The Introduction of Optional

When Optional came out there was a lot of noise about how it was supposed to replace Null and help do away with NullPointerExceptions. So people started using Optional instead of Null in places where it really wasn’t appropriate, didn’t make sense, and certainly didn’t make code any cleaner or easy to understand.

You’d see things like this:

if (Optional.ofNull(variable).isPresent()) then {
           someCode();
}

Which, of course, isn’t any clearer than just checking if variable is Null. Its’ possible this might have led to a backlash against Optional, and it seems to me that it’s not used today as much as it should be. So let’s look at what Optional really does.

The Purpose of Optional

To better understand how Optional should be used, we’ll use an example that could be a real world programming situation:

An Example

Imagine that we have a system with customer accounts and invoices. Somewhere, there’s a persistence layer that will look up a customer account and instantiate a domain object for the account. In most systems I’ve seen, it’ll have a signature something like this:

public CustAccount getAccount(String accountNumber)

In this system, if we have an invoice and it has an account number referenced on it, then that account has to exist on file. If not, then we have data corruption and that’s a real problem. But if someone enters an account number into a screen to do a lookup, then they may have just typed in an invalid account, and that’s not necessarily a problem.

These two different situations need to be resolved differently, and the situation is all about the context of the code that calls the getAccount() method. Usually, systems are written so that getAccount() just returns Null if the account isn’t on file, and it’s up to the calling code to remember to check for Null, and decide what it means if the answer is Null.

But that’s not really how you are supposed to use Null, and Java has always had an alternative that seems to be rarely used, and that’s Checked Exceptions.

Checked Exceptions

Checked Exceptions are a way of declaring to any calling routines that your method might legitimately not be able to return a result, and forcing them to deal with that possibility. Of course, many programmers just put an empty catch clause (or put a stackdump in it), or throw the exception all the way up to the top where it’s mishandled.

So, in order to set this up properly, we really need two methods in our persistence layer, one for the case where the account needs to be on file, and one where it might not be there. Something like this:

public CustAccount getAccountIfPresent(String accountNumber) throws MissingAccountException

public CustAccount getAccount(String accountNumber)

MissingAccountException is a checked exception. If getAccount() cannot find the account on file, then it throws some sort of RuntimeException, maybe something called DataCorruptionException. Neither method is capable of returning a Null value.

The real problem with getAccountIfPresent() is that it’s cumbersome and forces awkward code structures around the try-catch blocks and scoping of variables. So almost nobody ever uses this. They just pass Null when the account can’t be found.

The problem with that, though, is that there’s no way tell a consumer of getAccount() that Null is a reasonable return value or what it means without checking the code itself to see how the result is determined. And that’s a problem.

Optional to the Rescue

This is where Optional comes in. With Optional you can do away with the checked exceptions and lock the answer in a wrapper that requires the calling method to deal with the possibility that the answer might be missing when you take it out. So the method signature would change to this:

public Optional<CustAccount> getAccountIfPresent(String accountNumber)

With this, there’s only one way to get that CustAccount out of the Optional without handling its possible absence That’s the Optional.get() method, which should only be used inside an if statement that has already checked Optional.isPresent() first. Until Java 9, if you had Optional in a stream, you needed to use something like this:

stream().filter(Optional::isPresent).map(Optional::get)....

But now the Optional class has it’s own stream() method, which does the filter and get in one step. So you can do this:

stream().flatMap(Optional::stream)...

All of which means that there’s no longer any place that the Java language forces you to use Optional.get() any more.

Brian Goetz has said in a famous StackOverflow response:

(Public service announcement: NEVER call Optional.get unless you can prove it will never be null; instead use one of the safe methods like orElse or ifPresent. In retrospect, we should have called get something like getOrElseThrowNoSuchElementException or something that made it far clearer that this was a highly dangerous method that undermined the whole purpose of Optional in the first place. Lesson learned. (UPDATE: Java 10 has Optional.orElseThrow(), which is semantically equivalent to get(), but whose name is more appropriate.))

Many programmers first experience with Optional is when they start using the Streams facility, and encounter it with the .findAny(), or findFirst() methods. These methods return Optional because it’s entirely possible that your stream, especially after filtering, is empty and there may legitimately be no stream elements returned by these methods.

Learning From Kotlin

Kotlin handles this concept in a much more natural manner and it’s instructive to take a look at it to understand Optional better.

Most Variables Cannot be Null

In Kotlin, the default for a data type is that it cannot be null. When you declare a variable, it needs to be initialized right away. If it’s a field, it needs to be initialized in the constructor or in the init{} block that runs right after the constructor.

