neo4j 2.3.1 support? about neo4j-mazerunner

[rotten]

It is somewhat cumbersome, but the approach I took this afternoon that seems to work using the kbastani/spark-neo4j docker based solution. [ https://github.com/kbastani/spark-neo4j ]

1. Use neo4j-shell to export a graphml of my subgraph of interest by running it with a cypher query from my production 2.3.2 database.
2. Use a chain of 'sed' commands to rename all of my edges in the graphml export file to have the same name. (The sed rename is much faster than cypher in this case)
3. Open the app_graphdb_1 docker shell, and install the 2.3.2 version of neo4j-shell-tools. Restart the shell to pick them up. (stop/start neo4j doesn't work in my docker vm anyway)
4. scp the edge-renamed graphml into my docker instance.
5. Import it with neo4j-shell.
6. Run the pagerank service.
7. Export the nodes with their page rank back out. (neo4j-shell again)
8. Copy that file over to the original graph database instance.
9. Load the page ranks back in.

So far it seems to be much faster than Graphaware's noderank (which runs for days and never converges), and doesn't run out of memory like Python Graph-Tool. (which can also soak up the graphml export). I'm still experimenting with the end-to-end process, meanwhile I wanted to share what I came up with today to leverage these really sweet tools until @kbastani has time to help us get to direct support of the 2.3.x neo4j releases.

[rotten]

I'm using 2.3.2. I was unable to get mazerunner to work with that either. There isn't an easy way to downgrade from 2.3.2 to 2.2.1. Any other suggestions for getting it to work?

If I wanted to export the edge list from Neo4j with Cypher and then upload the page rank results independently as well, is there a subset of these docker containers that I can use to still leverage some of the fine work here, but not have to figure out how to reload my hundreds of millions of nodes and relationships into an older Neo4j instance?