📦 PART 4: Collections Framework (Q106–150)

Iterable<T>
  └── Collection<T>
        ├── List<T>
        │     ├── ArrayList
        │     ├── LinkedList
        │     ├── Vector (legacy)
        │     │     └── Stack (legacy)
        │     └── CopyOnWriteArrayList
        ├── Set<T>
        │     ├── HashSet
        │     │     └── LinkedHashSet
        │     ├── TreeSet (SortedSet → NavigableSet)
        │     ├── ConcurrentSkipListSet
        │     └── CopyOnWriteArraySet
        └── Queue<T>
              ├── LinkedList
              ├── PriorityQueue
              ├── ArrayDeque
              └── BlockingQueue (interface)
                    ├── LinkedBlockingQueue
                    ├── ArrayBlockingQueue
                    └── PriorityBlockingQueue

Map<K,V> (does NOT extend Collection)
  ├── HashMap
  │     └── LinkedHashMap
  ├── TreeMap (SortedMap → NavigableMap)
  ├── Hashtable (legacy)
  │     └── Properties
  ├── WeakHashMap
  ├── IdentityHashMap
  ├── EnumMap
  └── ConcurrentHashMap

Default choice: ArrayList. Use LinkedList only when frequent O(1) insertions/deletions at head/middle with iteration. In practice, ArrayList is almost always faster due to CPU cache effects even for some "LinkedList scenarios".

LinkedList excels as a Deque (double-ended queue):

LinkedList<String> deque = new LinkedList<>();
deque.addFirst("first"); // O(1)
deque.addLast("last");   // O(1)
deque.removeFirst();      // O(1)
deque.removeLast();       // O(1)

Use cases:

1. Queue/Deque implementation: Where you frequently add/remove from both ends
2. LRU Cache: Remove from tail, add to head (but LinkedHashMap is usually better)
3. Stack with frequent push/pop at head

Prefer ArrayDeque over LinkedList for most queue/deque scenarios – ArrayDeque is faster (array-based, better cache behavior) and uses less memory.

Both are resizable arrays, but:

Vector is legacy – use ArrayList + explicit synchronization (Collections.synchronizedList()) or CopyOnWriteArrayList for concurrent use.

Vector synchronizes every method at the object level, even when no concurrent access is happening. This is:

1. Too coarse-grained: Even iteration requires locking every element access
2. Not composable: if (!v.isEmpty()) v.get(0) is still NOT thread-safe (race between isEmpty and get)
3. Performance penalty: Even single-threaded code pays sync cost

Modern alternatives:

• ArrayList for non-concurrent
• CopyOnWriteArrayList for read-heavy concurrent
• Collections.synchronizedList(new ArrayList<>()) for simple sync need
• ConcurrentLinkedDeque for lock-free concurrent deque

HashMap uses an array of buckets (Node[]) with linked list (Java 7) or red-black tree (Java 8+ when chain length ≥ 8).

Array (buckets):
[0] -> null
[1] -> Node("Alice", 30) -> Node("Bob", 25)  // collision - same bucket
[2] -> Node("Charlie", 35)
[3] -> null
...
[15] -> Node("Dave", 28)

put(key, value) flow:

1. key.hashCode() computed
2. Hash spread: h ^ (h >>> 16) (reduces clustering)
3. Bucket index: hash & (capacity - 1) (bitwise AND, fast when capacity is power of 2)
4. If bucket empty: insert new Node
5. If bucket has entries: iterate/compare with equals()

• Key found: update value
• Key not found: append (add to tree if chain ≥ 8)

get(key) flow:

1. Compute hash → bucket index
2. Traverse nodes in bucket comparing with equals()
3. Return value or null

Treeification (Java 8): When a single bucket's linked list reaches 8 nodes AND capacity ≥ 64, converts to red-black tree → O(log n) vs O(n) for long chains.

Default capacity: 16 (always power of 2)

Default load factor: 0.75

Threshold = capacity × load factor. When entries exceed threshold, HashMap rehashes (creates new array of double size, re-inserts all entries).

Example: Default map with 16 capacity, threshold = 12. When 13th entry is added, resize to 32.

Why power of 2? hash & (n-1) is faster than hash % n and distributes evenly.

Why 0.75? Balance between time (fewer collisions, bigger array) and space (fewer rehashes, smaller array). Below 0.6 = too many empty buckets. Above 0.85 = too many collisions.

