Concurrent Run

The QtConcurrent::run() function runs a function in a separate thread. The return value of the function is made available through the QFuture API.

QtConcurrent::run() is an overloaded method. You can think of these overloads as slightly different modes. In basic mode, the function passed to QtConcurrent::run() is able to report merely a single computation result to its caller. In run with promise mode, the function passed to QtConcurrent::run() can make use of the additional QPromise API, which enables multiple result reporting, progress reporting, suspending the computation when requested by the caller, or stopping the computation on the caller's demand.

This function is a part of the Qt Concurrent framework.

Concurrent Run (basic mode)

The function passed to QtConcurrent::run() may report the result through its return value.

Running a Function in a Separate Thread

To run a function in another thread, use QtConcurrent::run():

extern void aFunction();
QFuture<void> future = QtConcurrent::run(aFunction);

This will run aFunction in a separate thread obtained from the default QThreadPool. You can use the QFuture and QFutureWatcher classes to monitor the status of the function.

To use a dedicated thread pool, you can pass the QThreadPool as the first argument:

extern void aFunction();
QThreadPool pool;
QFuture<void> future = QtConcurrent::run(&pool, aFunction);

Passing Arguments to the Function

Passing arguments to the function is done by adding them to the QtConcurrent::run() call immediately after the function name. For example:

extern void aFunctionWithArguments(int arg1, double arg2, const QString &string);

int integer = ...;
double floatingPoint = ...;
QString string = ...;

QFuture<void> future = QtConcurrent::run(aFunctionWithArguments, integer, floatingPoint, string);

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

Note that QtConcurrent::run does not support calling overloaded functions directly. For example, the code below won't compile:

void foo(int arg);
void foo(int arg1, int arg2);
...
QFuture<void> future = QtConcurrent::run(foo, 42);

The easiest workaround is to call the overloaded function through lambda:

QFuture<void> future = QtConcurrent::run([] { foo(42); });

Or you can tell the compiler which overload to choose by using a static_cast:

QFuture<void> future = QtConcurrent::run(static_cast<void(*)(int)>(foo), 42);

Or qOverload:

QFuture<void> future = QtConcurrent::run(qOverload<int>(foo), 42);

Returning Values from the Function

Any return value from the function is available via QFuture:

extern QString functionReturningAString();
QFuture<QString> future = QtConcurrent::run(functionReturningAString);
...
QString result = future.result();

If you don't need the result (for example, because the function returns void), using the QThreadPool::start() overload taking a function object is more efficient.

As documented above, passing arguments is done like this:

extern QString someFunction(const QByteArray &input);

QByteArray bytearray = ...;

QFuture<QString> future = QtConcurrent::run(someFunction, bytearray);
...
QString result = future.result();

Note that the QFuture::result() function blocks and waits for the result to become available. Use QFutureWatcher to get notification when the function has finished execution and the result is available.

Additional API Features

Using Member Functions

QtConcurrent::run() also accepts pointers to member functions. The first argument must be either a const reference or a pointer to an instance of the class. Passing by const reference is useful when calling const member functions; passing by pointer is useful for calling non-const member functions that modify the instance.

For example, calling QByteArray::split() (a const member function) in a separate thread is done like this:

// call 'QList<QByteArray>  QByteArray::split(char sep) const' in a separate thread
QByteArray bytearray = "hello world";
QFuture<QList<QByteArray> > future = QtConcurrent::run(&QByteArray::split, bytearray, ' ');
...
QList<QByteArray> result = future.result();

Calling a non-const member function is done like this:

// call 'void QImage::invertPixels(InvertMode mode)' in a separate thread
QImage image = ...;
QFuture<void> future = QtConcurrent::run(&QImage::invertPixels, &image, QImage::InvertRgba);
...
future.waitForFinished();
// At this point, the pixels in 'image' have been inverted

Using Lambda Functions

Calling a lambda function is done like this:

QFuture<void> future = QtConcurrent::run([=]() {
    // Code in this block will run in another thread
});
...

Calling a function modifies an object passed by reference is done like this:

static void addOne(int &n) { ++n; }
...
int n = 42;
QtConcurrent::run(&addOne, std::ref(n)).waitForFinished(); // n == 43

Using callable object is done like this:

struct TestClass
{
    void operator()(int s1) { s = s1; }
    int s = 42;
};

...

