Working with Stream API

1. Creating Streams

Source	Factory	Notes
Collection	`coll.stream()` / `parallelStream()`	Most common
Array	`Arrays.stream(arr)`	Primitive overloads
Values	`Stream.of(a, b, c)`	Varargs
Range	`IntStream.range(0, n)`	Half-open
Generate	`Stream.generate(supplier)`	Infinite
Iterate	`Stream.iterate(seed, fn)` / `iterate(seed, hasNext, next)`	Bounded form Java 9+
Files	`Files.lines(path)`	Auto-closed
Builder	`Stream.builder().add(x).build()`	Imperative add
Empty	`Stream.empty()`	Avoid null

2. Using Intermediate Operations

Operation	Stateless?	Purpose
`filter(Predicate)`	Yes	Keep matching
`map(Function)`	Yes	1:1 transform
`flatMap(Function)`	Yes	1:N flatten
`mapMulti` Java 16+	Yes	Imperative flatten
`distinct()`	Stateful	Dedup via `equals`
`sorted()`	Stateful	Buffers all elements
`peek(Consumer)`	Yes	Debug only
`limit(n)` / `skip(n)`	Stateful	Slicing
`takeWhile` / `dropWhile`	Stateful	Java 9+

3. Using Terminal Operations

Operation	Returns	Short-circuits
`forEach` / `forEachOrdered`	void	No
`collect(Collector)`	R	No
`toList()` Java 16+	Unmodifiable List	No
`reduce`	T or Optional<T>	No
`count`	long	No (sized → yes)
`min` / `max`	Optional	No
`findFirst` / `findAny`	Optional	Yes
`anyMatch` / `allMatch` / `noneMatch`	boolean	Yes
`toArray`	Object[] / T[]	No

4. Understanding Stream Pipeline

Source → [Intermediate ops fused] → Terminal op
        (lazy, no work)            (triggers traversal)

Per element: filter → map → flatMap → ... → terminal
(pull model — terminal pulls one element through the pipeline)

Property	Meaning
Lazy	Intermediate ops do nothing until terminal
Single use	Stream cannot be reused after terminal
Fusion	Adjacent stateless ops combined
Encounter order	Preserved unless `unordered()`

5. Working with Primitive Streams

Stream	Box	Specialized Ops
`IntStream`	Integer	`sum, average, summaryStatistics, range`
`LongStream`	Long	Same as IntStream
`DoubleStream`	Double	Same minus `range`

Example: Avoid boxing

int total = orders.stream().mapToInt(Order::amount).sum();
IntSummaryStatistics st = orders.stream().mapToInt(Order::amount).summaryStatistics();

6. Using Collectors

Collector	Result
`toList() / toSet()`	List / Set
`toUnmodifiableList()`	Immutable list
`toMap(k, v[, merge])`	Map
`groupingBy(classifier[, downstream])`	Map<K, List<V>>
`partitioningBy(predicate)`	Map<Boolean, List>
`counting()`	Long count
`summingInt` / `averagingInt`	Aggregation
`joining(delim, prefix, suffix)`	String
`mapping` / `filtering` / `flatMapping`	Downstream adapters
`teeing(c1, c2, merger)` Java 12+	Combine two collectors

Example: Group + summarize

Map<String, Long> byDept = emps.stream()
    .collect(Collectors.groupingBy(Employee::dept, Collectors.counting()));

Map<String, Double> avgSalary = emps.stream()
    .collect(Collectors.groupingBy(Employee::dept, Collectors.averagingDouble(Employee::salary)));

7. Creating Custom Collectors (Collector.of())

Argument	Type	Role
supplier	`Supplier<A>`	New mutable container
accumulator	`BiConsumer<A,T>`	Add element
combiner	`BinaryOperator<A>`	Merge for parallel
finisher	`Function<A,R>`	Optional final transform
characteristics	`CONCURRENT, UNORDERED, IDENTITY_FINISH`	Hints

Example: Top-N collector

public static <T> Collector<T, ?, List<T>> topN(int n, Comparator<? super T> cmp) {
    return Collector.of(
        () -> new PriorityQueue<T>(cmp),
        (q, x) -> { q.add(x); if (q.size() > n) q.poll(); },
        (a, b) -> { b.forEach(a::add); while (a.size() > n) a.poll(); return a; },
        q -> q.stream().sorted(cmp.reversed()).toList()
    );
}

8. Using Parallel Streams (parallelStream())

Aspect	Detail
Pool	Common ForkJoinPool (size = cores − 1)
Custom pool	Submit stream task to `ForkJoinPool`
Good fit	Large N, CPU-bound, splittable source
Bad fit	Small N, blocking I/O, ordered ops
Reduce	Identity must satisfy `combiner(identity, x) == x`

Warning: Avoid shared mutable state and side effects in parallel streams; collisions cause non-determinism.

9. Understanding Stream Performance

Factor	Impact
Auto-boxing	Use primitive streams
Stateful ops	Buffer entire stream (sorted, distinct)
Splitability	ArrayList good; LinkedList poor
Short-circuit	Place selective filters early
Object overhead	vs plain for-loop ~ 2-5×

10. Handling Infinite Streams

Pattern	Example
Bounded by limit	`Stream.iterate(0, i->i+1).limit(100)`
Bounded form	`Stream.iterate(0, i->i<100, i->i+1)`
takeWhile	`stream.takeWhile(x -> x < 50)`
findFirst on infinite	OK if predicate eventually matches

Warning: Never call count(), collect(), or sorted() on an unbounded stream.

11. Using FlatMap for Nested Structures

Variant	Maps Each Element To
`flatMap`	Stream<R>
`flatMapToInt/Long/Double`	Primitive stream
`mapMulti`	Pushes 0..N to consumer

Example: Flatten orders

List<Item> allItems = orders.stream()
    .flatMap(o -> o.items().stream())
    .toList();

12. Debugging Streams (peek())

Tool	Use
`peek`	Inspect elements between ops
Logger.atDebug	Conditional logging
IntelliJ Stream Trace	Visualize per-stage values
Stream Debugger plugin	Step through pipeline

Note: JIT may elide peek if its result is unused after a non-stateful terminal — don't rely on it for production side effects.

13. Using takeWhile() and dropWhile() (Java 9+)

Method	Behavior	Ordered Stream
`takeWhile(p)`	Take prefix while p holds	Stops at first false
`dropWhile(p)`	Skip prefix while p holds	Emits rest
vs `filter`	Position-aware vs all elements	Different semantics

14. Using Stream.toList() (Java 16+)

Aspect	`toList()`	`collect(toList())`
Mutability	Unmodifiable	Mutable ArrayList
Nulls	Allowed	Allowed
Performance	Slightly faster (presized)	Baseline
Recommendation	Default choice	Use only when mutability needed

Related Cheat Sheets

Spring Boot