Pre-sizing: If you know approximate size, provide it:

Map<String, String> map = new HashMap<>(expectedSize / 0.75 + 1);
// Or use Guava: Maps.newHashMapWithExpectedSize(expectedSize)

Collision: two keys hash to the same bucket.

Java 7: Separate chaining with linked list. New entries added to head (causes a subtle ABA problem in concurrent use that leads to infinite loops in multi-threaded HashMap – don't use HashMap from multiple threads!).

Java 8:

• Linked list for ≤ 8 nodes per bucket
• Red-black tree for > 8 nodes (if capacity ≥ 64)
• Tree degrades back to list when nodes ≤ 6

Why switch to tree? In worst case (all keys hash to same bucket), linked list = O(n). Tree = O(log n). Mitigates hash DoS attacks where attacker crafts many keys with same hash.

String's hashCode() is deterministic and used to be exploitable – Java uses random hash seed (-XX:+UseStringDeduplication) to mitigate.

Hashtable is legacy – use ConcurrentHashMap for thread-safe map (much better than Hashtable).

Java 8 ConcurrentHashMap internals: No segment locks. Uses CAS for most operations. Only locks individual buckets during tree restructuring. size() uses LongAdder-like approach.

Key methods:

• putIfAbsent(k, v): Atomic put-if-missing
• computeIfAbsent(k, fn): Compute and store if missing
• merge(k, v, fn): Atomic merge
• compute(k, fn): Atomic compute

Use TreeMap when:

• Need keys in sorted order
• Need range queries: "all entries with key between 100 and 200"
• Need floor/ceiling key operations

TreeMap<Integer, String> scores = new TreeMap<>();
scores.subMap(70, 100); // entries with keys 70-99
scores.floorKey(85); // highest key ≤ 85
scores.ceilingKey(70); // lowest key ≥ 70
scores.headMap(50); // keys < 50

LinkedHashMap maintains insertion order (or access order).

// Insertion order (default)
Map<String, Integer> map = new LinkedHashMap<>();
map.put("banana", 2); map.put("apple", 1); map.put("cherry", 3);
// Iterates: banana, apple, cherry

// Access order (LRU behavior)
Map<String, Integer> lru = new LinkedHashMap<>(16, 0.75f, true) {
    @Override protected boolean removeEldestEntry(Map.Entry<String, Integer> eldest) {
        return size() > MAX_CACHE_SIZE; // evict oldest access
    }
};

Use cases:

1. Preserve insertion order in map iteration
2. LRU Cache (access-order mode + removeEldestEntry)
3. Ordered JSON/config properties
4. Event logging where insertion order matters

Conceptual mismatch:

• Collection<E> stores elements
• Map<K,V> stores key-value pairs

If Map extended Collection, what would E be? Map.Entry<K,V>? But then:

• add(Map.Entry<K,V>) doesn't make sense vs put(K, V)
• contains(Object) – contains key? value? entry?
• remove(Object) – remove by key? value?
• size() – conflicts in contract

The operations and semantics are fundamentally different. Map provides entrySet(), keySet(), values() which return Collection views.

Josh Bloch (Java Collections designer): "Maps are not proper Collections because the element concept doesn't fit the Map abstraction."

HashSet is backed by a HashMap (values are a dummy PRESENT object):

public class HashSet<E> {
    private transient HashMap<E, Object> map;
    private static final Object PRESENT = new Object();
    
    public boolean add(E e) { return map.put(e, PRESENT) == null; }
    public boolean contains(Object o) { return map.containsKey(o); }
}

Uniqueness enforced via hashCode() and equals():

1. Compute hashCode() → find bucket
2. Compare with equals() against each element in bucket
3. If match found: not added (duplicate)
4. If no match: added

This is why overriding equals() without hashCode() in Set elements causes duplicates:

class Point { int x, y; }
// Without hashCode override:
Set<Point> set = new HashSet<>();
set.add(new Point(1,1));
set.add(new Point(1,1)); // Different hashCode! Different bucket! Both added!

