Using the New Null Conditional Operator in C# 6
Like this article? We recommend
Like this article? We recommend
Like this article? We recommend
One of the most versatile and useful additions to the C# language in version 6 is the null conditional operator. As I've been using C# 6 in my projects, I'm finding more and more scenarios in which this operator is the simplest and clearest way to express my intent.
Ask yourself how much of your code must check a variable against the null value. Chances are, it's a lot of code. (If not, I'd worry about the quality of your codebase.) In every one of those null checks, the null conditional operator may help you to write cleaner, more concise code. We all want our code to be as clear and concise as possible, so let's explore this feature.
Null Conditional Operator Syntax
The null conditional operator (?.) is colloquially referred to as the "Elvis operator" because of its resemblance to a pair of dark eyes under a large quiff of hair. The null conditional is a form of a member access operator (the .). Here's a simplified explanation for the null conditional operator:
The expression A?.B evaluates to B if the left operand (A) is non-null; otherwise, it evaluates to null.
Many more details fully define the behavior:
- The type of the expression A?.B is the type of B, in cases where B is a reference type. If B is a value type, the expression A?.B is the nullable type that wraps the underlying value type represented by B.
- The specification for the feature mandates that A be evaluated no more than once.
- The null conditional operator short-circuits, which means that you can chain multiple ?. operators, knowing that the first null encountered prevents the remaining (rightmost) components of the expression from being evaluated.
Let's look at some examples to explain those behaviors. Consider this simplified Person class:
public class Person
{
public string FirstName { get; set; }
public string LastName { get; set; }
public int Age { get; set; }
}
Assume that p represents a person. Consider these two statements:
var name = p?.FirstName; var age = p?.Age;
The variable name is a string. The value of name depends on the value of p. If p is null, name is null. If p is not null, name is the value of p.FirstName. Note that p.FirstName may be null even when p is not.
The variable age is an int? (which is another way of specifying a Nullable<int>). As with name, the value of age depends on the value of p. If p is null, age is an int? with no value. If p is non-null, age is the wrapped value of p.Age.
That's the basics. The power of this feature come from all the scenarios where this feature enables cleaner code.
Code Cleanup with the Null Conditional Operator
Suppose people is a variable that represents an IList<Person>. Now, we have a couple of levels of member access to navigate, and one of those levels uses the indexer syntax ([ ]). We could write this statement:
var thisName = people?[3]?.FirstName;
The ?[] syntax has the same semantics as the ?. operator: It's how you access the indexer on an array, or a class that implements an indexer. The rules for its behavior are the same. If people is null, thisName is assigned the value null. If people[3] is null, thisName is assigned the value null. Otherwise, thisName is assigned the value of people[3].FirstName. However, if people is not null, but has fewer than four elements, accessing people[3] will still throw an OutOfRangeException.
In the earlier example, I used the null conditional operator on both member accesses. That's a typical pattern because the null conditional operator short-circuits. The evaluation proceeds from left to right, and it stops when the expression evaluates to null.
Let's look at a second example. Consider this enhancement (shown in bold) to the Person class so that it contains a reference to a person's spouse:
public class Person
{
public string FirstName { get; set; }
public string LastName { get; set; }
public int Age { get; set; }
public Person Spouse { get; set; }
}
You would retrieve the spouse's name as follows:
var spouseName = p?.Spouse?.FirstName;
Semantically, this is roughly equivalent to the following:
var spouseName = (p == null) ? null : (p.Spouse == null) ? null : p.Spouse.FirstName;
or, in a more verbose form:
var spouseName = default(string);
if (p != null)
{
if (p.Spouse != null)
{
spouseName = p.Spouse.FirstName;
}
}
This example shows how much cleaner code becomes by using the null conditional operator. The more lengthy form is quite a bit more verbose. While this example used the ?. operator on each member access, that's not required. You can freely mix the null conditional operator with normal member access. If the above assignment were used in a routine where p had already validated to be non-null, you could assign the spouse's name as follows:
var spouseName = p.Spouse?.FirstName;
Or, if a particular scenario will be called only using people that are married, you can assume the Spouse property will never be null:
var spouseName = p?.Spouse.FirstName;
When you mix the null conditional operator with the traditional member access operator, the resulting expression will return null if the left operand of ?. evaluates to null, and throw a NullReferenceException if the left operand of ?. evaluates to null. Remember that the short-circuiting still applies, so p?.Spouse.FirstName returns null when p is null, whereas p.Spouse?.FirstName throws a NullReferenceException when p is null.
