Java SE 8: Why Should I Care?
Like this article? We recommend
Like this article? We recommend
Like this article? We recommend
Java SE 8 may be one of the most profound Java releases in its 14-year history. When Java was released in 1996 and adopted by the masses in the form of applets (and later Servlets, JSP, and even user interfaces), the consumer computer world was a different place. Most of us had a single CPU in our computers with a single core. Java worked well on a single CPU and, as time and technology evolved, we developed ways to manage concurrency. Java was multi-threaded from its inception, and it implemented a lock-based synchronization strategy for managing access to shared data.
But that was then, this is now. Most of us have multiple CPUs with multiple cores on our desktops, and our servers have even more power. Furthermore, with the advent of the cloud, distributed and parallel computing have grown at an unprecedented rate. In order to stay current with the times, parallel computing requires a paradigm shift away from traditional object-oriented programming to a more functional model. Thus, in the past few years we have seen the evolution of new functional programming languages like Scala and Haskel, as well as the reemergence of older functional programming languages like Erlang.
So what is to become of Java? Java has had a rich history, but if it is to continue to be the language of choice of the mainstream commercial world, it must evolve. And with Java 8, evolve it has!
This article is not exhaustive in its review of the new features added to Java SE 8 (you can read a more exhaustive article series here), but it highlights the specific subset of features that modernize the Java programming language with functional programming concepts to allow it meet the demands of modern computing. In other words, if you are going to stay relevant as a Java programmer in the coming years, then these are the new features in Java that you need to understand.
Java Methods and Functions
We are familiar with passing variables and values to methods, but Java 8 permits methods and functions to be passed to methods as well. This allows you to write a generic method and pass in code that allows it to build a specific result. For example, consider a list of cars. If we want to filter the list of cars to include only coupes or only sedans, we can define methods that perform this check for us. Listing 1 shows a Car class with two static methods that determine if a car is a coupe or a sedan.
Listing 1. Car.java
public class Car {
public static Boolean isSedan( Car car ) {
return car.getType().equals( "sedan" );
}
public static Boolean isCoupe( Car car ) {
return car.getType().equals( "coupe" );
}
}
Listing 1 shows a snippet from the Car class that contains two methods which, when provided with a Car, interpret the type: isSedan() returns true if the type is a “sedan” and isCoupe() returns true if the type is a “coupe”. Listing 2 shows the filter() method from a CarInventory class that manages a list of cars.
Listing 2. CarInventory.java
public class CarInventory {
public List<Car> filter( Predicate<Car> p ) {
List<Car> results = new ArrayList<Car>();
for( Car car : carList ) {
if( p.test( car ) ) {
results.add( car );
}
}
return results;
}
}
The filter() method iterates over a list of cars and builds a result that matches the predicate condition. This probably leads to the question, what is a Predicate? The Predicate interface is defined in the java.util.function package and looks similar to Listing 3.
Listing 3. Predicate.java
public interface Predicate<T> {
public boolean test( T t );
}
A predicate, in mathematics, is a function that accepts a value and returns true or false. Predicate<Car> could likewise have been written as Function<Car,Boolean>, but Predicate<Car> is more succinct.
Finally, Listing 4 shows how you can pass the correct Car method to the CarInventory filter() method.
Listing 4. Using the filter() method
CarInventory carInventory = new CarInventory(); List<Car> coupes = carInventory.filter( Car::isCoupe ); List<Car> sedans = carInventory.filter( Car::isSedan );
Using the “::” operator, we are able to pass a method to the filter() method and, as we say in Listing 2, that method will be executed inside the filter() method itself.
It does become burdensome to write static methods in our classes just to be passed as predicates, so we have the option to instead create anonymous functions, which are also called lambdas. Lambdas, in general are defined as follows:
( Input values ) -> Expression that optionally produces a response
Given an input, a lambda can do something, which may produce an output. Predicates are special types of lambdas that are of the following form:
( Input values ) -> Expression that evaluates to a boolean
For example, we could retrieve our coupes and sedans as follows:
List<Car> coupes = carInventory.filter( ( Car c ) -> c.getType().equals( "coupe" ) ); List<Car> sedans = carInventory.filter( ( Car c ) -> c.getType().equals( "sedan" ) );
These expressions read as follows: Given a Car c, return true of the type is a “coupe” (or “sedan”). This is functionally equivalent to passing the method (Car::isCoupe).
Streams
Anonymous functions, or lambda expressions, are nice, but they were included in Java 8 for more than syntactic eye candy. In order to better facilitate parallel processing, Java 8 introduced the Streams API, which we will see works hand-in-hand with lambda expressions.
The Streams API allows you to connect multiple methods together such that the output from one method serves as the input to the next method. Furthermore, one method does not have to complete before its output can be used by the next method in the stream. Consider how streams work on a simple Linux command line:
ls –l | grep txt
ls –l retrieves a list of filenames in the current directory, and then grep txt only shows files that have the string “txt” in their name. ls –l returns filenames one at a time, so if the first file is “file1.txt” then the grep command will process that filename potentially before ls –l returns the second filename. The Streams API follows this model and, if you allow it to, it can execute operations in parallel. For example, if it is performing operations against a collection of elements, it could process more than one record at a time.
Because Java applications frequently operate on collections of data, Streams are intimately connected to Collection classes. Two new methods have been added to the Collection APIs:
- stream(): Creates a Stream object that can be used to operate on the collection.
- parallelStream(): Creates a Stream object that can be used to operate on the collection in parallel.
With a Stream in hand, you can execute one of the following methods (the following is a subset of methods that I find most interesting), passing it a lambda expression:
- filter(): Only passes values that match the supplied predicate to the next stream.
- distinct(): Ensures that all values in the stream are distinct; in other words if “apple” appears twice, only one “apple” will be passed to the next stream.
- limit(): Only passes the first n elements to the next stream; for example, limit(3) would only pass the first three elements to the next stream.
- sorted(): Sorts the items in the stream into their natural order.
- max()/min(): Returns the maximum or minimum element in the stream.
- forEach(): Does not return a stream, but instead allows you to perform an operation on every element in the stream.
- collect(): Ends the stream processing and returns the completed stream in a more consumable fashion, such as a List.
With this description, we could rewrite our coupe/sedan search as follows:
List<Car> cars = new ArrayList<Car>();
// Add cars to the list...
List<Car> coupes = cars.stream().filter(( Car c ) -> c.getType().equals( "coupe" ) )
.collect( toList() );
List<Car> sedans = cars.stream().filter(( Car c ) -> c.getType().equals( "sedan" ) )
.collect( toList() );
The stream() method converts the List to a Stream (or more specifically, it provides Stream access to the List), the filter() method accepts the predicate that compares the car type to the string “coupe” or “sedan”, and finally the collect() method converts the result to a List. Likewise, if we wanted to find all coupes, but perform the operation in parallel, we could do so with the following command:
List<Car> coupes = cars.parallelStream().filter(( Car c ) -> c.getType().equals( "coupe" ) )
.collect( toList() );
parallelStream() provides a Stream that can read the collection, but the JVM can now execute the filter in parallel (on multiple cores on multiple CPUs) and then collect the results into a single list. If our cars List had millions of cars in it, a parallel stream could process the list much faster than a standard stream. The Streams API has the option to create as many threads as it deems necessary and partition the cars into sub-lists for parallel processing. And, as a Java programmer, you are able to gain this level of parallel processing by invoking the parallelStream() method instead of the stream() method. Think about how complicated the code that you would have to write would be to partition the list into sub-lists, create multiple threads and assign each thread a sub-list to process, and then correlate the results into a single response. I hope you can appreciate the value that the Streams API offers.
Returning to our example, if we want to get a little crazy, let’s combine some of the Stream methods to return the five least expensive red coupes:
List<Car> coupes = cars.parallelStream().filter(( Car c ) -> c.getType().equals( "coupe" ) )
.filter( ( Car c ) -> c.getColor().equals( "red" ) )
.sorted( comparing( Car::getPrice ) )
.limit( 5 )
.collect( toList() );
The first filter returns only coupes, and the second filter returns only red cars (and yes, you could combine both of these filters into one predicate). Next, we sort the stream by price. The Comparator class now has a static comparing() method to which we can pass a function. In this case, we pass a reference to the Car class’s getPrice() method. The natural sorting order for numbers is lowest to highest, so this will sort the cars by ascending price. Next, we invoke limit(5), which returns only the first five elements in the stream (again, sorted by ascending price). Finally, we call collect() to build a List that contains our five cars.
You may have noticed that Streams allow you to manipulate collections in a declarative way, or in other words, they allow you to define the type of operation to perform on the collection without having to write all of the plumbing code to make it work. Furthermore, when the Streams API is used in parallel, it not only dramatically improves performance, but also removes some very complicated plumbing code!
This section only touched on the Streams API, but hopefully it has whet your appetite to learn more.
Static Interface Methods and Default Methods
Before leaving this introduction to Java 8, I thought it important to review two additional features that enabled the developers at Oracle to update the collection APIs without breaking a horde of existing code. In addition to the standard collection classes, various developers have built implementations of the collection classes, but adhered to the interfaces to ensure compatibility. Should Oracle have required all of them to update their code to add the new stream() and parallelStream() methods? In Java 7, they would have no choice. So in Java 8, Oracle added the following two capabilities:
- Static interface methods
- Default interface methods
Java 8 allows you to implement static methods in your interfaces. In Java 7, all method implementations, static or not static, needed to be implemented in classes. Now you are free to implement static methods in interfaces.
Similarly, interfaces are now able to implement default methods, using the new default keyword. For example, the Collection<E> interface that all collection classes implement (ArrayList, TreeSet, etc.) defines a new default method called stream() that returns “a sequential Stream over the elements in this collection.” This means that any collection class that implements the Collection interface can now be used through the Streams API framework. The stream() method is defined as follows:
public interface Collection<E> {
...
default Stream<E> stream() {
// Build the stream...
}
}
In addition to defining methods that collection classes must implement, the Collection interface was able to build the implementation of the stream() method for those classes. Classes that implement the Collection interface are free to override the stream() method, but if they do not, then the default implementation will be used. Whether or not you will leverage this feature in your code has yet to be seen, but it is what empowered Oracle to make changes without breaking existing code.
Summary
As computer hardware has evolved, functional programming languages have slowly replaced object-oriented programming languages because of their ability to operate on multiple CPUs and multiple cores. Java has a rich history, and has become the dominant choice of commercial enterprises, but if it does not evolve then it can be replaced. Fortunately, when designing Java 8, Oracle saw this disparity and added functional programming capabilities to Java.
This article provided a high-level overview of some of the more important features introduced in Java 8 that support functional programming paradigms. Specifically this article reviewed:
- Passing functions to methods, as well as defining anonymous functions (lambdas) and passing those to methods
- The Streams API, which is used to perform parallel operations without requiring the programmer to write complicated threading code
- Static and default interface methods, which allow developers to provide default implementations of methods in interfaces
This article has been only a small introduction. For more details about the specific features in Java 8, refer to this article series Java SE 8's New Language Features.