// TreeSet range operations
TreeSet<Integer> set = new TreeSet<>(Set.of(1, 5, 3, 7, 9));
set.headSet(5);        // [1, 3] - elements < 5
set.tailSet(5);        // [5, 7, 9] - elements >= 5
set.subSet(3, 7);      // [3, 5] - elements >= 3 and < 7
set.floor(6);          // 5 - greatest element <= 6
set.ceiling(6);        // 7 - smallest element >= 6

TreeSet maintains a red-black tree (self-balancing BST). Every add/remove/contains requires:

1. Traversal from root to correct position: O(log n)
2. Potentially rebalancing the tree: O(log n) rotations

HashSet's hash-based approach achieves O(1) average case (single bucket lookup if no collisions).

For n=1,000,000: HashSet find = 1 hash + ~1 comparison; TreeSet find = ~20 comparisons (log₂(1,000,000) ≈ 20).

Comparable<T> is an interface for defining natural ordering of a class.

public interface Comparable<T> {
    int compareTo(T other);
}

• Returns negative: this < other
• Returns 0: this == other
• Returns positive: this > other

public class Employee implements Comparable<Employee> {
    private String name;
    private double salary;
    
    @Override
    public int compareTo(Employee other) {
        return Double.compare(this.salary, other.salary); // sort by salary
    }
}

Used by TreeSet, TreeMap, Collections.sort(), Arrays.sort() when no explicit Comparator provided.

Contract: Must be consistent with equals. If compareTo() returns 0, equals() should return true.

// Comparable - in the class
class Person implements Comparable<Person> {
    public int compareTo(Person other) { return this.age - other.age; }
}

// Comparator - external, multiple options
Comparator<Person> byName = Comparator.comparing(Person::getName);
Comparator<Person> byAgeDesc = Comparator.comparingInt(Person::getAge).reversed();
Comparator<Person> complex = Comparator.comparing(Person::getLastName)
                                        .thenComparing(Person::getFirstName)
                                        .thenComparingInt(Person::getAge);

List<Employee> employees = ...;

// By salary descending
employees.sort(Comparator.comparingDouble(Employee::getSalary).reversed());

// Multiple criteria
employees.sort(Comparator.comparing(Employee::getDepartment)
    .thenComparing(Employee::getName)
    .thenComparingDouble(Employee::getSalary));

// Null-safe
employees.sort(Comparator.comparing(Employee::getManager, 
    Comparator.nullsLast(Comparator.naturalOrder())));

// Stream
employees.stream()
    .sorted(Comparator.comparing(Employee::getHireDate))
    .collect(Collectors.toList());

Fail-fast iterators (ArrayList, HashMap, HashSet): Throw ConcurrentModificationException if collection is structurally modified during iteration (not through the iterator).

Mechanism: modCount counter tracks structural modifications. Iterator checks modCount on each next() call.

List<String> list = new ArrayList<>(List.of("a", "b", "c"));
for (String s : list) {
    if (s.equals("b")) list.remove(s); // ConcurrentModificationException!
}
// Fix: list.removeIf(s -> s.equals("b"));
// Or: use iterator.remove()

Fail-safe iterators (CopyOnWriteArrayList, ConcurrentHashMap): Iterate over a snapshot or weakly-consistent view; no exception thrown, but may not reflect latest changes.

CopyOnWriteArrayList<String> cowList = new CopyOnWriteArrayList<>(List.of("a", "b"));
for (String s : cowList) {
    cowList.add("new"); // no exception, iterator sees original snapshot
}

List<String> list = new ArrayList<>(List.of("a", "b", "c"));
ListIterator<String> it = list.listIterator();
while (it.hasNext()) {
    String s = it.next();
    it.set(s.toUpperCase()); // replace current element
}
while (it.hasPrevious()) {
    System.out.println(it.previous()); // C, B, A
}

Calling list.remove() during for-each throws ConcurrentModificationException. Iterator's remove() is the safe way to remove during iteration.

Iterator<String> it = list.iterator();
while (it.hasNext()) {
    String s = it.next();
    if (condition(s)) it.remove(); // safe - updates modCount
}

it.remove() removes the last element returned by next(). Must call next() first. Can only call once per next().

Modern alternative (Java 8+):

list.removeIf(s -> condition(s)); // cleaner

Use Iterator when: simple sequential traversal with possible removal. Use Stream when: transformations, filtering, aggregations, parallel processing.

