9.2 The Collectors API: Downstream Reductions, Grouping, and Partitioning

9.2 The Collectors API: Downstream Reductions, Grouping, and Partitioning

In Java Stream pipelines, terminal operations convert intermediate streams into concrete domain collections, primitive aggregations, or complex grouped data structures. While Stream.reduce() performs immutable folding, the mutable reduction mechanism provided by Stream.collect(Collector) is the industry standard for high-performance data processing. The utility class java.util.stream.Collectors provides a rich set of pre-built, reusable implementations. The 1Z0-829 examination tests your ability to choose, configure, and nest collectors, handle map key collisions, apply downstream reductions, and use modern additions like partitioningBy and Java 12's teeing collector.

---

The Architecture of Collector<T, A, R>

A Collector encapsulates the recipe for performing a mutable reduction into an intermediate accumulator container A, ultimately producing a final result R from elements of type T.

The Collector<T, A, R> interface defines four core functional components:

1. Supplier<A> supplier(): Creates and returns a new, mutable intermediate container (e.g., ArrayList::new, StringBuilder::new, or HashMap::new).
2. BiConsumer<A, T> accumulator(): Folds a single stream element T into the mutable container A (e.g., List::add).
3. BinaryOperator<A> combiner(): Merges two partial accumulation containers into one. This method is critical for parallel streams when different threads process distinct sub-tasks.
4. Function<A, R> finisher(): Transforms the intermediate container A into the final result type R. If container A requires no final transformation, the collector reports the IDENTITY_FINISH characteristic and the identity function is used.

Collectors can also report optimization characteristics via characteristics():

• IDENTITY_FINISH: Indicates the finisher is an unchecked identity cast from A to R.
• CONCURRENT: Indicates the accumulator can be invoked concurrently by multiple threads on a shared container without synchronization (used primarily with concurrent data structures).
• UNORDERED: Indicates the collection operation does not depend on preserving encounter order.

---

Basic Collection Targets: Lists, Sets, and Custom Collections

List<String> items = Stream.of("alpha", "beta", "gamma").toList(); // Java 16+ unmodifiable list

// 1. Collectors.toList(): Typically returns a modifiable ArrayList (no guarantee)
List<String> modifiableList = items.stream().collect(Collectors.toList());

// 2. Collectors.toSet(): Collects into a HashSet (eliminates duplicates, drops order)
Set<String> set = items.stream().collect(Collectors.toSet());

// 3. Collectors.toCollection(Supplier): Collects into an explicit collection type
TreeSet<String> sortedSet = items.stream()
    .collect(Collectors.toCollection(TreeSet::new));
LinkedList<String> linkedList = items.stream()
    .collect(Collectors.toCollection(LinkedList::new));

// 4. Collectors.toUnmodifiableList() / toUnmodifiableSet() (Java 10+)
List<String> immutableList = items.stream().collect(Collectors.toUnmodifiableList());

---

Map Collection and the Duplicate Key Trap

Transforming stream elements into a Map using Collectors.toMap() requires careful attention to key collisions and null handling.

The Three toMap() Overloads

1. 2-Argument: toMap(keyMapper, valueMapper)

   • Uses elements to generate keys and values.
   • Throws IllegalStateException if duplicate keys are generated!

2. 3-Argument: toMap(keyMapper, valueMapper, mergeFunction)

   • When duplicate keys collide, invokes mergeFunction.apply(existingVal, newVal) to resolve the collision.

3. 4-Argument: toMap(keyMapper, valueMapper, mergeFunction, mapSupplier)

   • Supplies a custom Map implementation (e.g., TreeMap::new, LinkedHashMap::new).

record Employee(int id, String department, double salary) {}

List<Employee> staff = List.of(
    new Employee(101, "Engineering", 90000),
    new Employee(102, "HR", 65000),
    new Employee(103, "Engineering", 95000)
);

// 2-arg toMap: id is unique -> SUCCESS
Map<Integer, Double> idSalaryMap = staff.stream()
    .collect(Collectors.toMap(Employee::id, Employee::salary));

// 2-arg toMap on duplicate department -> THROWS IllegalStateException!
// Map<String, Double> deptSalaryMap = staff.stream()
//     .collect(Collectors.toMap(Employee::department, Employee::salary));

// 3-arg toMap with mergeFunction: keep the higher salary on collision
Map<String, Double> topDeptSalary = staff.stream()
    .collect(Collectors.toMap(
        Employee::department,
        Employee::salary,
        (existing, incoming) -> Math.max(existing, incoming)
    ));

// 4-arg toMap: preserves sorted key order via TreeMap
TreeMap<String, Double> sortedDeptMap = staff.stream()
    .collect(Collectors.toMap(
        Employee::department,
        Employee::salary,
        Double::sum,
        TreeMap::new
    ));

---

String Joining and Numeric Aggregations

Collectors.joining()

Joins elements of CharSequence (or String) into a single output string.

