Java 8 Stream API - Selecting only values after Collectors.groupingBy(..)

Question

Say I have the following collection of Student objects which consist of Name(String), Age(int) and City(String).

I am trying to use Java's Stream API to achieve the following sql-like behavior:

SELECT MAX(age)
FROM Students
GROUP BY city

Now, I found two different ways to do so:

final List<Integer> variation1 =
            students.stream()
                    .collect(Collectors.groupingBy(Student::getCity, Collectors.maxBy((s1, s2) -> s1.getAge() - s2.getAge())))
                    .values()
                    .stream()
                    .filter(Optional::isPresent)
                    .map(Optional::get)
                    .map(Student::getAge)
                    .collect(Collectors.toList());

And the other one:

final Collection<Integer> variation2 =
            students.stream()
                    .collect(Collectors.groupingBy(Student::getCity,
                            Collectors.collectingAndThen(Collectors.maxBy((s1, s2) -> s1.getAge() - s2.getAge()),
                                    optional -> optional.get().getAge())))
                    .values();

In both ways, one has to .values() ... and filter the empty groups returned from the collector.

Is there any other way to achieve this required behavior?

These methods remind me of over partition by sql statements...

Thanks

---

Edit: All the answers below were really interesting, but unfortunately this is not what I was looking for, since what I try to get is just the values. I don't need the keys, just the values.

Answer 1

Do not always stick with groupingBy. Sometimes toMap is the thing you need:

Collection<Integer> result = students.stream()
    .collect(Collectors.toMap(Student::getCity, Student::getAge, Integer::max))
    .values();

Here you just create a Map where keys are cities and values are ages. In case when several students have the same city, merge function is used which just selects maximal age here. It's faster and cleaner.

Answer 2

As addition to Tagir’s great answer using toMap instead of groupingBy, here the short solution, if you want to stick to groupingBy:

Collection<Integer> result = students.stream()
    .collect(Collectors.groupingBy(Student::getCity,
                 Collectors.reducing(-1, Student::getAge, Integer::max)))
    .values();

Note that this three arg reducing collector already performs a mapping operation, so we don’t need to nest it with a mapping collector, further, providing an identity value avoids dealing with Optional. Since ages are always positive, providing -1 is sufficient and since a group will always have at least one element, the identity value will never show up as a result.

Still, I think Tagir’s toMap based solution is preferable in this scenario.

---

The groupingBy based solution becomes more interesting when you want to get the actual students having the maximum age, e.g

Collection<Student> result = students.stream().collect(
   Collectors.groupingBy(Student::getCity, Collectors.reducing(null, BinaryOperator.maxBy(
     Comparator.nullsFirst(Comparator.comparingInt(Student::getAge)))))
).values();

well, actually, even this can also be expressed using the toMap collector:

Collection<Student> result = students.stream().collect(
    Collectors.toMap(Student::getCity, Function.identity(),
        BinaryOperator.maxBy(Comparator.comparingInt(Student::getAge)))
).values();

You can express almost everything with both collectors, but groupingBy has the advantage on its side when you want to perform a mutable reduction on the values.

Answer 3

The second approach calls get() on an Optional; this is usually a bad idea as you don't know if the optional will be empty or not (use orElse(), orElseGet(), orElseThrow() methods instead). While you might argue that in this case there always be a value since you generate the values from the student list itself, this is something to keep in mind.

Based on that, you might turn the variation 2 into:

final Collection<Integer> variation2 =
     students.stream()
             .collect(collectingAndThen(groupingBy(Student::getCity,
                                                   collectingAndThen(
                                                      mapping(Student::getAge, maxBy(naturalOrder())),
                                                      Optional::get)), 
                                        Map::values));

Although it really starts to be difficult to read, I'll probably use the variant 1:

final List<Integer> variation1 =
        students.stream()
            .collect(groupingBy(Student::getCity,
                                mapping(Student::getAge, maxBy(naturalOrder()))))
            .values()
            .stream()
            .map(Optional::get)
            .collect(toList());

Answer 4

Here is my implementation 

    public class MaxByTest {

        static class Student {

        private int age;
        private int city;        

        public Student(int age, int city) {
            this.age = age;
            this.city = city;
        }

        public int getCity() {
            return city;
        }

        public int getAge() {
            return age;
        }

        @Override
        public String toString() {
            return " City : " + city + " Age : " + age;
        }



    }

    static List<Student> students = Arrays.asList(new Student[]{
        new Student(10, 1),
        new Student(9, 2),        
        new Student(8, 1),        
        new Student(6, 1),
        new Student(4, 1),
        new Student(8, 2),
        new Student(9, 2),
        new Student(7, 2),        
    });

    public static void main(String[] args) {
        final Comparator<Student> comparator = (p1, p2) -> Integer.compare( p1.getAge(), p2.getAge());
        final List<Student> studets =
            students.stream()
                    .collect(Collectors.groupingBy(Student::getCity, 
                            Collectors.maxBy(comparator))).values().stream().map(Optional::get).collect(Collectors.toList());
        System.out.println(studets);
    }
}

Answer 5

        List<BeanClass> list1 = new ArrayList<BeanClass>();
        DateFormat formatter = new SimpleDateFormat("yyyy-MM-dd");
        list1.add(new BeanClass(123,abc,99.0,formatter.parse("2018-02-01")));
        list1.add(new BeanClass(456,xyz,99.0,formatter.parse("2014-01-01")));
        list1.add(new BeanClass(789,pqr,95.0,formatter.parse("2014-01-01")));
        list1.add(new BeanClass(1011,def,99.0,formatter.parse("2014-01-01")));
        Map<Object, Optional<Double>> byDate = list1.stream()
       .collect(Collectors.groupingBy(p -> formatter.format(p.getCurrentDate()),
        Collectors.mapping(BeanClass::getAge, Collectors.maxBy(Double::compare))));