A Stream is a sequence of elements upon which sequential and parallel aggregate operations can be performed. Any given Stream can potentially have an unlimited amount of data flowing through it. As a result, data received from a Stream is processed individually as it arrives, as opposed to performing batch processing on the data altogether. When combined with lambda expressions they provide a concise way to perform operations on sequences of data using a functional approach.

Example: (see it work on Ideone)

Stream<String> fruitStream = Stream.of("apple", "banana", "pear", "kiwi", "orange");

fruitStream.filter(s -> s.contains("a"))
           .map(String::toUpperCase)
           .sorted()
           .forEach(System.out::println);

Output:

APPLE
BANANA
ORANGE
PEAR

The operations performed by the above code can be summarized as follows:

1. Create a Stream<String> containing a sequenced ordered Stream of fruit String elements using the static factory method Stream.of(values).

2. The filter() operation retains only elements that match a given predicate (the elements that when tested by the predicate return true). In this case, it retains the elements containing an "a". The predicate is given as a lambda expression.

3. The map() operation transforms each element using a given function, called a mapper. In this case, each fruit String is mapped to its uppercase String version using the method-reference String::toUppercase.

   Note that the map() operation will return a stream with a different generic type if the mapping function returns a type different to its input parameter. For example on a Stream<String> calling .map(String::isEmpty) returns a Stream<Boolean>

4. The sorted() operation sorts the elements of the Stream according to their natural ordering (lexicographically, in the case of String).

5. Finally, the forEach(action) operation performs an action which acts on each element of the Stream, passing it to a Consumer. In the example, each element is simply being printed to the console. This operation is a terminal operation, thus making it impossible to operate on it again.

   Note that operations defined on the Stream are performed because of the terminal operation. Without a terminal operation, the stream is not processed. Streams can not be reused. Once a terminal operation is called, the Stream object becomes unusable.

Operations (as seen above) are chained together to form what can be seen as a query on the data.

Closing Streams

Note that a Stream generally does not have to be closed. It is only required to close streams that operate on IO channels. Most Stream types don't operate on resources and therefore don't require closing.

The Stream interface extends AutoCloseable. Streams can be closed by calling the close method or by using try-with-resource statements.

An example use case where a Stream should be closed is when you create a Stream of lines from a file:

try (Stream<String> lines = Files.lines(Paths.get("somePath"))) {
    lines.forEach(System.out::println);
}

The Stream interface also declares the Stream.onClose() method which allows you to register Runnable handlers which will be called when the stream is closed. An example use case is where code which produces a stream needs to know when it is consumed to perform some cleanup.

public Stream<String>streamAndDelete(Path path) throws IOException {
    return Files.lines(path).onClose(() -> someClass.deletePath(path));
}

The run handler will only execute if the close() method gets called, either explicitly or implicitly by a try-with-resources statement.

Processing Order

A Stream object's processing can be sequential or parallel.

In a sequential mode, the elements are processed in the order of the source of the Stream. If the Stream is ordered (such as a SortedMap implementation or a List) the processing is guaranteed to match the ordering of the source. In other cases, however, care should be taken not to depend on the ordering (see: is the Java HashMap keySet() iteration order consistent?).

Example:

List<Integer> integerList = Arrays.asList(0, 1, 2, 3, 42); 

// sequential 
long howManyOddNumbers = integerList.stream()
                                    .filter(e -> (e % 2) == 1)
                                    .count(); 

System.out.println(howManyOddNumbers); // Output: 2

Live on Ideone

Parallel mode allows the use of multiple threads on multiple cores but there is no guarantee of the order in which elements are processed.

If multiple methods are called on a sequential Stream, not every method has to be invoked. For example, if a Stream is filtered and the number of elements is reduced to one, a subsequent call to a method such as sort will not occur. This can increase the performance of a sequential Stream — an optimization that is not possible with a parallel Stream.

Example:

// parallel
long howManyOddNumbersParallel = integerList.parallelStream()
                                            .filter(e -> (e % 2) == 1)
                                            .count();

System.out.println(howManyOddNumbersParallel); // Output: 2

Live on Ideone

Differences from Containers (or Collections)

While some actions can be performed on both Containers and Streams, they ultimately serve different purposes and support different operations. Containers are more focused on how the elements are stored and how those elements can be accessed efficiently. A Stream, on the other hand, doesn't provide direct access and manipulation to its elements; it is more dedicated to the group of objects as a collective entity and performing operations on that entity as a whole. Stream and Collection are separate high-level abstractions for these differing purposes.

