Some code for trying out neo4j with python

api.py: a simple flask api

n4j.py: the cypher queries (that the api uses)

createdb.py: the file that generated fullset.csv, see that python file for the rules how the users are connected

A user has one property, "ourid", not to be confused with the internal id of the node.

Neo4j url

The url with username/password for Neo4j is hard coded in n4j.py

Starting the api

Dev environment

export FLASK_APP=`pwd`/api.py
export FLASK_DEBUG=1
flask run

Using gunicorn

gunicorn -w 2 api:app

Loadtest

npm install loadtest
node node_modules/loadtest/bin/loadtest.js -c 10 --rps 1000 "http://127.0.0.1:8000/2nd3rd/?ourid=3"

Importing test data (about 2 million rows)

Remove the WITH row LIMIT 30 to import the full 2 million entries also adjust the path to wherever this directory is.

Comment out dbms.directories.import in .neo4j.conf so neo4j will allow importing from files outside its own dir.

If it can't handle the full import at once, or you don't want to import all test data, use the chunks initdata/xa* instead (100k in each)

(notice that you have to bunzip2 the initdata/* files first)

Cypher code for import

CREATE CONSTRAINT ON (u:User) ASSERT u.ourid IS UNIQUE;

USING PERIODIC COMMIT
LOAD CSV WITH HEADERS FROM "file:///some/path/neo4jplayground/initdata/fullset.csv" AS row
WITH row LIMIT 30

MERGE(u:User {ourid:row.user})

FOREACH (f IN split(row.friends, ",") |
  MERGE(u2:User {ourid:f})
  MERGE(u)-[:FRIEND]->(u2))