thoughts? Thanks for any feedback!

Note: I found this S.O. question on multiple label performance, but it's only addressing a scenario with a few labels (~4), rather than 10 / 20 / 100.

I have a C# object model that I want to persist with Neo4j. It contains dozens of Projects, which each has 1000-4000 children. I currently use the official C# driver, but I could switch to something else. Doing this one child at a time is, unsurprisingly, inefficient. This takes several minutes for one project: