Parameter Null Checking in C#

Parameter Null Checking in C#

Last month, some eagle-eyed folk spotted a change in the C# 11 pipeline. This triggered a fair bit of discussion about the new feature. In this post I’ll dig a bit into the decision and explain a bit more about the feature.

What is parameter null checking?

Let’s start by looking at the feature. For a couple of decades we have been writing the following code to check that an argument we were passed isn’t null:

public void RunSomething(object myParam)
{
    if (myParam is null)
    {
        throw new ArgumentNullException(nameof(myParam));
    }

    // Code I want to read
}

The more parameters a method has, the further you have to scroll to get to the code you want to read. This is the driver behind the parameter null checking feature.

Parameter null checking provides a syntax for achieving the same checks without writing all the code. You just add !! to the end of the parameter name.

public void RunSomething(object myParam!!)
{
    // Code I want to read
}

Now the code I want to read is right at the top of the method. Hooray. Algorithmically, I save (4 lines x parameters) - (2 characters x parameters) across my whole codebase. That’s approximately a gazelle-ian lines of code.

Under the hood, it doesn’t generate exactly the same code you might have written yourself, it generates a more performant version. This interests me, so here’s what the compiler turns the code into…

public void RunSomething(object myParam)
{
    <PrivateImplementationDetails>.ThrowIfNull(myParam, "myParam");

    // Code I want to read
}

The PrivateImplementationDetails alluded to in the above output is below. You could implement a version of this in your own codebase right now if you really wanted to.

[CompilerGenerated]
internal sealed class <PrivateImplementationDetails>
{
    internal static void Throw(string paramName)
    {
        throw new ArgumentNullException(paramName);
    }

    internal static void ThrowIfNull(object argument, string paramName)
    {
        if (argument == null)
        {
            Throw(paramName);
        }
    }
}

Why?!

If you followed the discussion, you will have found many questions or comments relate to the choice of adding two exclamation marks to the parameter name. This can be broken down into the following component questions:

To summarise the main theme of the questions, the choices that people didn’t understand were:

• Why did they choose !!
• Why does the !! go on the parameter name not the parameter type?

The answers to both questions came from Jared Par, who joined the Languages and Runtime Community Stand-up in February to discuss this topic.

Why was !! chosen? There is a chance that this feature might be made available inside methods – not just for the parameters. The syntax choice had to allow for this possible future development. The single exclamation mark syntax ! already exists within method bodies since C’# 8. This is the null suppression operator (or unary postfix ! operator to its friends). This operator forgives nulls, so the so the use of the double !! for holding a grudge against nulls falls into line with this concept without clashing with the existing feature. myParam! forgives and myParam!! would want revenge (or, throw an exception).

It might be used something like the below code, which is not (yet) valid in any version of C#.

public void RunSomething(object myParam)
{
    if (_someCondition) 
    {
        myParam!!;
        // Use myParam
    }

    // Code that doesn't need to use myParam
}

Why does the !! go on the parameter name not the parameter type? Because this isn’t type information; it’s a modification of the behaviour in response to the value.

With this little insight into the choice of syntax, we can understand a bit more how it fits into the language. It might not be perfect, but it does make sense.

This syntax […] didn’t have people jumping up in the room saying: ‘This is perfect, we have found the solution, this is great!’ […] this was the syntax we found that worked.

Jared Par

Summing up

If you take great exception to this syntax, you can carry on without it. You can write your conditional checks, or steal that compiler generated ThrowIfNull helper and at least reduce the lines of code by 75%. Just remember the amazing maths presented before!

(4 lines x parameters) - (2 characters x parameters)