Spliterator (Splittable Iterator) supports parallel traversal of a data source.

public interface Spliterator<T> {
    boolean tryAdvance(Consumer<? super T> action); // process one element
    Spliterator<T> trySplit(); // split for parallel processing
    long estimateSize(); // rough count
    int characteristics(); // ORDERED, SORTED, SIZED, DISTINCT, IMMUTABLE, etc.
}

Stream API uses Spliterators internally for parallel streams:

1. trySplit() recursively splits the Spliterator
2. Each split processed by different threads in ForkJoinPool
3. Results combined

Custom Spliterators: Implement for custom data structures to support parallel stream processing.

Queue: FIFO – insert at tail, remove from head.

Queue<String> queue = new LinkedList<>();
queue.offer("a"); // add to tail
queue.poll();     // remove from head
queue.peek();     // look at head without removing

Deque (Double-Ended Queue): Insert/remove at both ends.

Deque<String> deque = new ArrayDeque<>();
deque.offerFirst("a"); // add to front
deque.offerLast("b");  // add to back
deque.pollFirst();      // remove from front
deque.pollLast();       // remove from back

ArrayDeque is preferred over LinkedList for queue/deque use cases (faster, less memory).

ArrayDeque as stack: push() = addFirst(), pop() = removeFirst() – faster than Stack.

PriorityQueue uses a min-heap (binary heap stored in array). Smallest element (by natural order or Comparator) is always at the head.

PriorityQueue<Integer> pq = new PriorityQueue<>();
pq.offer(5); pq.offer(1); pq.offer(3);
pq.poll(); // returns 1 (minimum)

Internal structure: Array where element at index i has children at 2i+1 and 2i+2, parent at (i-1)/2.

Operations:

• offer(e): Add at end, sift up – O(log n)
• poll(): Remove root (min), replace with last, sift down – O(log n)
• peek(): Return root – O(1)
• remove(o): Find and remove arbitrary element – O(n) to find

Not sorted (heap order, not sorted order). Iteration is NOT in priority order; only poll() guarantees order.

Max-heap:

PriorityQueue<Integer> maxHeap = new PriorityQueue<>(Comparator.reverseOrder());

BlockingQueue extends Queue with blocking operations for producer-consumer patterns:

BlockingQueue<Task> queue = new LinkedBlockingQueue<>(100); // capacity 100

// Producer
queue.put(task);    // blocks if full
queue.offer(task, 5, TimeUnit.SECONDS); // blocks up to 5s

// Consumer
Task task = queue.take();    // blocks if empty
Task task = queue.poll(5, TimeUnit.SECONDS); // blocks up to 5s

Implementations:

• LinkedBlockingQueue: Unbounded (or capacity-limited), linked nodes
• ArrayBlockingQueue: Fixed capacity, array-backed, fair/unfair option
• PriorityBlockingQueue: Priority ordered, unbounded
• DelayQueue: Elements available after a delay
• SynchronousQueue: No storage; direct handoff between producer and consumer
• LinkedTransferQueue: More efficient for some producer-consumer scenarios

Used in: Thread pool's work queue (Executors.newFixedThreadPool uses LinkedBlockingQueue), message passing between threads.

ConcurrentSkipList*: Lock-free, sorted, O(log n) operations. Alternative to synchronized TreeMap/TreeSet.

ConcurrentLinkedDeque: Lock-free, non-blocking concurrent deque. High throughput.

CopyOnWriteArrayList: On every write (add, set, remove), creates a new copy of the entire array. Reads are non-blocking (use snapshot of array at time of iterator creation).

CopyOnWriteArrayList<String> list = new CopyOnWriteArrayList<>();
// Read: No locking, uses current array reference
// Write: Creates new array, copies, modifies, switches reference atomically

Best for: Read-heavy, write-rare scenarios (observer lists, listener registries, routing tables).

Terrible for: High write frequency (copies entire array on each write – expensive for large lists).

Trade-offs:

• Reads: Very fast, no locking
• Writes: Slow (O(n) copy), expensive
• Memory: May hold multiple array versions during GC
• Iteration: Safe (snapshot), won't see concurrent modifications

WeakHashMap: Keys are held with weak references. If no strong reference to a key exists elsewhere, GC can collect the key and the entry is automatically removed.

