SoundCloud Follower Maze Code Submission

Problem description can be found here.

Requirements

This solution was implemented using Java 8. It requires an Oracle JDK version 1.8 or higher. It will not work with previous JDKs. It was not tested on Open JDK.

Please make sure your system points to the correct JDK by setting the value of the $JAVA_HOME environment variable.

On macOS, you can find the path of a specific JDK version using the java_home command:

/usr/libexec/java_home --version 1.8

Starting the server

./run.sh

The server will listen on ports 9090 and 9099.

Running the test suite

./test.sh

Regarding documentation

I'm a big fan of Clean Code. I strongly believe that code should be readable and document itself without the need of comments.

That's why I avoided writing obvious comments just for the sake of it. Instead, I wrote comments only where I felt that they would make a difference, in the form of high-level JavaDocs.

Regarding tests

Most of the code is fully covered by some form of unit tests or white-box integration tests. The only place that is not tested is the Server class. Since this is the highest-level component, I'd rather have black-box integration tests or functional tests to cover it. Otherwise, I'd need to orchestrate several mocks in a complicated setup for a questionable gain.

If I had more time I would probably write those black-box tests using Mountebank to record and stub TCP iteractions.

In additional to functional tests, we also need load and performance tests. These can be made using JMeter or another tool that supports testing TCP sockets.

Future improvements

If I had more time I would try to do the following:

• Instead of relying on an memory-based queue data structure, use a persistent message queue such as RabbitMQ, ActiveMQ or ØMQ. This way, if the server dies and needs to be restarted, requests won't be lost.

• Instead of persisting the follow/unfollow status in a memory-based data structure, store this relationship status in a persistent key-value store, such as Redis. This way, if the server dies and needs to be restarted, this information won't be lost.

• Run the server in a container (eg: Docker). Orchestrate possible dependencies (eg: Redis, RabbitMQ) using docker-compose or a similar tool.

• Improve monitoring (for instance, by adding a health status check).