New language features since Java 8 to 14

Enhancements to the Java language you should know

Last updated on 2020/03/31 to include changes up to JDK 14.

When Java 8 introduced Streams and Lambdas it was a big change, enabling functional programming style to be expressed with much less boilerplate. While recent versions did not add such impactful features, lots of smaller improvements were made to the language.

This post summarizes language enhancements included in Java versions released after Java 8. For an overview of all the JEPs shaping the new platform check this post.

Switch Expressions

Available since: JDK 14 (Preview in JDK 12 JDK 13)

The good old switch got a facelift in Java 14. While Java keeps supporting the old switch statement, it adds the new switch expression to the language:

int numLetters = switch (day) {
    case MONDAY, FRIDAY, SUNDAY -> 6;
    case TUESDAY                -> 7;
    default      -> {
        String s = day.toString();
        int result = s.length();
        yield result;
    }
};

The most striking difference is that this new form can be used as an expression. It can be used to populate variables as demonstrated in the example above, and it can be used wherever an expression is accepted:

int k = 3;
System.out.println(
    switch (k) {
        case  1 -> "one";
        case  2 -> "two";
        default -> "many";
    }
);

However, there are some other, more subtle differences between switch expressions and switch statements.

First, for switch expressions cases don’t fall-through. So no more subtle bugs caused by missing breaks. To avoid the need for fall-through, multiple constants can be specified for each case in a comma separated list.

Second, each case has its own scope.

switch (k) {
    case  1 -> {
        String temp = "one";
        return temp.length();
    }
    case  2 -> {
        String temp = "two";
        return temp.length();
    }
    default -> "many";
}

Just for comparison, a similar piece of code would result in a compile error, because the block for a switch case is threated as a single scope and we would attempt to define temp twice.

Third, cases of a switch expression are exhaustive. This means that for String, primitive types and their wrappers the default case always has to be defined.

int k = 3;
switch (k) {
    case  1 -> "one";
    case  2 -> "two";
    default -> "many";
}

For enums either a default case has to be present, or all cases have to be explicitly covered. Relying on the latter is quite nice to ensure that all values are considered. Adding an extra value to the enum will result in a compile error for all switch expressions where it’s used.

enum Day {
   MONDAY, TUESDAY, WEDNESDAY, THURSDAY, FRIDAY, SATURDAY, SUNDAY
}

Day day = Day.TUESDAY;
switch (day) {
    case  MONDAY -> ":(";
    case  TUESDAY, WEDNESDAY, THURSDAY -> ":|";
    case  FRIDAY -> ":)";
    case  SATURDAY, SUNDAY -> ":D";
}

For all these reasons preferring switch expressions over switch statements can lead to more maintainable code.

Shameless plug
We write articles like this regularly. Join our mailing list and let's keep in touch.

⚠️ Gotcha: use arrow syntax

Switch expression can not only used with the lambda-like arrow-form cases. The old switch statement with its colon-form cases can also be used as an expression using yield:

int result = switch (s) {
    case "foo":
    case "bar":
        yield 2;
    default:
        yield 3;
};

This version can also be used as an expression, but it’s more similar to the old switch statement because

  • cases fall through
  • cases share the same scope

My advice? Don’t use this form, use switch expressions with the arrow syntax instead to get all the benefits.

Helpful NullPointerExceptions

Available since: JDK 14 (enabled with -XX:+ShowCodeDetailsInExceptionMessages)

This little gem is not really a language feature, but it’s so nice that I wanted to include it in this list.

Traditionally, experiencing a NullPointerException was like this:

node.getElementsByTagName("name").item(0).getChildNodes().item(0).getNodeValue();

Exception in thread "main" java.lang.NullPointerException
        at Unlucky.method(Unlucky.java:83)

From the exception it’s not obvious which method returned null in this case. For this reason many developers used to spread such statements over multiple lines to make sure they’ll be able to figure out which step led to the exception.

From Java 14, there’s no need to do that because NPE’s describe which part was null in the statement:

Exception in thread "main" java.lang.NullPointerException:
  Cannot invoke "org.w3c.dom.Node.getChildNodes()" because
  the return value of "org.w3c.dom.NodeList.item(int)" is null
        at Unlucky.method(Unlucky.java:83)

(Check the example project on GitHub)

The detailed message contains the action that could not be performed (Cannot invoke getChildNodes) and the reason for the failure (item(int) is null), making it much easier to find the exact source of the problem.

So overall this feature is good for debugging, and also good for code readability as there’s one less reason to sacrifice it for a technical reason.

The Helpful NullPointerExceptions extension is implemented in the JVM so you get the same benefits for code compiled with older Java versions, and when using other JVM languages, such as Scala or Kotlin.

Note, that not all NPEs get this extra info, just the ones that are created and thrown by the JVM upon:

  • reading or writing a field on null
  • invoking method on null
  • accessing or assigning an element of an array (indices are not part of the error message)
  • unboxing null

Also note that this feature does not support serialization. For example, when an NPE is thrown on the remote code executed via RMI, the exception will not include the helpful message.

Currently the Helpful NullPointerExceptions are disabled by default, and have to be enabled with the -XX:+ShowCodeDetailsInExceptionMessages flag.

⚠️ Gotcha: Check your tooling

According to the JEP Helpful NullPointerExceptions are planned to be enabled by default in the future. The fact that they are not gives users and tool vendors some time to prepare for this feature.

Make sure to check your application and infrastructure to ensure:

  • sensitive variable names not end up in log files and web server responses
  • log parsing tools can handle the new message format
  • the additional overhead required to construct the additional details is okay

Local-Variable Type Inference

Available since: JDK 11 (Without lambda support in JDK 10)

Probably the most significant language improvement since Java 8 is the addition of the var keyword. It was initially introduced in Java 10, and was further improved in Java 11.

This feature allows us to reduce the ceremony of a local variable declaration by omitting the explicit type specification:

var greetingMessage = "Hello!";

While it looks similar to Javascript’s var keyword, this is not about dynamic typing.

Take this quote from the JEP:

We seek to improve the developer experience by reducing the ceremony associated with writing Java code, while maintaining Java’s commitment to static type safety.

The type of the declared variables is inferred at compile time. In the example above the inferred type is String. Using var instead of an explicit type makes this piece of code less redundant, thus, easier to read.

Here’s another good candidate for type inference:

MyAwesomeClass awesome = new MyAwesomeClass();

It’s clear that in many cases this feature can improve code quality. However, sometimes it’s better to stick with the explicit type declaration. Let’s see a few examples where replacing a type declaration with var can backfire.

⚠️ Gotcha: Keep readability in mind

The first case is when removing explicit type information from the source code makes it less readable.

Of course, IDEs can help in this regard, but during code-reviews or when you just quickly scanning the code it might hurt readability. For example, consider factories or builders: you have to find the code responsible for object initialization to determine the type.

Here’s a little puzzle. The following piece of code is using Java 8’s Date/Time API. Guess the types of the variables in the following snippet:

var date = LocalDate.parse("2019-08-13");
var dayOfWeek = date.getDayOfWeek();
var dayOfMonth = date.getDayOfMonth();

Done? Here’s the solution:

The first one is pretty intuitive, the parse method returns a LocalDate object. However, for the next two, you should be a little bit more familiar with the API: dayOfWeek returns a java.time.DayOfWeek, while dayOfMonth simply returns an int.

Another potential problem is that with var the reader has to rely more on the context. Consider the following:

private void longerMethod() {
    // ...
    // ...
    // ...

    var dayOfWeek = date.getDayOfWeek();

    // ...
    // ...
    // ...
}

Based on the previous example, I bet you’d guess it’s a java.time.DayOfWeek. But this time, it’s an integer, because the date in this example is from Joda time. It’s a different API, behaving slightly differently, but you can’t see it because it’s a longer method, and you did not read all the lines. (JavaDoc: Joda time / Java 8 Date/Time API)

If the explicit type declaration was present, figuring out what type dayOfWeek has would be trivial. Now, with var, the reader first has to find out the type of the date variable and check what getDayOfWeek does. This is simple with an IDE, not so simple when just scanning the code.

⚠️ Gotcha: Pay attention to preserve important type information

The second case is when using var removes all available type information, so it can not be even inferred. In most cases, these situations are caught by the Java compiler. For example, var cannot infer type for lambdas or method references, because for these features the compiler relies on the left-hand side expression to figure out the types.

However, there are a few exceptions. For example, var does not play nicely with the Diamond Operator. The Diamond operator is a nice feature to remove some verbosity from the right-hand side of an expression when creating a generic instance:

Map<String, String> myMap = new HashMap<String, String>(); // Pre Java 7
Map<String, String> myMap = new HashMap<>(); // Using Diamond operator

Because it only deals with the generic types, there is still redundancy to be removed. Let’s try to make it terser with var:

var myMap = new HashMap<>();

This example is valid, and Java 11 it does not even emit in compiler warnings about it. However, with all these type inference we ended up not specifying the generic types at all, and the type will be Map<Object, Object>.

Of course, this can be solved easily by removing the Diamond Operator:

var myMap = new HashMap<String, String>();

Another set of problems can arise when var is used with primitive data types:

byte   b = 1;
short  s = 1;
int    i = 1;
long   l = 1;
float  f = 1;
double d = 1;

Without explicit type declaration, the type of all these variables would be inferred to int. Use type literals (e.g. 1L) when working with primitive data types, or don’t use var in this case at all.

⚠️ Gotcha: Make sure to read the official style guides

It’s ultimately up to you to decide when to use type inference and make sure that it does not hurt readability and correctness. As a rule of thumb, sticking to good programming practices, such as good naming and minimizing the scope of local variables certainly helps a lot. Make sure to read the official style guide and FAQ about var.

Because var has so many gotchas, it’s great that it was introduced conservatively and can only be used on local variables, which scope is usually pretty limited.

Also, it has been introduced cautiously, var is not a new keyword but a reserved type name. This means that it only has special meaning when it’s used as a type name, everywhere else var is continuing to be a valid identifier.

Currently, var does not have an immutable counterpart (such as val or const) to declare a final variable and infer its type with a single keyword. Hopefully, we’ll get it in a future release, until then, we can resort to final var.