Collect with toList() and toSet()

Elements from a Stream can be easily collected into a container by using the Stream.collect operation:

System.out.println(Arrays
    .asList("apple", "banana", "pear", "kiwi", "orange")
    .stream()
    .filter(s -> s.contains("a"))
    .collect(Collectors.toList())
);
// prints: [apple, banana, pear, orange]

Other collection instances, such as a Set, can be made by using other Collectors built-in methods. For example, Collectors.toSet() collects the elements of a Stream into a Set.

Explicit control over the implementation of List or Set

According to documentation of Collectors#toList() and Collectors#toSet(), there are no guarantees on the type, mutability, serializability, or thread-safety of the List or Set returned.

For explicit control over the implementation to be returned, Collectors#toCollection(Supplier) can be used instead, where the given supplier returns a new and empty collection.

// syntax with method reference
System.out.println(strings
        .stream()
        .filter(s -> s != null && s.length() <= 3)
        .collect(Collectors.toCollection(ArrayList::new))
);

// syntax with lambda
System.out.println(strings
        .stream()
        .filter(s -> s != null && s.length() <= 3)
        .collect(Collectors.toCollection(() -> new LinkedHashSet<>()))
);

Collecting Elements using toMap

Collector accumulates elements into a Map, Where key is the Student Id and Value is Student Value.

  List<Student> students = new ArrayList<Student>(); 
    students.add(new Student(1,"test1"));
    students.add(new Student(2,"test2"));
    students.add(new Student(3,"test3"));
    
    Map<Integer, String> IdToName = students.stream()
        .collect(Collectors.toMap(Student::getId, Student::getName));
    System.out.println(IdToName);

Output :

{1=test1, 2=test2, 3=test3}

The Collectors.toMap has another implementation Collector<T, ?, Map<K,U>> toMap(Function<? super T, ? extends K> keyMapper, Function<? super T, ? extends U> valueMapper, BinaryOperator<U> mergeFunction).The mergeFunction is mostly used to select either new value or retain old value if the key is repeated when adding a new member in the Map from a list.

The mergeFunction often looks like: (s1, s2) -> s1 to retain value corresponding to the repeated key, or (s1, s2) -> s2 to put new value for the repeated key.

Collecting Elements to Map of Collections

Example: from ArrayList to Map<String, List<>>

Often it requires to make a map of list out of a primary list. Example: From a student of list, we need to make a map of list of subjects for each student.

    List<Student> list = new ArrayList<>();
    list.add(new Student("Davis", SUBJECT.MATH, 35.0));
    list.add(new Student("Davis", SUBJECT.SCIENCE, 12.9));
    list.add(new Student("Davis", SUBJECT.GEOGRAPHY, 37.0));

    list.add(new Student("Sascha", SUBJECT.ENGLISH, 85.0));
    list.add(new Student("Sascha", SUBJECT.MATH, 80.0));
    list.add(new Student("Sascha", SUBJECT.SCIENCE, 12.0));
    list.add(new Student("Sascha", SUBJECT.LITERATURE, 50.0));

    list.add(new Student("Robert", SUBJECT.LITERATURE, 12.0));

    Map<String, List<SUBJECT>> map = new HashMap<>();
    list.stream().forEach(s -> {
                map.computeIfAbsent(s.getName(), x -> new ArrayList<>()).add(s.getSubject());
            });
    System.out.println(map);

Output:

{ Robert=[LITERATURE], 
Sascha=[ENGLISH, MATH, SCIENCE, LITERATURE], 
Davis=[MATH, SCIENCE, GEOGRAPHY] }

Example: from ArrayList to Map<String, Map<>>

    List<Student> list = new ArrayList<>();
    list.add(new Student("Davis", SUBJECT.MATH, 1, 35.0));
    list.add(new Student("Davis", SUBJECT.SCIENCE, 2, 12.9));
    list.add(new Student("Davis", SUBJECT.MATH, 3, 37.0));
    list.add(new Student("Davis", SUBJECT.SCIENCE, 4, 37.0));

    list.add(new Student("Sascha", SUBJECT.ENGLISH, 5, 85.0));
    list.add(new Student("Sascha", SUBJECT.MATH, 1, 80.0));
    list.add(new Student("Sascha", SUBJECT.ENGLISH, 6, 12.0));
    list.add(new Student("Sascha", SUBJECT.MATH, 3, 50.0));

    list.add(new Student("Robert", SUBJECT.ENGLISH, 5, 12.0));

