Java Virtual Threads (Project Loom) Cheat Sheet

// Direct creationThread vt = Thread.ofVirtual().name("worker-1").start(() -> {    System.out.println("running on " + Thread.currentThread());});vt.join();// Factory for use with ExecutorService-style codeThreadFactory factory = Thread.ofVirtual().name("worker-", 0).factory();// Unstarted, then start laterThread t = Thread.ofVirtual().unstarted(() -> doWork());t.start();

try (ExecutorService executor = Executors.newVirtualThreadPerTaskExecutor()) {    List<Future<String>> futures = new ArrayList<>();    for (int i = 0; i < 10_000; i++) {        int id = i;        futures.add(executor.submit(() -> fetch("https://example.com/" + id)));    }    for (var f : futures) {        System.out.println(f.get());    }} // executor.close() waits for all tasks, one virtual thread per submitted task

// java.util.concurrent.StructuredTaskScope (preview in recent JDKs)try (var scope = new StructuredTaskScope.ShutdownOnFailure()) {    Subtask<String> user = scope.fork(() -> fetchUser(id));    Subtask<String> orders = scope.fork(() -> fetchOrders(id));    scope.join();           // wait for both, or fail fast on first exception    scope.throwIfFailed();  // rethrow if any subtask failed    return new Response(user.get(), orders.get());}

// java.lang.ScopedValue (finalized JDK 21+/25 depending on release)final static ScopedValue<String> REQUEST_ID = ScopedValue.newInstance();void handleRequest(String id) {    ScopedValue.where(REQUEST_ID, id).run(() -> {        // REQUEST_ID.get() is visible to this thread and any it forks        process();    });}void process() {    log.info("handling " + REQUEST_ID.get());}// Unlike ThreadLocal:// - value is bound only for the dynamic scope of run()/call(), then unbound// - no inheritance leaks between pooled/reused threads (moot anyway since//   virtual threads aren't pooled), no explicit remove() needed// - immutable within the scope: no accidental mutation from nested code

// Default carrier pool size = Runtime.availableProcessors()// Override at JVM startup:// -Djdk.virtualThreadScheduler.parallelism=16// -Djdk.virtualThreadScheduler.maxPoolSize=256// -Djdk.virtualThreadScheduler.minRunnable=8// Inspect at runtime:// jcmd <pid> Thread.dump_to_file -format=json threads.json// jcmd <pid> VM.info | grep -i virtual// Rule of thumb: parallelism should roughly match available CPU cores for// CPU-bound work interleaved with virtual threads; raising it does NOT help// I/O-bound throughput once carriers are no longer the bottleneck, since a// blocking virtual thread unmounts from its carrier automatically.

Semaphore permits = new Semaphore(50); // cap concurrent in-flight callstry (var executor = Executors.newVirtualThreadPerTaskExecutor()) {    List<Future<String>> futures = new ArrayList<>();    for (String url : urls) {        futures.add(executor.submit(() -> {            permits.acquire();            try {                return httpClient.send(url);            } finally {                permits.release();            }        }));    }    for (var f : futures) System.out.println(f.get());}// Blocking on permits.acquire() is cheap here: it parks the virtual thread// without pinning or wasting an OS thread, unlike blocking a platform thread.

// Collect all successful results, throw if any subtask failedvar joiner = StructuredTaskScope.Joiner.<String>allSuccessfulOrThrow();try (var scope = StructuredTaskScope.open(joiner)) {    scope.fork(() -> fetchFromMirrorA());    scope.fork(() -> fetchFromMirrorB());    scope.fork(() -> fetchFromMirrorC());    List<String> results = scope.join();}// Race: return the first successful result, cancel the restvar race = StructuredTaskScope.Joiner.<String>anySuccessfulResultOrThrow();try (var scope = StructuredTaskScope.open(race)) {    scope.fork(() -> fetchFromMirrorA());    scope.fork(() -> fetchFromMirrorB());    String first = scope.join();}// Cancellation of the losing subtasks is automatic and propagates interrupts// to any blocking I/O they're doing, which unparks their virtual threads.