How I Learned to Live with Nullable Reference Types

12 Feb 2022

I hated the nullable reference types feature introduced in C# 8.0 for quite some time and avoided using it.

It made new warnings, and lots of them for some projects. Especially for existing projects. I was not a fan of this wave of yellow warnings after a build. Those of you that have grown up on teams that had a strict no warning compile standard should be able to relate.

It is 2022, and I decided to flip the bit, add some new styles to my coding, and dive into nullable reference types. Having the compiler help you remove bugs is a good thing. I was starting a new project so let’s start it right.

Enabling Nullable Reference Types

Prior to C# 8.0 all types were nullable and that was just how code worked. If you had a reference then it could be null. You were a developer, so you checked for null when required. It was called programming.

Enabling the feature was simple. Add the following to your .csproj file, or check a box in your IDE. You are ready!

<Project Sdk="Microsoft.NET.Sdk">
    <PropertyGroup>
        <Nullable>enable</Nullable>
    </PropertyGroup>
</Project>

Now the compiler is checking your code for what it calls null-state. If you are going to use a variable the compiler will determine if:

1. The variable has been assigned to a value that is known to be not null.
2. or the variable has been checked against null and hasn’t been modified since that check.

The official docs can be read here for more details. You just need to keep doing what you have, hopefully, been doing. Either assign the variable to a known value, or check the variable against null, before using the variable.

I have been doing that already. Except when I wasn’t.

Am I Loving Nullable Reference Types?

I would not say that I am loving it, but I am enthusiastically supporting it. Every developer has their own coding style. It is how, when presented with code you swear you never touched, barring a damning git commit entry, you actually come to understand that you did write it. No more denying it because it has your style all over it.

Here is what I have been doing to adopt using nullable reference types:

• Accept the warnings. This was actually tough to do because, as I stated before, the warnings introduced a level of stress. I was not used to having warnings in my projects and surely not as many as enabling this feature introduced (let’s be real - there are some there, but they are usually only a few and accepted for a reason).
• Fix the warnings. Let’s face it, the warnings are there because your code is wrong (now). It could allow a null reference to sneak in and cause the associated pain. This is where I finally grew up and accepted the feature in its entirety. You might say that I have been enthusiastically supporting it!
• Use nullable variable annotations. This is the type annotation ? that says this type is nullable. It is appended to the type declaration as in string? name;. Adding this to your existing code is really documenting what already exists. Prior to enabling nullable reference types, all types had this annotation applied implicitly. You are now just declaring what you already had. My first step was to add it where needed to get a clean compile. It can be removed later as my null sensibility increases.
• Keep checking parameters. Just because the compiler tells you that the null state of a variable of not-null doesn’t mean some joker (generally yourself) will not pass a null value into it (see the null-suppression operator below). Do not remove your unit tests for these conditions. The compiler is helping you, but it is not writing your code for you.
• Use the null-suppression operator. Use this operator to remove warnings, as you fix them, when you must pass null into a method. I use this liberally in my unit tests to ensure the parameters are checked and the appropriate exceptions are still thrown. Please remember, the compiler is helping remove the chances of getting a System.NullReferenceException. There are no guarantees.

Next Steps for Better Coding

Part of the coding process is to document your code. I do this in two (2) ways:

1. Write clear code that is easy to understand. I do not try to get fancy. That should be enough (but generally is not).
2. Document all your public classes and public methods on those classes using XML doc comments. You may never actually generate the documentation for browsing but the IDEs of today will display the help when asked.

I am going to be adding a third tool to the documentation tool chest. That is the attributes for null-state static analysis. I am not currently planning on overhauling any codebase with attributes. This is a process and the introduction will be gradual. I do believe that once incorporated into my style it will help.

I know finally enabling nullable reference types was an overall plus for me.