    Map<String, Map<SUBJECT, List<Double>>> map = new HashMap<>();

    list.stream().forEach(student -> {
        map.computeIfAbsent(student.getName(), s -> new HashMap<>())
                .computeIfAbsent(student.getSubject(), s -> new ArrayList<>())
                .add(student.getMarks());
    });

    System.out.println(map);

Output:

{ Robert={ENGLISH=[12.0]}, 
Sascha={MATH=[80.0, 50.0], ENGLISH=[85.0, 12.0]}, 
Davis={MATH=[35.0, 37.0], SCIENCE=[12.9, 37.0]} }

Cheat-Sheet

It is possible to generate a Stream that does not end. Calling a terminal method on an infinite Stream causes the Stream to enter an infinite loop. The limit method of a Stream can be used to limit the number of terms of the Stream that Java processes.

This example generates a Stream of all natural numbers, starting with the number 1. Each successive term of the Stream is one higher than the previous. By calling the limit method of this Stream, only the first five terms of the Stream are considered and printed.

// Generate infinite stream - 1, 2, 3, 4, 5, 6, 7, ...
IntStream naturalNumbers = IntStream.iterate(1, x -> x + 1);

// Print out only the first 5 terms
naturalNumbers.limit(5).forEach(System.out::println);

Output:

1
2
3
4
5

Another way of generating an infinite stream is using the Stream.generate method. This method takes a lambda of type Supplier.

// Generate an infinite stream of random numbers
Stream<Double> infiniteRandomNumbers = Stream.generate(Math::random);

// Print out only the first 10 random numbers
infiniteRandomNumbers.limit(10).forEach(System.out::println);

A Stream will only be traversed when there is a terminal operation, like count(), collect() or forEach(). Otherwise, no operation on the Stream will be performed.

In the following example, no terminal operation is added to the Stream, so the filter() operation will not be invoked and no output will be produced because peek() is NOT a terminal operation.

IntStream.range(1, 10).filter(a -> a % 2 == 0).peek(System.out::println);

Live on Ideone

This is a Stream sequence with a valid terminal operation, thus an output is produced.

You could also use forEach instead of peek:

IntStream.range(1, 10).filter(a -> a % 2 == 0).forEach(System.out::println); 

Live on Ideone

Output:

2
4
6
8

After the terminal operation is performed, the Stream is consumed and cannot be reused.

Although a given stream object cannot be reused, it's easy to create a reusable Iterable that delegates to a stream pipeline. This can be useful for returning a modified view of a live data set without having to collect results into a temporary structure.

List<String> list = Arrays.asList("FOO", "BAR");
Iterable<String> iterable = () -> list.stream().map(String::toLowerCase).iterator();

for (String str : iterable) {
    System.out.println(str);
}
for (String str : iterable) {
    System.out.println(str);
}

Output:

foo
bar
foo
bar

This works because Iterable declares a single abstract method Iterator<T> iterator(). That makes it effectively a functional interface, implemented by a lambda that creates a new stream on each call.

In general, a Stream operates as shown in the following image:

NOTE: Argument checks are always performed, even without a terminal operation:

try {
    IntStream.range(1, 10).filter(null);
} catch (NullPointerException e) {
    System.out.println("We got a NullPointerException as null was passed as an argument to filter()");
}

Live on Ideone

Output:

We got a NullPointerException as null was passed as an argument to filter()

The groupingBy(classifier, downstream) collector allows the collection of Stream elements into a Map by classifying each element in a group and performing a downstream operation on the elements classified in the same group.

A classic example of this principle is to use a Map to count the occurrences of elements in a Stream. In this example, the classifier is simply the identity function, which returns the element as-is. The downstream operation counts the number of equal elements, using counting().

Stream.of("apple", "orange", "banana", "apple")
      .collect(Collectors.groupingBy(Function.identity(), Collectors.counting()))
      .entrySet()
      .forEach(System.out::println);

The downstream operation is itself a collector (Collectors.counting()) that operates on elements of type String and produces a result of type Long. The result of the collect method call is a Map<String, Long>.

This would produce the following output:

banana=1
orange=1
apple=2

Note: Before deciding which Stream to use please have a look at ParallelStream vs Sequential Stream behavior.

When you want to perform Stream operations concurrently, you could use either of these ways.

List<String> data = Arrays.asList("One", "Two", "Three", "Four", "Five");
Stream<String> aParallelStream = data.stream().parallel();

Or:

Stream<String> aParallelStream = data.parallelStream();

To execute the operations defined for the parallel stream, call a terminal operator:

aParallelStream.forEach(System.out::println);

(A possible) output from the parallel Stream:

Three
Four
One
Two
Five

The order might change as all the elements are processed in parallel (Which may make it faster). Use parallelStream when ordering does not matter.

Performance impact

In case networking is involved, parallel Streams may degrade the overall performance of an application because all parallel Streams use a common fork-join thread pool for the network.

On the other hand, parallel Streams may significantly improve performance in many other cases, depending of the number of available cores in the running CPU at the moment.

You may need to convert a Stream emitting Optional to a Stream of values, emitting only values from existing Optional. (ie: without null value and not dealing with Optional.empty()).

 Optional<String> op1 = Optional.empty();
 Optional<String> op2 = Optional.of("Hello World");

 List<String> result = Stream.of(op1, op2)
                             .filter(Optional::isPresent)
                             .map(Optional::get)
                             .collect(Collectors.toList());

 System.out.println(result); //[Hello World]

All java Collection<E>s have stream() and parallelStream() methods from which a Stream<E> can be constructed:

Collection<String> stringList = new ArrayList<>();
Stream<String> stringStream = stringList.parallelStream();

A Stream<E> can be created from an array using one of two methods:

String[] values = { "aaa", "bbbb", "ddd", "cccc" };
Stream<String> stringStream = Arrays.stream(values);
Stream<String> stringStreamAlternative = Stream.of(values);

The difference between Arrays.stream() and Stream.of() is that Stream.of() has a varargs parameter, so it can be used like:

Stream<Integer> integerStream = Stream.of(1, 2, 3);

There are also primitive Streams that you can use. For example:

IntStream intStream = IntStream.of(1, 2, 3);
DoubleStream doubleStream = DoubleStream.of(1.0, 2.0, 3.0);

These primitive streams can also be constructed using the Arrays.stream() method:

IntStream intStream = Arrays.stream(new int[]{ 1, 2, 3 });

It is possible to create a Stream from an array with a specified range.

int[] values= new int[]{1, 2, 3, 4, 5};
IntStream intStram = Arrays.stream(values, 1, 3);

Note that any primitive stream can be converted to boxed type stream using the boxed method :

Stream<Integer> integerStream = intStream.boxed();

This can be useful in some case if you want to collect the data since primitive stream does not have any collect method that takes a Collector as argument.

Reusing intermediate operations of a stream chain

Stream is closed when ever terminal operation is called. Reusing the stream of intermediate operations, when only terminal operation is only varying. we could create a stream supplier to construct a new stream with all intermediate operations already set up.

Supplier<Stream<String>> streamSupplier = () -> Stream.of("apple", "banana","orange", "grapes", "melon","blueberry","blackberry")
.map(String::toUpperCase).sorted();

  streamSupplier.get().filter(s ->   s.startsWith("A")).forEach(System.out::println);

// APPLE

  streamSupplier.get().filter(s -> s.startsWith("B")).forEach(System.out::println);

  // BANANA
  // BLACKBERRY
  // BLUEBERRY

int[] arrays can be converted to List<Integer> using streams

int[] ints = {1,2,3};
List<Integer> list = IntStream.of(ints).boxed().collect(Collectors.toList());

Java 8 provides classes called IntSummaryStatistics, DoubleSummaryStatistics and LongSummaryStatistics which give a state object for collecting statistics such as count, min, max, sum, and average.

List<Integer> naturalNumbers = Arrays.asList(1, 2, 3, 4, 5, 6, 7, 8, 9, 10);
IntSummaryStatistics stats = naturalNumbers.stream()
                                           .mapToInt((x) -> x)
                                           .summaryStatistics();
System.out.println(stats);

Which will result in:

IntSummaryStatistics{count=10, sum=55, min=1, max=10, average=5.500000}

Example: Get a Stream of 30 elements, containing 21st to 50th (inclusive) element of a collection.

final long n = 20L; // the number of elements to skip
final long maxSize = 30L; // the number of elements the stream should be limited to
final Stream<T> slice = collection.stream().skip(n).limit(maxSize);

Notes:

• IllegalArgumentException is thrown if n is negative or maxSize is negative
• both skip(long) and limit(long) are intermediate operations
• if a stream contains fewer than n elements then skip(n) returns an empty stream
• both skip(long) and limit(long) are cheap operations on sequential stream pipelines, but can be quite expensive on ordered parallel pipelines

Variable declaration for examples:

Collection<String> abc = Arrays.asList("a", "b", "c");
Collection<String> digits = Arrays.asList("1", "2", "3");
Collection<String> greekAbc = Arrays.asList("alpha", "beta", "gamma");

Example 1 - Concatenate two Streams

final Stream<String> concat1 = Stream.concat(abc.stream(), digits.stream());

concat1.forEach(System.out::print);
// prints: abc123

Example 2 - Concatenate more than two Streams

final Stream<String> concat2 = Stream.concat(
    Stream.concat(abc.stream(), digits.stream()),
    greekAbc.stream());

System.out.println(concat2.collect(Collectors.joining(", ")));
// prints: a, b, c, 1, 2, 3, alpha, beta, gamma

Alternatively to simplify the nested concat() syntax the Streams can also be concatenated with flatMap():

final Stream<String> concat3 = Stream.of(
    abc.stream(), digits.stream(), greekAbc.stream())
    .flatMap(s -> s);
    // or `.flatMap(Function.identity());` (java.util.function.Function)

System.out.println(concat3.collect(Collectors.joining(", ")));
// prints: a, b, c, 1, 2, 3, alpha, beta, gamma

Be careful when constructing Streams from repeated concatenation, because accessing an element of a deeply concatenated Stream can result in deep call chains or even a StackOverflowException.

Java does not have a Char Stream, so when working with Strings and constructing a Stream of Characters, an option is to get a IntStream of code points using String.codePoints() method. So IntStream can be obtained as below:

public IntStream stringToIntStream(String in) {
  return in.codePoints();
}

It is a bit more involved to do the conversion other way around i.e. IntStreamToString. That can be done as follows:

public String intStreamToString(IntStream intStream) {
  return intStream.collect(StringBuilder::new, StringBuilder::appendCodePoint, StringBuilder::append).toString();
}

List<String> data = new ArrayList<>();
data.add("Sydney");
data.add("London");
data.add("New York");
data.add("Amsterdam");
data.add("Mumbai");
data.add("California");

System.out.println(data);

List<String> sortedData = data.stream().sorted().collect(Collectors.toList());

System.out.println(sortedData);

Output:

[Sydney, London, New York, Amsterdam, Mumbai, California]
[Amsterdam, California, London, Mumbai, New York, Sydney]

It's also possible to use different comparison mechanism as there is a overloaded sorted version which takes a comparator as its argument.

Also, you can use a lambda expression for sorting:

List<String> sortedData2 = data.stream().sorted((s1,s2) -> s2.compareTo(s1)).collect(Collectors.toList());

This would output [Sydney, New York, Mumbai, London, California, Amsterdam]

You can use Comparator.reverseOrder() to have a comparator that imposes the reverse of the natural ordering.

List<String> reverseSortedData = data.stream().sorted(Comparator.reverseOrder()).collect(Collectors.toList());

Java provides specialized Streams for three types of primitives IntStream (for ints), LongStream (for longs) and DoubleStream (for doubles). Besides being optimized implementations for their respective primitives, they also provide several specific terminal methods, typically for mathematical operations. E.g.:

IntStream is = IntStream.of(10, 20, 30);
double average = is.average().getAsDouble(); // average is 20.0

Analog to get a collection for a Stream by collect() an array can be obtained by the Stream.toArray() method:

List<String> fruits = Arrays.asList("apple", "banana", "pear", "kiwi", "orange");

String[] filteredFruits = fruits.stream()
    .filter(s -> s.contains("a"))
    .toArray(String[]::new);     

// prints: [apple, banana, pear, orange]
System.out.println(Arrays.toString(filteredFruits));

String[]::new is a special kind of method reference: a constructor reference.

It is possible to find the first element of a Stream that matches a condition.

For this example, we will find the first Integer whose square is over 50000.

IntStream.iterate(1, i -> i + 1) // Generate an infinite stream 1,2,3,4...
    .filter(i -> (i*i) > 50000) // Filter to find elements where the square is >50000
    .findFirst(); // Find the first filtered element

This expression will return an OptionalInt with the result.

Note that with an infinite Stream, Java will keep checking each element until it finds a result. With a finite Stream, if Java runs out of elements but still can't find a result, it returns an empty OptionalInt.

Streams of elements usually do not allow access to the index value of the current item. To iterate over an array or ArrayList while having access to indexes, use IntStream.range(start, endExclusive).

String[] names = { "Jon", "Darin", "Bauke", "Hans", "Marc" };

IntStream.range(0, names.length)
    .mapToObj(i -> String.format("#%d %s", i + 1, names[i]))
    .forEach(System.out::println);

The range(start, endExclusive) method returns another ÌntStream and the mapToObj(mapper) returns a stream of String.

Output:

#1 Jon
#2 Darin
#3 Bauke
#4 Hans
#5 Marc

This is very similar to using a normal for loop with a counter, but with the benefit of pipelining and parallelization:

for (int i = 0; i < names.length; i++) {
    String newName = String.format("#%d %s", i + 1, names[i]);
    System.out.println(newName);
}

A Stream of items that are in turn streamable can be flattened into a single continuous Stream:

Array of List of Items can be converted into a single List.

List<String> list1 =  Arrays.asList("one", "two");
      List<String> list2 =  Arrays.asList("three","four","five");
      List<String> list3 =  Arrays.asList("six");
          List<String> finalList = Stream.of(list1, list2, list3).flatMap(Collection::stream).collect(Collectors.toList());
System.out.println(finalList);

// [one, two, three, four, five, six]

Map containing List of Items as values can be Flattened to a Combined List

Map<String, List<Integer>> map = new LinkedHashMap<>();
map.put("a", Arrays.asList(1, 2, 3));
map.put("b", Arrays.asList(4, 5, 6));

List<Integer> allValues = map.values() // Collection<List<Integer>>
        .stream()                      // Stream<List<Integer>>
        .flatMap(List::stream)         // Stream<Integer>
        .collect(Collectors.toList());

System.out.println(allValues);
// [1, 2, 3, 4, 5, 6]

List of Map can be flattened into a single continuous Stream

List<Map<String, String>> list = new ArrayList<>();
Map<String,String> map1 = new HashMap();
map1.put("1", "one");
map1.put("2", "two");

Map<String,String> map2 = new HashMap();
map2.put("3", "three");
map2.put("4", "four");
list.add(map1);
list.add(map2);


Set<String> output= list.stream()  //  Stream<Map<String, String>>
    .map(Map::values)              // Stream<List<String>>
    .flatMap(Collection::stream)   // Stream<String>
    .collect(Collectors.toSet());  //Set<String>
// [one, two, three,four]

Simple case without duplicate keys

Stream<String> characters = Stream.of("A", "B", "C");

Map<Integer, String> map = characters
            .collect(Collectors.toMap(element -> element.hashCode(), element -> element));
// map = {65=A, 66=B, 67=C}

To make things more declarative, we can use static method in Function interface - Function.identity(). We can replace this lambda element -> element with Function.identity().

Case where there might be duplicate keys

The javadoc for Collectors.toMap states:

If the mapped keys contains duplicates (according to Object.equals(Object)), an IllegalStateException is thrown when the collection operation is performed. If the mapped keys may have duplicates, use toMap(Function, Function, BinaryOperator) instead.

Stream<String> characters = Stream.of("A", "B", "B", "C");

Map<Integer, String> map = characters
            .collect(Collectors.toMap(
                element -> element.hashCode(),
                element -> element,
                (existingVal, newVal) -> (existingVal + newVal)));

// map = {65=A, 66=BB, 67=C}

The BinaryOperator passed to Collectors.toMap(...) generates the value to be stored in the case of a collision. It can:

• return the old value, so that the first value in the stream takes precedence,
• return the new value, so that the last value in the stream takes precedence, or
• combine the old and new values

Grouping by value

You can use Collectors.groupingBy when you need to perform the equivalent of a database cascaded "group by" operation. To illustrate, the following creates a map in which people's names are mapped to surnames:

List<Person> people = Arrays.asList(
    new Person("Sam", "Rossi"),
    new Person("Sam", "Verdi"),
    new Person("John", "Bianchi"),
    new Person("John", "Rossi"),
    new Person("John", "Verdi")
);

Map<String, List<String>> map = people.stream()
        .collect(
                // function mapping input elements to keys
                Collectors.groupingBy(Person::getName, 
                // function mapping input elements to values,
                // how to store values
                Collectors.mapping(Person::getSurname, Collectors.toList()))
        );

// map = {John=[Bianchi, Rossi, Verdi], Sam=[Rossi, Verdi]}

Live on Ideone

It is sometimes useful to create random Strings, maybe as Session-ID for a web-service or an initial password after registration for an application. This can be easily achieved using Streams.

First we need to initialize a random number generator. To enhance security for the generated Strings, it is a good idea to use SecureRandom.

Note: Creating a SecureRandom is quite expensive, so it is best practice to only do this once and call one of its setSeed() methods from time to time to reseed it.

private static final SecureRandom rng = new SecureRandom(SecureRandom.generateSeed(20)); 
//20 Bytes as a seed is rather arbitrary, it is the number used in the JavaDoc example

When creating random Strings, we usually want them to use only certain characters (e.g. only letters and digits). Therefore we can create a method returning a boolean which can later be used to filter the Stream.

//returns true for all chars in 0-9, a-z and A-Z
boolean useThisCharacter(char c){
    //check for range to avoid using all unicode Letter (e.g. some chinese symbols)
    return c >= '0' && c <= 'z' && Character.isLetterOrDigit(c);
}

Next we can utilize the RNG to generate a random String of specific length containing the charset which pass our useThisCharacter check.

public String generateRandomString(long length){
    //Since there is no native CharStream, we use an IntStream instead
    //and convert it to a Stream<Character> using mapToObj.
    //We need to specify the boundaries for the int values to ensure they can safely be cast to char
    Stream<Character> randomCharStream = rng.ints(Character.MIN_CODE_POINT, Character.MAX_CODE_POINT).mapToObj(i -> (char)i).filter(c -> this::useThisCharacter).limit(length);

    //now we can use this Stream to build a String utilizing the collect method.
    String randomString = randomCharStream.collect(StringBuilder::new, StringBuilder::append, StringBuilder::append).toString();
    return randomString;
}

Streams, and especially IntStreams, are an elegant way of implementing summation terms (∑). The ranges of the Stream can be used as the bounds of the summation.

E.g., Madhava's approximation of Pi is given by the formula (Source: wikipedia):

This can be calculated with an arbitrary precision. E.g., for 101 terms:

double pi = Math.sqrt(12) * 
            IntStream.rangeClosed(0, 100)
                     .mapToDouble(k -> Math.pow(-3, -1 * k) / (2 * k + 1))
                     .sum();

Note: With double's precision, selecting an upper bound of 29 is sufficient to get a result that's indistinguishable from Math.Pi.

Method references make excellent self-documenting code, and using method references with Streams makes complicated processes simple to read and understand. Consider the following code:

public interface Ordered {
    default int getOrder(){
        return 0;
    }
}

public interface Valued<V extends Ordered> {
    boolean hasPropertyTwo();
    V getValue();
}

public interface Thing<V extends Ordered> {
    boolean hasPropertyOne();
    Valued<V> getValuedProperty();
}

public <V extends Ordered> List<V> myMethod(List<Thing<V>> things) {
    List<V> results = new ArrayList<V>();
    for (Thing<V> thing : things) {
        if (thing.hasPropertyOne()) {
            Valued<V> valued = thing.getValuedProperty();
            if (valued != null && valued.hasPropertyTwo()){
                V value = valued.getValue();
                if (value != null){
                    results.add(value);
                }
            }
        }
    }
    results.sort((a, b)->{
        return Integer.compare(a.getOrder(), b.getOrder());
    });
    return results;
}

This last method rewritten using Streams and method references is much more legible and each step of the process is quickly and easily understood - it's not just shorter, it also shows at a glance which interfaces and classes are responsible for the code in each step:

public <V extends Ordered> List<V> myMethod(List<Thing<V>> things) {
    return things.stream()
        .filter(Thing::hasPropertyOne)
        .map(Thing::getValuedProperty)
        .filter(Objects::nonNull)
        .filter(Valued::hasPropertyTwo)
        .map(Valued::getValue)
        .filter(Objects::nonNull)
        .sorted(Comparator.comparing(Ordered::getOrder))
        .collect(Collectors.toList());
}    

When you have a Stream you need to map but want to preserve the initial values as well, you can map the Stream to a Map.Entry<K,V> using a utility method like the following:

public static <K, V> Function<K, Map.Entry<K, V>> entryMapper(Function<K, V> mapper){
    return (k)->new AbstractMap.SimpleEntry<>(k, mapper.apply(k));
}

Then you can use your converter to process Streams having access to both the original and mapped values:

Set<K> mySet;
Function<K, V> transformer = SomeClass::transformerMethod;
Stream<Map.Entry<K, V>> entryStream = mySet.stream()
    .map(entryMapper(transformer));

You can then continue to process that Stream as normal. This avoids the overhead of creating an intermediate collection.

Stream operations fall into two main categories, intermediate and terminal operations, and two sub-categories, stateless and stateful.

Intermediate Operations:

An intermediate operation is always lazy, such as a simple Stream.map. It is not invoked until the stream is actually consumed. This can be verified easily:

Arrays.asList(1, 2 ,3).stream().map(i -> {
    throw new RuntimeException("not gonna happen");
    return i;
});

Intermediate operations are the common building blocks of a stream, chained after the source and are usually followed by a terminal operation triggering the stream chain.

