Unlike C++ or Java, JavaScript does not support threads. WebWorkers aren’t threads, since they don’t allow shared state between each other. They are more like processes: they are isolated and can communicate via asynchronous channels. There’s also no synchronization primitives in WebWorkers.

However, Java relies on threads heavily. There’s synchronous Thread.sleep method, but no asynchronous alternative in standard library. Most of classic IO is synchronous. There are libraries which use threads. And even if you write a new code, specially for TeaVM, it’s still idiomatic for Java developer to create threads to perform several computations simultaneously.

Fortunately, TeaVM comes with solution for this. TeaVM supports coroutines. Coroutines are special kinds of methods that can be suspended by runtime in a certain point and then resumed from this point. In particular, TeaVM uses coroutines to emulate Thread class together with few simple synchronization primitives.

In most cases you don’t need anything special to use coroutines. You just create a new thread and start it. You can use synchronized keyword, Object.wait, Object.notify, and so on. The only case you should care of is interaction with JavaScript API. The remaining part of this page describes some low-level details of TeaVM coroutines to help you to use them properly in some corner cases.

@Async annotation

Most of JavaScript APIs are asynchronous, while Java often provides synchronous APIs. For example, let’s we have a JavaScript function foo that’s been called like this:

foo("some argument", function(result) {

We want to call it synchronously from Java. In this case we need @Async annotation:

public static native String foo(String arg);
private static void foo(String arg, AsyncCallback<String> callback) {
    fooAsync(arg, result -> callback.complete(result));

@JsBody(params = { "arg", "callback" }, script = "return foo(arg, callback);")
public static native void fooAsync(String arg, JsConsumer<String> callback);

So, to expose an asynchronous JavaScript API as a synchronous Java API, you need:

  • to declare a native method;
  • to mark this method with @Async annotation;
  • to declare a second method in the same class with the same name, and almost same signature, except for the second method should return void and should take additional parameter of AsyncCallback<T> class (where T should correspond to return type of the first method);
  • call either callback.complete or callback.error when operation completes.

The following example shows how to perform HTTP requests synchronously:

public class Ajax {
    public static native String get(String url) throws IOException;
    private static void get(String url, AsyncCallback<String> callback) {
        XMLHttpRequest xhr = XMLHttpRequest.create();"get", url);
        xhr.setOnReadyStateChange(() -> {
            if (xhr.getReadyState() != XMLHttpRequest.DONE) {
            int statusGroup = xhr.getStatus() / 100;
            if (statusGroup != 2 && statusGroup != 3) {
                callback.error(new IOException("HTTP status: " + 
                        xhr.getStatus() + " " + xhr.getStatusText()));
            } else {

Interaction with JavaScript.

Sometimes JavaScript API expects you to pass function. For example, expects mapping function. This API expects passed function to complete immediately and produce result. However, when passing Java lambda to such function, this lambda may be a coroutine. To prevent this, TeaVM prohibits to pass coroutine to a JavaScript function. See following example:

@JsBody(params = { "array", "mapper" }, body = "return;")
static native <T, S> JsArray<S> map(JsArray<T> array, JsMapper<T, S> mapper);

static JsArray<String> getAll(JsArray<String> urls) {
    map(urls, url -> {  // <-- TeaVM throws exception here
        return Ajax.get(url);

Improve this page