Resources:

Allow private methods in interfaces

Available since: JDK 9 (Milling Project Coin)

Since Java 8 it is possible to add default methods to interfaces. With Java 9 these default methods can even call private methods to share code in case you are in a need for reuse, but do not want to expose functionality publicly.

Although it’s not a huge deal, it’s a logical addition that allows to tidy up code in default methods.

Diamond operator for anonymous inner classes

Available since: JDK 9 (Milling Project Coin)

Java 7 introduced the Diamond Operator (<>) to reduce verbosity by letting the compiler infer the parameter types for constructors:

List<Integer> numbers = new ArrayList<>();

However, this feature did not work with anonymous inner classes before. According to the discussion on the project’s mailing list this was not added as part of the original Diamond Operator feature, because it required a substantial JVM change.

With Java 9, this small rough edge is removed, making the operator more universally applicable:

List<Integer> numbers = new ArrayList<>() {
    // ...
}

Allow effectively-final variables to be used as resources in the try-with-resources statement

Available since: JDK 9 (Milling Project Coin)

Another enhancement introduced by Java 7 is the try-with-resources, which frees the developer from having to worry about releasing resources.

To illustrate its power, first consider the effort made to properly close a resource in this typical pre-Java 7 example:

BufferedReader br = new BufferedReader(...);
try {
    return br.readLine();
} finally {
    if (br != null) {
        br.close();
    }
}

With try-with-resources resources can be automatically released, with much less ceremony:

try (BufferedReader br = new BufferedReader(...)) {
    return br.readLine();
}

Despite its power, try-with-resources had a few shortcomings that Java 9 addressed.

Although this construct can handle multiple resources, it can easily make the code harder to read. Declaring variables like this in a list after the try keyword is a bit unconventional compared to the usual Java code:

try (BufferedReader br1 = new BufferedReader(...);
    BufferedReader br2 = new BufferedReader(...)) {
    System.out.println(br1.readLine() + br2.readLine());
}

Also, in the Java 7 version, if you already have a variable that you want to handle with this construct, you had to introduce a dummy variable. (For an example, see JDK-8068948.)

To mitigate these criticisms, try-with-resources was enhanced to handle final or effectively final local variables in addition to newly created ones:

BufferedReader br1 = new BufferedReader(...);
BufferedReader br2 = new BufferedReader(...);
try (br1; br2) {
    System.out.println(br1.readLine() + br2.readLine());
}

In this example, the initialization of the variables is separated from their registration to the try-with-resources construct.

⚠️ Gotcha: Watch out for released resources

One caveat to keep in mind is that now it’s possible to reference variables that are already released by try-with-resources, which, in most cases will fail:

BufferedReader br = new BufferedReader(...);
try (br) {
    System.out.println(br.readLine());
}
br.readLine(); // Boom!

Underscore is no longer a valid identifier name.

Available since: JDK 9 (Milling Project Coin)

In Java 8, the compiler emits a warning when ‘_’ is used as an identifier. Java 9 took this a step further making the sole underscore character illegal as an identifier, reserving this name to have special semantics in the future:

int _ = 10; // Compile error

Improved Warnings

Available since: JDK 9

Finally, let’s say a word about the changes related to the compiler warnings in newer Java versions.

Now it’s possible to annotate a private method with @SafeVarargs to mark a Type safety: Potential heap pollution via varargs parameter warning false positive. (In fact, this change is part of the previously discussed JEP 213: Milling Project Coin). Read more about Varargs, Generics and the potential probems that can arise by combining these features in the official documentation.

Also since Java 9, the compiler does not issue a warning for import statements when a deprecated type is imported. These warnings were uninformative and redundant since a separate warning is always displayed at the actual usage of the deprecated members.

What’s next: Preview features in Java 14

Java 14 has 3 preview features that can be enabled with the --enable-preview -source 14 flags. Most likely they are the next improvements to the Java language. Here’s short teaser.

Records

Records introduce a new type declaration to the language, providing compact syntax to create data classes. Instead of the usual ceremony with private fields, getters, setters and constructors, it allows us to use a terse definition to create data structures:

record Point(int x, int y) { }

Personally I can’t wait for this to become a standard feature supported by popular libraries and frameworks!

Text Blocks

Text blocks are a multi-line string literals in Java.

Instead of doing this:

String html = "";
html += "<html>";
html += "  <body>";
html += "    <p>Hello, world</p>";
html += "  </body>";
html += "</html>";

It allows us to do this:

String html = """
          <html>
            <body>
              <p>Hello, world</p>
            </body>
          </html>
          """;

Pattern Matching for instanceof

In most cases, instanceof is usually followed by a cast:

if (obj instanceof String) {
    String s = (String) obj;
    // use s
}

JEP 305 extends instanceof to make this typical scenario a bit less verbose:

if (obj instanceof String s) {
    // use s
}

Summary

This post covered the improvements related to the Java language since Java 8. It’s important to keep an eye on the Java platform, as with the new rapid release cadence a new Java version is released every six months, delivering changes to the platform and to the language.

Last updated on 31 March 2020