Why should one use Objects.requireNonNull()?

Question

I have noted that many Java 8 methods in Oracle JDK use Objects.requireNonNull(), which internally throws NullPointerException if the given object (argument) is null.

public static <T> T requireNonNull(T obj) {
    if (obj == null)
        throw new NullPointerException();
    return obj;
}

But NullPointerException will be thrown anyway if a null object is dereferenced. So, why should one do this extra null check and throw NullPointerException?

One obvious answer (or benefit) is that it makes code more readable and I agree. I'm keen to know any other reasons for using Objects.requireNonNull() in the beginning of the method.

Answer 1

Because you can make things explicit by doing so. Like:

public class Foo {
  private final Bar bar;

  public Foo(Bar bar) {
    Objects.requireNonNull(bar, "bar must not be null");
    this.bar = bar;
  }

Or shorter:

  this.bar = Objects.requireNonNull(bar, "bar must not be null");

Now you know:

• when a Foo object was successfully created using new()
• then its bar field is guaranteed be non-null.

Compare that to: you create a Foo object today, and tomorrow you invoke a method that uses that field and throws. Most likely, you will not know tomorrow why that reference was null yesterday when it got passed to the constructor!

In other words: by explicitly using this method to check incoming references you can control the point in time when the exception will be thrown. And most of the time, you want to fail as fast as possible!

The major advantages are:

• as said, controlled behavior
• easier debugging - because you throw up in the context of the object creation. At a point in time where you have a certain chance that your logs/traces tell you what went wrong!
• and as shown above: the true power of this idea unfolds in conjunction with final fields. Because now any other code in your class can safely assume that bar isn't null - and thus you do not need any if (bar == null) checks in other places!

Answer 2

Fail-fast

The code should crash as soon as possible. It should not do half of the work, dereference the null and only then crash, leaving half of some work done causing the system to be in an invalid state.

This is commonly called "fail early" or "fail-fast".

Answer 3

But NullPointerException will be thrown anyway if a null object is dereferenced. So, why should one do this extra null check and throw NullPointerException?

It means you detect the problem immediately and reliably.

Consider:

• The reference may not be used until later in the method, after your code has already performed some side-effects
• The reference may not be dereferenced in this method at all
  • It could be passed to completely different code (i.e. cause and error are far apart in code space)
  • It could be used much later (i.e. cause and error are far apart in time)
• It may be used somewhere that a null reference is valid, but has an unintended effect

.NET makes this better by separating NullReferenceException ("you dereferenced a null value") from ArgumentNullException ("you shouldn't have passed in null as an argument - and it was for this parameter). I wish Java did the same thing, but even with just a NullPointerException, it's still much easier to fix code if the error is thrown at the earliest point at which it can be detected.

Answer 4

Using requireNonNull() as first statements in a method allow to identify right now/fast the cause of the exception.
The stacktrace indicates clearly that the exception was thrown as soon as the entry of the method because the caller didn't respect the requirements/contract. Passing a null object to another method may indeed provoke an exception at a time but the cause of the problem may be more complicated to understand as the exception will be thrown in a specific invocation on the null object that may be much further.

---

Here is a concrete and real example that shows why we have to favor fail fast in general and more particularly using Object.requireNonNull() or any way to perform a no null check on parameters designed to be not null.

Suppose a Dictionary class that composes a LookupService and a List of String representing words contained in. These fields are designed to be not null and one of these is passed in the Dictionary constructor.

Now suppose a "bad" implementation of Dictionary without null check in the method entry (here that is the constructor):

public class Dictionary {

    private final List<String> words;
    private final LookupService lookupService;

    public Dictionary(List<String> words) {
        this.words = this.words;
        this.lookupService = new LookupService(words);
    }

    public boolean isFirstElement(String userData) {
        return lookupService.isFirstElement(userData);
    }        
}


public class LookupService {

    List<String> words;

    public LookupService(List<String> words) {
        this.words = words;
    }

    public boolean isFirstElement(String userData) {
        return words.get(0).contains(userData);
    }
}

Now, let's invoke the Dictionary constructor with a null reference for the words parameter :

Dictionary dictionary = new Dictionary(null); 

// exception thrown lately : only in the next statement
boolean isFirstElement = dictionary.isFirstElement("anyThing");

The JVM throws the NPE at this statement :

return words.get(0).contains(userData); 

Exception in thread "main" java.lang.NullPointerException
    at LookupService.isFirstElement(LookupService.java:5)
    at Dictionary.isFirstElement(Dictionary.java:15)
    at Dictionary.main(Dictionary.java:22)

The exception is triggered in the LookupService class while the origin of it is well earlier (the Dictionary constructor). It makes the overall issue analysis much less obvious.
Is words null? Is words.get(0) null ? Both ? Why the one, the other or maybe both are null ? Is it a coding error in Dictionary (constructor? invoked method?) ? Is it a coding error in LookupService ? (constructor? invoked method?) ?
Finally, we will have to inspect more code to find the error origin and in a more complex class maybe even use a debugger to understand more easily what it happened.
But why a simple thing (a lack of null check) become a complex issue ?
Because we allowed the initial bug/lack identifiable on a specific component leak on lower components.
Imagine that LookupService was not a local service but a remote service or a third party library with few debugging information or imagine that you didn't have 2 layers but 4 or 5 layers of object invocations before that the null be detected ? The problem would be still more complex to analyse.