WeakHashMap<Object, String> cache = new WeakHashMap<>();
Object key = new Object();
cache.put(key, "value");
key = null; // Remove strong reference
// After GC, the entry may be removed from the map
System.gc();
cache.size(); // may be 0

Use cases:

1. Caches where values should be collected when key is no longer used
2. Metadata/attributes attached to objects (without preventing GC)
3. Canonical mapping: canonicalize instances (no duplicate instances of same logical object)

Pitfall: If key is a string literal (interned, always strongly reachable), it's never collected. WeakHashMap with String keys often doesn't work as expected.

IdentityHashMap: Uses == (reference equality) instead of .equals() for key comparison, and System.identityHashCode() instead of hashCode().

IdentityHashMap<String, Integer> map = new IdentityHashMap<>();
String a = new String("key");
String b = new String("key");
map.put(a, 1);
map.put(b, 2);
map.size(); // 2 - different objects, same content but different identity

Use cases:

1. Object graph traversal: Detect cycles using object identity (not equality)
2. Serialization: Track already-serialized objects
3. Deep clone: Track already-copied objects
4. Proxy systems: Map by object identity

EnumMap<K extends Enum<K>, V>: HashMap optimized for enum keys using a compact array internally.

enum Day { MON, TUE, WED, THU, FRI, SAT, SUN }

EnumMap<Day, String> schedule = new EnumMap<>(Day.class);
schedule.put(Day.MON, "Team meeting");
schedule.put(Day.FRI, "Code review");

Advantages over HashMap with enum keys:

• Faster: Array indexed by enum ordinal – O(1) with no hash computation or collision
• Memory-efficient: Compact array representation
• Ordered: Iterates in enum declaration order
• No null keys (enum values can't be null)

Always prefer EnumMap over HashMap when keys are enums.

Key principles:

1. Pre-size collections: new ArrayList<>(expectedSize) or new HashMap<>(capacity/0.75)
2. ArrayList vs LinkedList: ArrayList almost always wins in practice (cache efficiency)
3. HashSet for contains(): O(1) vs ArrayList's O(n)
4. Avoid Iterator wrapping: Direct array iteration > stream > iterator for simple cases
5. Primitive collections: Use IntStream, or libraries like Eclipse Collections, Trove for primitive lists (avoid boxing)
6. Immutable collections: List.of() (Java 9) – more memory-efficient than ArrayList with no-modification
7. EnumMap/EnumSet: For enum-keyed maps and sets – very fast

*LinkedList remove is O(n) if you need to find the element first; O(1) with iterator.

HashMap worst case (all collisions): O(n). After Java 8 treeification: O(log n) worst case.

1. Use concurrent collections (best):

Map<K,V> map = new ConcurrentHashMap<>();
List<E> list = new CopyOnWriteArrayList<>();

1. Collections.synchronizedXxx() (synchronized wrapper):

List<String> syncList = Collections.synchronizedList(new ArrayList<>());
Map<K,V> syncMap = Collections.synchronizedMap(new HashMap<>());
// Must synchronize on collection for iteration:
synchronized (syncList) { for (String s : syncList) {} }

1. Manual synchronization: synchronized block or ReentrantLock

1. Immutability: List.of(), Collections.unmodifiableList() – no sync needed for reads

Choose based on use pattern:

• Read-heavy: CopyOnWriteArrayList / Collections.unmodifiableList()
• Write-heavy, all operations: Collections.synchronizedList() + external lock on iteration
• Concurrent read+write: ConcurrentHashMap, ConcurrentSkipListMap
• Queue for producer-consumer: BlockingQueue implementations
• Immutable: List.of(), Map.of() (Java 9+) – zero synchronization needed

java.util.Collections: Static methods for Collection operations:

• sort(), reverse(), shuffle(), rotate()
• min(), max(), frequency(), disjoint()
• unmodifiableList(), synchronizedList()
• singletonList(), emptyList(), nCopies()
• binarySearch() (on sorted list)
• fill(), copy(), swap()

java.util.Arrays: Static methods for array operations:

• sort(), parallelSort()
• binarySearch()
• copyOf(), copyOfRange()
• fill()
• equals(), deepEquals()
• toString(), deepToString()
• asList() – convert array to fixed-size List (backed by array, no add/remove)
• stream() – create stream from array

Any hash-based collection (HashMap, HashSet, HashTable) uses both:

1. hashCode() to find the bucket
2. equals() to confirm the key match

Breaking the contract causes:

