Nearmap Go Test (Code Review Version)

Instructions for the standard Nearmap Go test are provided below.

For this exercise, a full solution has been provided (there is no need to provide any code yourself). However, we would like you to review the code so that we can have an in-depth technical discussion around the solution and how it can be improved.

The purpose of this assignment is to test your familiarity with Go, distributed systems concepts and unit testing.

Background

The source code that you are given is a very simple imitation of a key/value store:

• Database represents a client to the central store that takes a long time (500ms) to store and retrieve data.
• DistributedCache represents a client to the distributed cache (Redis for example) that takes much less time to turn around (100ms to store or retrieve).

This scenario is a simplified example of a typical high performance server cluster with a database, a distributed cache and multiple worker nodes.

Assumptions and requirements

• Data in Database never changes and can be cached forever.

• If Database.Value() returns nil for a key, the requested data item does not exist and will never exist.

• DistributedCache is initially empty.

• For a frequently-requested item your DataSource.Value() implementation should have a better response time than the distributed cache store (ie < 100ms).

• The user of the DataSource interface must not have to deal with thread synchronisation.

• Sufficient unit test coverage for the DataSource implementation.

• The solution should aim to minimise calls to the database.

• Use 10 goroutines (simulating separate threads on a single worker node) each making 50 consecutive requests for a random key in the range (key0-key9). I.e. there should be a total of 500 requests.

• For each request, print the requested key name, returned value, time to complete that request; similar to the following example:

  [1] Request 'key1', response 'value1', time: 50.05 ms
  [2] Request 'key2', response 'value2', time: 50.05 ms