So the way to favor is :

public Dictionary(List<String> words) {
    this.words = Objects.requireNonNull(words);
    this.lookupService = new LookupService(words);
}

In this way, no headache : we get the exception thrown as soon as this is received :

// exception thrown early : in the constructor 
Dictionary dictionary = new Dictionary(null);

// we never arrive here
boolean isFirstElement = dictionary.isFirstElement("anyThing");

Exception in thread "main" java.lang.NullPointerException
    at java.util.Objects.requireNonNull(Objects.java:203)
    at com.Dictionary.(Dictionary.java:15)
    at com.Dictionary.main(Dictionary.java:24)

Note that here I illustrated the issue with a constructor but a method invocation could have the same non null check constraint.

Answer 5

NPE (Null Pointer Exception) is thrown when you access a member of an object which is null at a later point. Objects.requireNonNull() backtracks from where the value was accessed to where it was initialized as null, thus allowing to focus on real source of NPE.

A simplest scenario:

// Global String
String nullString = null;

Calling the null value:

public void useString() {
    nullString.length();  // The source of exception, without using Objects.requireNonNull()
}

Now if we use Objects.requireNonNull() during initialization:

// Global String
String nullString = Objects.requireNonNull(null); // The source of exception which is indeed the actual source of NPE

Answer 6

As a side note, this fail fast before Object#requireNotNull was implemented slightly different before java-9 inside some of the jre classes themselves. Suppose the case :

 Consumer<String> consumer = System.out::println;

In java-8 this compiles as (only the relevant parts)

getstatic Field java/lang/System.out
invokevirtual java/lang/Object.getClass

Basically an operation as : yourReference.getClass - which would fail if yourRefercence is null.

Things have changed in jdk-9 where the same code compiles as

getstatic Field java/lang/System.out
invokestatic java/util/Objects.requireNonNull

Or basically Objects.requireNotNull (yourReference)

Answer 7

Apart from the other answers - to me use of requireNonNull could make a code a bit convenient, (and sometimes easy to read)

For example - lets examine the code below,

private int calculateStringLength(String input) {
        return Objects.
                requireNonNull(input, "input cannot be null").
                length();
    }

This code returns length of the string passed to it as argument - however it will throw an NPE if input is null.

As you can see, with use of requireNonNull - there’s no reason to perform null checks manually anymore.

The other useful thing is the "exception message" is my own hand-written (input cannot be null in this case).

Answer 8

There is also the added benefit that static analysis tools generally know @NonNull and @Nullable for standard library (which is the same as checking for null explicitly) :

public static <T> @NonNull T requireNonNull(@Nullable T obj) {
    if (obj == null)
        throw new NullPointerException();
    return obj;
}

Answer 9

The basic usage is checking and throwing NullPointerException immediately.

One better alternative (shortcut) to cater to the same requirement is @NonNull annotation by lombok.

Answer 10

I think it should be used in copy constructors and some other cases like DI whose input parameter is an object, you should check if the parameter is null. In such circumstances, you can use this static method conveniently.

Answer 11

In the context of compiler extensions that implement Nullability checking (eg: uber/NullAway), Objects.requireNonNull should be used somewhat sparingly for occasions when you have a nullable field that you happen to know is not null at a certain point in your code.

In this way, there are two main usages:

• Validation

  • already covered by other responses here
  • runtime check with overhead and potential NPE

• Nullability marking (changing from @Nullable to @Nonnull)

  • minimal use of runtime check in favor of compile-time checking
  • only works when annotations are correct (enforced by the compiler)

Example usage of Nullability marking:

@Nullable
Foo getFoo(boolean getNull) { return getNull ? null : new Foo(); }

// Changes contract from Nullable to Nonnull without compiler error
@Nonnull Foo myFoo = Objects.requireNonNull(getFoo(false));

Answer 12

In addition to all the correct answers:

We use it in reactive streams. Usually the resulting NullpointerExceptions are wrapped into other Exceptions depending on their occurance in the stream. Hence, we can later easily decide how to handle the error.

Just an example: Imagine you have

<T> T parseAndValidate(String payload) throws ParsingException { ... };
<T> T save(T t) throws DBAccessException { ... };

where parseAndValidate wrapps the NullPointerException from requireNonNull in a ParsingException.

Now you can decide, e.g. when to do a retry or not:

...
.map(this::parseAndValidate)
.map(this::save)
.retry(Retry.<T>allBut(ParsingException.class))

Without the check, the NullPointerException will occure in the save method, which will result in endless retries. Even worth, imagine a long living subscription, adding

.onErrorContinue(
    throwable -> throwable.getClass().equals(ParsingException.class),
    parsingExceptionConsumer()
)

Now the RetryExhaustException will kill your subscription.