Using docker for testing

The other day, I wrote a tiny application using neo4j. While neo4j is a nice piece of software, it supports just one running database. Running tests required a clean database, so I always had to stop the neo4j instance, change the database path to test.db, run the tests and do the same hamsterdance again to have a development database in place.

Very tedious!

So why not use docker to spawn a fresh database? It’s the perfect solution for a throwaway test db. Here’s how I did it:

Create a shell script that starts the database and runs tests

  #!/bin/bash
  echo "Starting docker containers"
  docker run -p 7474:7474 --name 'test\_neo4j' -d tpires/neo4j
  echo "Running tests"
  sleep 10
  if [ `uname` == "Darwin" ]; then
    # we have boot2docker on OSX
    neo_ip=`boot2docker ip | grep -Eo '[0-9]{1,3}\.[0-9]{1,3}\.[0-9]{1,3}\.[0-9]{1,3}'`
  else
    neo_ip=`docker inspect test\_neo4j | grep IPAddress | cut -d '"' -f 4`
  fi

  # running tests
  NEO\_PORT\_7474\_TCP\_ADDR=$neo\_ip bundle exec rspec

  # catching exit code
  rc=$?
  echo "Stopping docker containers"
  docker stop test\_neo4j

  echo "Removing obsolete containers"
  docker rm test\_neo4j
  # exiting with exit code of tests
  exit $rc

This shell script

• starts a neo4j instance
• runs ruby rspec tests
• catches the exit code of the tests
• stops and removes the neo4j container
• exits with the test exit code

Only thing you have to do application wise, is to use the environment parameter NEO_PORT_7474_TCP_ADDR for connecting to the neo4j.

But, as we are all big fans of 12 factor apps, we do that anyways, right?

So far this approach has worked out really well. No more annoying “how do I have to setup the test db again?” question. Just install docker and you’re ready to roll!