• joining(): Concatenates elements with no delimiter.
• joining(delimiter): Concatenates elements separated by the delimiter.
• joining(delimiter, prefix, suffix): Concatenates with delimiter, enclosed by prefix and suffix.

String csv = Stream.of("red", "green", "blue")
    .collect(Collectors.joining(", ", "[", "]")); // "[red, green, blue]"

Numeric and Statistical Collectors

Rather than mapping to primitive streams (mapToInt()), you can aggregate directly in collect():

• counting(): Returns Collector<T, ?, Long> counting elements.
• summingInt(ToIntFunction), summingLong(), summingDouble(): Returns the sum as boxed Integer, Long, or Double.
• averagingInt(), averagingLong(), averagingDouble(): Always returns Double (even for averagingInt()).
• summarizingInt(), summarizingLong(), summarizingDouble(): Returns rich summary objects (IntSummaryStatistics, DoubleSummaryStatistics) providing count, sum, min, max, and average in a single pass.

IntSummaryStatistics stats = staff.stream()
    .collect(Collectors.summarizingInt(e -> (int) e.salary()));
System.out.println("Max: " + stats.getMax() + ", Avg: " + stats.getAverage());

Post-Processing: Collectors.collectingAndThen()

Performs a collection and then immediately applies a finishing transformation function to the accumulated result:

List<String> readOnlyList = staff.stream()
    .map(Employee::department)
    .collect(Collectors.collectingAndThen(
        Collectors.toList(),
        Collections::unmodifiableList
    ));

---

Classification with groupingBy

Collectors.groupingBy() organizes stream elements into groups based on a classification function, analogous to GROUP BY in SQL.

The Three groupingBy Overloads

1. 1-Argument: groupingBy(classifier)
   • Defaults to a HashMap<K, List<T>>.
2. 2-Argument: groupingBy(classifier, downstreamCollector)
   • Classifies by key, but processes values in each bucket using the specified downstream collector instead of accumulating into a List.
3. 3-Argument: groupingBy(classifier, mapSupplier, downstreamCollector)
   • Allows choosing the concrete Map class (e.g., TreeMap::new).

// 1-arg: Map<String, List<Employee>>
Map<String, List<Employee>> byDept = staff.stream()
    .collect(Collectors.groupingBy(Employee::department));

// 2-arg: Map<String, Long> (count employees per department)
Map<String, Long> headCount = staff.stream()
    .collect(Collectors.groupingBy(Employee::department, Collectors.counting()));

// 2-arg: Map<String, Double> (average salary per department)
Map<String, Double> avgSalary = staff.stream()
    .collect(Collectors.groupingBy(Employee::department, Collectors.averagingDouble(Employee::salary)));

// 3-arg: TreeMap<String, Set<Double>>
TreeMap<String, Set<Double>> salariesByDept = staff.stream()
    .collect(Collectors.groupingBy(
        Employee::department,
        TreeMap::new,
        Collectors.mapping(Employee::salary, Collectors.toSet())
    ));

Downstream Collectors: mapping(), filtering(), and flatMapping()

• mapping(mapper, downstream): Transforms bucket elements before downstream collection.
• filtering(predicate, downstream) (Java 9+): Filters bucket elements. Crucial Advantage: Unlike a pipeline .filter() which eliminates keys entirely if all their items are dropped, filtering() preserves the key with an empty downstream collection!
• flatMapping(mapper, downstream) (Java 9+): Flattens nested collections generated from bucket elements.

---

Partitioning with partitioningBy

Collectors.partitioningBy() is a specialized form of grouping that divides stream elements into two categories based on a Predicate<? super T>.

• 1-Argument: partitioningBy(predicate) -> Returns Map<Boolean, List<T>>.
• 2-Argument: partitioningBy(predicate, downstreamCollector) -> Returns Map<Boolean, D>.

Map<Boolean, List<Employee>> highEarners = staff.stream()
    .collect(Collectors.partitioningBy(e -> e.salary() >= 90000));

// highEarners.get(true) contains employees earning >= 90000
// highEarners.get(false) contains employees earning < 90000