• Missed lookups: equal objects land in different buckets (not found)
• Duplicate insertions: Two "equal" objects in same Set
• Memory leaks: Keys added but never removable

class BadKey {
    int value;
    @Override public boolean equals(Object o) { ... } // overrides equals
    // NO hashCode override!
}

Set<BadKey> set = new HashSet<>();
BadKey k1 = new BadKey(5);
set.add(k1);
set.contains(new BadKey(5)); // FALSE! Different hashCode → different bucket

1. Mutable key whose hashCode changes after insertion:

List<String> key = new ArrayList<>(List.of("a"));
map.put(key, "value");
key.add("b"); // hashCode changes! Entry now in wrong bucket!
// map.get(key) returns null even though key is in map
// Old entry is leaked - impossible to reach without iterating all buckets

1. Keys not removed: static HashMap accumulating entries indefinitely

1. ThreadLocal HashMap: If thread pool threads survive but application context changes, ThreadLocal + HashMap can leak class loaders in web app redeployment scenarios.

1. Inner class in static context: Anonymous inner class holding reference to outer instance.

ArrayList: Index-based, cursor variable incremented:

// Simplified Iterator for ArrayList
int cursor = 0;
public boolean hasNext() { return cursor < size; }
public E next() { 
    checkForModification(); // modCount check
    return elementData[cursor++]; 
}

HashMap: Iterates through buckets array, then through linked list/tree nodes within each bucket:

// Visits all capacity buckets even if empty (why iteration is O(capacity))
for (int i = 0; i < table.length; i++) {
    Node<K,V> node = table[i];
    while (node != null) { visit(node); node = node.next; }
}

LinkedList: Follows next pointers from head.

A structural modification is any operation that changes the size of a collection: add, remove, clear.

Modifying an element's value (via set()) is NOT a structural modification.

Why it matters: Fail-fast iterators track structural modifications via modCount. Any structural modification during iteration (except via the iterator itself) causes ConcurrentModificationException.

list.set(0, "new value"); // NOT structural, safe during iteration (via index, not iterator)
list.add("item");         // structural, throws if done during fail-fast iteration

Immutable: Cannot be modified (contents AND structure). Throw UnsupportedOperationException.

Java 9+ factory methods:

List<String> list = List.of("a", "b", "c"); // immutable
Set<String> set = Set.of("x", "y");         // immutable
Map<String, Integer> map = Map.of("a", 1, "b", 2); // immutable

Properties:

• No null elements/keys
• set(), add(), remove() throw UnsupportedOperationException
• Compact memory representation (no backing array overhead)
• Iteration order: List – insertion; Set/Map – unspecified

Guava: ImmutableList.of(), ImmutableMap.of(), ImmutableSet.of() – similar but with richer API.

Unmodifiable (Collections.unmodifiableList()): A view wrapping the original. Throws exception on mutation, but the underlying collection can still be mutated by direct reference.

List<String> original = new ArrayList<>(List.of("a", "b"));
List<String> unmod = Collections.unmodifiableList(original);

unmod.add("c"); // UnsupportedOperationException
original.add("c"); // OK! unmod now shows "a", "b", "c"

Immutable (List.of()): No underlying mutable collection. Truly frozen.

List<String> immutable = List.of("a", "b");
// No way to modify this - period

Use List.of() when you have control over creation. Use Collections.unmodifiableList() to expose a view of a mutable collection.

Deep vs shallow immutability: Both are shallowly immutable – can't add/remove elements, but the elements themselves can be mutated if they're mutable objects.

// List
List<String> list = List.of("a", "b", "c");
List<String> copy = List.copyOf(existingCollection);

// Set
Set<String> set = Set.of("x", "y", "z");
Set<String> copy = Set.copyOf(existingCollection);

// Map
Map<String, Integer> map = Map.of("a", 1, "b", 2); // up to 10 entries
Map<String, Integer> map = Map.ofEntries(
    Map.entry("a", 1),
    Map.entry("b", 2),
    Map.entry("c", 3)
    // no limit
);
Map<String, Integer> copy = Map.copyOf(existingMap);

Key characteristics vs old approaches:

• Throw NullPointerException for null elements/keys/values
• Throw IllegalArgumentException for duplicate elements/keys (in Set/Map)
• Unspecified iteration order (Set, Map)
• More memory-efficient than Arrays.asList() + Collections wrappers
• Serializable