Misusing System.Random

This is a series on common ways that System.Random is misused, based on my experience reviewing code test submissions from job candidates.

We’ve been interviewing for a new position at work recently, and we give all our candidates a coding test to do, usually between the first and second interview. The test involves design around caching of a key/value store in a multi-threaded system. Part of the test requires generating a random set of keys for exercising the cache. I’ve reviewed submissions from candidates so far and, to my surprise, not a single one used System.Random correctly.

I was amazed at how little candidates for a mid to senior-level software development position knew about generating random numbers, much less multi-threading. I don’t know if it says more about the candidates, or the design of System.Random, but I thought I’d write a few posts to highlight the problems and show the right way (or ways) to do things.

How can I misuse thee? Let me count the ways

Here is a list of the misuse I’ve seen, and the number of interview candidates of the I’ve reviewed who misused System.Random in their code in this way.

Misuse	Candidates
Incorrect arguments for .Next(...)	22 (63%)
Incorrect seeding	12 (34%)
Singleton for multiple threads	21 (60%)
Instance per invocation	7 (20%)

The following snippet shows examples of all of these misuses of System.Random in the context of generating random numbers from 0-9, inclusive.

var rnd1 = new Random(0);  // Incorrect seeding

new Thread(DoStuff).Start();
new Thread(DoStuff).Start();

void DoStuff()
{
    // Random is not thread-safe
    var value1 = rnd1.Next(0, 9);  // Incorrect parameters

    var value2 = new Random().Next(10);  // Instance per invocation
}

What’s next?

In this blog post series I’m going to dive into each of these wrong ways of using System.Random, explain why they’re wrong, and show the correct usage. I’ll finish up with thoughts on abstracting random number generation to enable thread-safety, dependency injection and unit testing.