Concurrent Task

QtConcurrent::task provides an alternative interface for running a task in a separate thread. The return value of the function is made available through the QFuture API.

If you want to just run a function in a separate thread without adjusting any parameters, use QtConcurrent::run as that lets you write less code. The QtConcurrent::task is designed for cases where you need to perform extra configurations steps.

This function is a part of the Qt Concurrent framework.

Fluent interface

The QtConcurrent::task returns an instance of an auxiliary class called QtConcurrent::QTaskBuilder. Normally, you don't need to create an instance of this class manually. The QtConcurrent::QTaskBuilder provides an interface to adjust different task parameters in a chain-like manner. This approach is known as a fluent interface.

You can just set the parameters you need and then kick a task off. In order to finalize the configuration of a task you must invoke QtConcurrent::QTaskBuilder::spawn. This function is non-blocking (i.e. returns a future object immediately), but it's not guaranteed that the task starts immediately. You can use the QFuture and QFutureWatcher classes to monitor the status of the task.

See more examples and explanations below.

Running a task in a separate thread

To run a function in another thread, use QtConcurrent::QTaskBuilder::spawn:

 QtConcurrent::task([]{ qDebug("Hello, world!"); }).spawn();

This will run a lambda function in a separate thread obtained from the default QThreadPool.

Passing arguments to the task

Invoking a function with arguments is done by passing them to QtConcurrent::QTaskBuilder::withArguments:

 auto task = [](const QString &s){ qDebug() << ("Hello, " + s); };
 QtConcurrent::task(std::move(task))
     .withArguments("world!")
     .spawn();

A copy of each argument is made at the point where QtConcurrent::QTaskBuilder::withArguments is called, and these values are passed to the thread when it begins executing the task. Changes made to the arguments after calling QtConcurrent::QTaskBuilder::withArguments are not visible to the thread.

If you want to run a function that accepts arguments by reference, you should use std::ref/cref auxiliary functions. These functions create thin wrappers around passed arguments:

 QString s("Hello, ");
 QtConcurrent::task([](QString &s){ s.append("world!"); })
     .withArguments(std::ref(s))
     .spawn();

Make sure that all wrapped objects live long enough. It is possible to get undefined behavior if a task outlives the object wrapped by std::ref/cref.

Returning values from the task

You can obtain the result of a task with the QFuture API:

 auto future = QtConcurrent::task([]{ return 42; }).spawn();
 auto result = future.result(); // result == 42

Note that QFuture::result() is a blocking call, it waits for the result to become available. Use QFutureWatcher to get a notification when the task has finished execution and the result is available.

In case you want to pass a result to another asynchronous task, you can use QFuture::then() to create a chain of dependent tasks. See the QFuture documentation for more details.

Additional API features

Using different types of callable objects

Strictly speaking, you can use any type of tasks and arguments that satisfy the following condition:

 std::is_invocable_v<std::decay_t<Task>, std::decay_t<Args>...>

You can use a free function:

 QVariant value(42);
 auto result = QtConcurrent::task(&qvariant_cast<int>)
                   .withArguments(value)
                   .spawn()
                   .result(); // result == 42

You can use a member function:

 QString result("Hello, world!");

 QtConcurrent::task(&QString::chop)
     .withArguments(&result, 8)
     .spawn()
     .waitForFinished(); // result == "Hello"

You can use a callable object with an operator():

 auto result = QtConcurrent::task(std::plus<int>())
                   .withArguments(40, 2)
                   .spawn()
                   .result() // result == 42

If you want to use an existing callable object, you need to either copy/move it to QtConcurrent::task or wrap it with std::ref/cref:

 struct CallableWithState
 {
     void operator()(int newState) { state = newState; }

     // ...
 };

 // ...

 CallableWithState object;

 QtConcurrent::task(std::ref(object))
    .withArguments(42)
    .spawn()
    .waitForFinished(); // The object's state is set to 42

Using custom thread pool

You can specify a custom thread pool:

 QThreadPool pool;
 QtConcurrent::task([]{ return 42; }).onThreadPool(pool).spawn();

Setting priority for a task

You can set the priority for a task:

 QtConcurrent::task([]{ return 42; }).withPriority(10).spawn();

If you don't need a future object, you can call QtConcurrent::QTaskBuilder::spawn(QtConcurrent::FutureResult::Ignore):

 QtConcurrent::task([]{ qDebug("Hello, world!"); }).spawn(FutureResult::Ignore);

You can access the promise object associated with the task by defining an additional argument of QPromise<T> & type inside the function. This additional argument must be the first argument passed to the function, and like in Concurrent Run With Promise mode, the function is expected to return void type. Result reporting is done through QPromise API:

 void increment(QPromise<int> &promise, int i)
 {
     promise.addResult(i + 1);
 }

 int result = QtConcurrent::task(&increment).withArguments(10).spawn().result(); // result == 11

Contents