This forces you to think about how your fields and variables are created and used. Unlike Java the default instantiation isn’t Null, and if you want to allow Null you have to put in effort to implement it.

Think about it. Nobody really wants Null. It’s a pain, and the reason that you have to deal with it can often be because somebody else didn’t take the time to deal with it properly in some method you rely on.

Nullable Variables

Kotlin allows you to declare a variable as “Nullable”. You do it like this, putting a “?” after the type name:

var myNumber : Int?

No big deal right? Think again. myNumber isn’t an Int, it’s a Nullable Int or an Int?. You can’t use it like an Int. You can’t do:

myNumber = myNumber + 2;

and you can’t do this, either:

if (myNumber < 20) {}

Because those are operations that work on Int, which myNumber is not.

Nullable Methods

Methods (or “functions” in Kotlin) can return Nullable types. So you could declare something like this:

fun myFunction():Int? {

}

But then you have the same constraints that a variable of type Int? would have. It’s not an Int and cannot be treated as one.

Null Safe Operations

Kotlin supplies the “Safe Call Operator” - ?. If you want to do something with a nullable type, you put that ? after the variable name. Then the method call will only happen if the variable isn’t Null. Something like this:

 val myString : String? = myNumber?.toString()

If myNumber is Null, the toString() won’t be called and the result will still be Null.

The “Elvis” Operator

Here’s something that you cannot do:

val myString: String = myNumber?.toString()

Why not?

Because myNumber?.toString() is not a String, it’s a String?. You cannot assign it to a String you cannot perform String operations on it (at least not without the Safe Call Operator). Essentially, myNumber is an Int locked away in a safe, and you cannot take it out without telling the compiler what to do if it’s Null.

Enter the “Elvis Operator”:

val myString: String = myNumber?.toString()?: "I'm empty"

If myNumber is not Null, then myString will have whatever myNumber translates to as a String. If it is Null then it will be assigned “I’m empty”.

The Dangerous Alternative

The last way to get a value out of a nullable type is the !! operator. The Kotlin documentation describes it this way:

The third option is for NPE-lovers: the not-null assertion operator (!!) converts any value to a non-null type and throws an exception if the value is null.

You can think of it as the “Look ma! No hands!” operator. Or the “Please sir, may I have another NullPointerExeception” operator. You can see how the Kotlin documentation clearly discourages its use. Event the choice of !! means it just jumps out of the code at you.

For the sake of completeness, this is how you would use it:

val myString : String = myNumber!!.toString()

This will throw an NPE if myNumber is Null. It’s really the only way to get an NPE in Kotlin.

How This Relates to Optional

I think you can see Kotlin nullable types relate to Optional in Java. The Safe Call Operator is fairly close to Optional.map(), the Elvis Operator is similar to Optional.orElse() and the !! operator corresponds to Optional.get().

The big difference between Kotlin and Java in this respect is that in Java the use of Optional is optional. It’s mandatory in Kotlin. Sure, you could thumb you nose at the “man” and say, “I’m just gonna make all my variables nullable”. But you’d pretty quickly see that your code had become an unreadable nightmare of Safe Call Operators and convoluted logic.

You couldn’t do this:

var index : Int? = 0
.
.
var counter: Int? = index++

because the “++” operator doesn’t work in Int?. So you’d have to do this:

var index : Int? = 0
.
.
var counter: Int? = index?.let{it++}

Eventually you’ll decide not to make all of your variables nullable, and then you’ll have to decide which ones should be nullable. Ultimately you’ll end up having to decide why a variable should be nullable, and what it means when it is nullable. At that point, it’s not really about nullable any more, and it’s more about the contexts where data and answers might not be available.

Conclusion

Understanding this purpose of Optional makes it much easier to use it appropriately. Ideally, you shouldn’t ever be creating an Optional unless it’s the return value of a method. Optional shouldn’t be a field in a class, and Optional shouldn’t be a parameter passed to a method. Further, there needs to be logic somewhere that unpacks the Optional and handles the situations where it’s empty in a contextually appropriate manner.

The motivation for Optional isn’t to avoid Null. Yes, it does this, but that’s a side-effect. Really, Optional is about handling situations where data might be missing, for whatever reason. It’s about telling the consumers of a method that, “Hey! There may not be a result here”, and forcing them to deal with it without having to look into your code to understand what Null means, or even if you method might return it.

Categories:

Updated: