Java 8 Streams FlatMap method example

Extract unique words sorted ASC from a list of phrases:

List<String> phrases = Arrays.asList(
        "sporadic perjury",
        "confounded skimming",
        "incumbent jailer",
        "confounded jailer");

List<String> uniqueWords = phrases
        .stream()
        .flatMap(phrase -> Stream.of(phrase.split("\\s+")))
        .distinct()
        .sorted()
        .collect(Collectors.toList());
System.out.println("Unique words: " + uniqueWords);

... and the output:

Unique words: [confounded, incumbent, jailer, perjury, skimming, sporadic]

Made up example

Imagine that you want to create the following sequence: 1, 2, 2, 3, 3, 3, 4, 4, 4, 4 etc. (in other words: 1x1, 2x2, 3x3 etc.)

With flatMap it could look like:

IntStream sequence = IntStream.rangeClosed(1, 4)
                          .flatMap(i -> IntStream.iterate(i, identity()).limit(i));
sequence.forEach(System.out::println);

where:

• IntStream.rangeClosed(1, 4) creates a stream of int from 1 to 4, inclusive
• IntStream.iterate(i, identity()).limit(i) creates a stream of length i of int i - so applied to i = 4 it creates a stream: 4, 4, 4, 4
• flatMap "flattens" the stream and "concatenates" it to the original stream

With Java < 8 you would need two nested loops:

List<Integer> list = new ArrayList<>();
for (int i = 1; i <= 4; i++) {
    for (int j = 0; j < i; j++) {
        list.add(i);
    }
}

Real world example

Let's say I have a List<TimeSeries> where each TimeSeries is essentially a Map<LocalDate, Double>. I want to get a list of all dates for which at least one of the time series has a value. flatMap to the rescue:

list.stream().parallel()
    .flatMap(ts -> ts.dates().stream()) // for each TS, stream dates and flatmap
    .distinct()                         // remove duplicates
    .sorted()                           // sort ascending
    .collect(toList());

Not only is it readable, but if you suddenly need to process 100k elements, simply adding parallel() will improve performance without you writing any concurrent code.

It doesn't make sense to flatMap a Stream that's already flat, like the Stream<Integer> you've shown in your question.

However, if you had a Stream<List<Integer>> then it would make sense and you could do this:

Stream<List<Integer>> integerListStream = Stream.of(
    Arrays.asList(1, 2), 
    Arrays.asList(3, 4), 
    Arrays.asList(5)
);

Stream<Integer> integerStream = integerListStream .flatMap(Collection::stream);
integerStream.forEach(System.out::println);

Which would print:

1
2
3
4
5

To do this pre-Java 8 you just need a loops:

List<List<Integer>> integerLists = Arrays.asList(
    Arrays.asList(1, 2), 
    Arrays.asList(3, 4), 
    Arrays.asList(5)
)

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

for (List<Integer> integerList : integerLists) {
    flattened.addAll(integerList);
}

for (Integer i : flattened) {
    System.out.println(i);
}