Terminal Operations

Terminal operations are what triggers the consumption of a stream. Some of the more common are Stream.forEach or Stream.collect. They are usually placed after a chain of intermediate operations and are almost always eager.

Stateless Operations

Statelessness means that each item is processed without the context of other items. Stateless operations allow for memory-efficient processing of streams. Operations like Stream.map and Stream.filter that do not require information on other items of the stream are considered to be stateless.

Stateful operations

Statefulness means the operation on each item depends on (some) other items of the stream. This requires a state to be preserved. Statefulness operations may break with long, or infinite, streams. Operations like Stream.sorted require the entirety of the stream to be processed before any item is emitted which will break in a long enough stream of items. This can be demonstrated by a long stream (run at your own risk):

// works - stateless stream
long BIG_ENOUGH_NUMBER = 999999999;
IntStream.iterate(0, i -> i + 1).limit(BIG_ENOUGH_NUMBER).forEach(System.out::println);

This will cause an out-of-memory due to statefulness of Stream.sorted:

// Out of memory - stateful stream
IntStream.iterate(0, i -> i + 1).limit(BIG_ENOUGH_NUMBER).sorted().forEach(System.out::println);

Use Spliterators.spliterator() or Spliterators.spliteratorUnknownSize() to convert an iterator to a stream:

Iterator<String> iterator = Arrays.asList("A", "B", "C").iterator();    
Spliterator<String> spliterator = Spliterators.spliteratorUnknownSize(iterator, 0);
Stream<String> stream = StreamSupport.stream(spliterator, false);

Reduction is the process of applying a binary operator to every element of a stream to result in one value.

The sum() method of an IntStream is an example of a reduction; it applies addition to every term of the Stream, resulting in one final value:

This is equivalent to (((1+2)+3)+4)

The reduce method of a Stream allows one to create a custom reduction. It is possible to use the reduce method to implement the sum() method:

IntStream istr;
    
//Initialize istr
    
OptionalInt istr.reduce((a,b)->a+b);

The Optional version is returned so that empty Streams can be handled appropriately.

Another example of reduction is combining a Stream<LinkedList<T>> into a single LinkedList<T>:

Stream<LinkedList<T>> listStream;
    
//Create a Stream<LinkedList<T>>
    
Optional<LinkedList<T>> bigList = listStream.reduce((LinkedList<T> list1, LinkedList<T> list2)->{
    LinkedList<T> retList = new LinkedList<T>();
    retList.addAll(list1);
    retList.addAll(list2);
    return retList;
});

You can also provide an identity element. For example, the identity element for addition is 0, as x+0==x. For multiplication, the identity element is 1, as x*1==x. In the case above, the identity element is an empty LinkedList<T>, because if you add an empty list to another list, the list that you are "adding" to doesn't change:

Stream<LinkedList<T>> listStream;

//Create a Stream<LinkedList<T>>

LinkedList<T> bigList = listStream.reduce(new LinkedList<T>(), (LinkedList<T> list1, LinkedList<T> list2)->{
    LinkedList<T> retList = new LinkedList<T>();
    retList.addAll(list1);
    retList.addAll(list2);
    return retList;
});

Note that when an identity element is provided, the return value is not wrapped in an Optional—if called on an empty stream, reduce() will return the identity element.

The binary operator must also be associative, meaning that (a+b)+c==a+(b+c). This is because the elements may be reduced in any order. For example, the above addition reduction could be performed like this:

This reduction is equivalent to writing ((1+2)+(3+4)). The property of associativity also allows Java to reduce the Stream in parallel—a portion of the Stream can be reduced by each processor, with a reduction combining the result of each processor at the end.

A use case that comes across frequently, is creating a String from a stream, where the stream-items are separated by a certain character. The Collectors.joining() method can be used for this, like in the following example:

Stream<String> fruitStream = Stream.of("apple", "banana", "pear", "kiwi", "orange");

String result = fruitStream.filter(s -> s.contains("a"))
           .map(String::toUpperCase)
           .sorted()
           .collect(Collectors.joining(", "));
           
System.out.println(result);

Output:

APPLE, BANANA, ORANGE, PEAR

The Collectors.joining() method can also cater for pre- and postfixes:

String result = fruitStream.filter(s -> s.contains("e"))
           .map(String::toUpperCase)
           .sorted()
           .collect(Collectors.joining(", ", "Fruits: ", "."));
           
System.out.println(result);

Output:

Fruits: APPLE, ORANGE, PEAR.

Live on Ideone