Other Scenarios
There are a couple more interesting scenarios that ?. enables. I've often used it for raising events. A typical scenario is when a type supports INotifyPropertyChanged. Let's expand the Person class to support this interface, and raise the PropertyChanged event whenever one of the properties changes.
Here is how I would implement the FirstName property:
public string FirstName
{
get { return firstName; }
set
{
if (value != firstName)
{
firstName = value;
PropertyChanged?.Invoke(this, new PropertyChangedEventArgs(nameof(FirstName)));
}
}
}
private string firstName;
Examine the highlighted line of code carefully. I'm also using the new nameof operator. (I'll cover that in more detail in a later article.) This line uses the null conditional operator to raise the PropertyChanged event only if code has registered a handler on that event. It would be nice if I could put the ? directly before the invocation , but that would lead to syntactic ambiguities. The C# 6 team disallowed this syntax. That's why I'm explicitly using the Invoke method on the System.Delegate class to invoke the event handler. Astute readers may be wondering if this usage is thread-safe. In earlier versions of C#, we would write this construct as follows:
var handler = PropertyChanged;
if (handler != null)
{
handler(this, new PropertyChangedEventArgs("FirstName"));
}
We would capture the current value of the event handler, and then test that value and invoke the handler if it was not null. The null conditional operator does that same work for us. It evaluates the left operand of the ?. operator only once, storing the result in a temporary variable. In this construct, that's important for thread safety. It's also important in many other scenarios, as I describe shortly.
Let's return to this example, with a small change:
var spouseName = GetPerson()?.Spouse?.FirstName;
Notice that the variable p has been replaced by a method call. That method call may have side effects or performance implications. For example, suppose GetPerson() makes a database call to find the current user. Earlier, I translated that expression to a longer version using if statements. The actual translation is more like the following code:
var spouseName = default(string);
var p = GetPerson();
if (p != null)
{
var pSpouse = p.Spouse;
if (pSpouse != null)
{
spouseName = p.Spouse.FirstName;
}
}
Notice that GetPerson() is called only once. Also, if GetPerson() returns a non-null object, GetPerson().Spouse is evaluated only once (through the temporary variable p). The result of this work is that you can use the null conditional operator in scenarios that reference return values from property accessors, indexers, or method access without worrying about possible side-effects.
The event-handling scenario is certainly the most common delegate usage for ?. but it isn't the only one. We can create filters that handle logging based on a delegate type:
public class Logger
{
private Func<Severity, bool> Publish;
public void GenerateLog(Severity severity, string message)
{
if (Publish?.Invoke(severity) ?? true)
{
SaveMessage(severity, message);
}
}
}
This portion of a Logger class uses the Publish delegate to determine whether a message should be written to the log. It uses the ?. operator to safely check an optional delegate that filters messages. It also leverages the existing ?? operator so that if the Publish delegate is null, all messages are published. It's syntactic sugar of the sweetest kind.
Finally, there is one other scenario in which the null conditional operator comes in quite handy: variables that may implement an interface. This usage is particularly useful with IDisposable. When I create libraries, I often create generic methods or classes that create and use objects. Those objects, depending on the type, may or may not implement IDisposable. The following code shows a quick way to call Dispose() on an object only if it implements IDisposable:
var thing = new TFoo(); // later (thing as IDisposable)?.Dispose();
In practice, I've only used this idiom when I create generic classes that create objects of the types specified by their type parameters.
Some Initial Guidance on Working with the Null Conditional Operator
I've been very aggressive in updating existing code bases with this feature because the new syntax is so much more concise and clear. I've replaced any number of null checks with the null conditional operator. If I combine it with the null propagating operator (??), I can often replace several lines of code with a single expression.
In the process, I've also found bugs that have lingered in a code base. As I described earlier in this article, the code generated by the ?. operator is carefully constructed to evaluate the left side of the operand only once. I've found that handwritten algorithms may not be so carefully managed. Because the replacement can change code behavior, it does require adding tests to make sure that no other code relies on the existing hand-coded algorithm. Overall, though, I've aggressively reviewed classes and replaced code to use the idioms shown in this article. This usage has reduced code size, reduced bug counts, and made my code more readable.