TestClass o;

// Modify original object
QtConcurrent::run(std::ref(o), 15).waitForFinished(); // o.s == 15

// Modify a copy of the original object
QtConcurrent::run(o, 42).waitForFinished(); // o.s == 15

// Use a temporary object
QtConcurrent::run(TestClass(), 42).waitForFinished();

// Ill-formed
QtConcurrent::run(&o, 42).waitForFinished(); // compilation error

Concurrent Run With Promise

The Run With Promise mode enables more control for the running task compared to basic mode of QtConcurrent::run(). It allows progress reporting of the running task, reporting multiple results, suspending the execution if it was requested, or canceling the task on caller's demand.

The mandatory QPromise argument

The function passed to QtConcurrent::run() in Run With Promise mode is expected to have an additional argument of QPromise<T> & type, where T is the type of the computation result (it should match the type T of QFuture<T> returned by QtConcurrent::run()), like e.g.:

extern void aFunction(QPromise<void> &promise);
QFuture<void> future = QtConcurrent::run(aFunction);

The promise argument is instantiated inside the QtConcurrent::run() function, and its reference is passed to the invoked aFunction, so the user doesn't need to instantiate it, nor pass it explicitly when calling QtConcurrent::run() in this mode.

The additional argument of QPromise type always needs to appear as a first argument on function's arguments list, like:

extern void aFunction(QPromise<void> &promise, int arg1, const QString &arg2);

int integer = ...;
QString string = ...;

QFuture<void> future = QtConcurrent::run(aFunction, integer, string);

Reporting results

In contrast to basic mode of QtConcurrent::run(), the function passed to QtConcurrent::run() in Run With Promise mode is expected to always return void type. Result reporting is done through the additional argument of QPromise type. It also enables multiple result reporting, like:

void helloWorldFunction(QPromise<QString> &promise)
{
    promise.addResult("Hello");
    promise.addResult("world");
}

QFuture<QString> future = QtConcurrent::run(helloWorldFunction);
...
QList<QString> results = future.results();

Suspending and canceling the execution

The QPromise API also enables suspending and canceling the computation, if requested:

void aFunction(QPromise<int> &promise)
{
    for (int i = 0; i < 100; ++i) {
        promise.suspendIfRequested();
        if (promise.isCanceled())
            return;

        // computes the next result, may be time consuming like 1 second
        const int res = ... ;
        promise.addResult(res);
    }
}

QFuture<int> future = QtConcurrent::run(aFunction);

... // user pressed a pause button after 10 seconds
future.suspend();

... // user pressed a resume button after 10 seconds
future.resume();

... // user pressed a cancel button after 10 seconds
future.cancel();

The call to future.suspend() requests the running task to hold its execution. After calling this method, the running task will suspend after the next call to promise.suspendIfRequested() in its iteration loop. In this case the running task will block on a call to promise.suspendIfRequested(). The blocked call will unblock after the future.resume() is called. Note, that internally suspendIfRequested() uses wait condition in order to unblock, so the running thread goes into an idle state instead of wasting its resources when blocked in order to periodically check if the resume request came from the caller's thread.

The call to future.cancel() from the last line causes that the next call to promise.isCanceled() will return true and aFunction will return immediately without any further result reporting.

Progress reporting

It's also possible to report the progress of a task independently of result reporting, like:

void aFunction(QPromise<int> &promise)
{
    promise.setProgressRange(0, 100);
    int result = 0;
    for (int i = 0; i < 100; ++i) {
        // computes some part of the task
        const int part = ... ;
        result += part;
        promise.setProgressValue(i);
    }
    promise.addResult(result);
}

QFutureWatcher<int> watcher;
QObject::connect(&watcher, &QFutureWatcher::progressValueChanged, [](int progress){
    ... ; // update GUI with a progress
    qDebug() << "current progress:" << progress;
});
watcher.setFuture(QtConcurrent::run(aFunction));

The caller installs the QFutureWatcher for the QFuture returned by QtConcurrent::run() in order to connect to its progressValueChanged() signal and update e.g. the graphical user interface accordingly.

Invoking functions with overloaded operator()()

By default, QtConcurrent::run() doesn't support functors with overloaded operator()() in Run With Promise mode. In case of overloaded functors the user needs to explicitly specify the result type as a template parameter passed to QtConcurrent::run(), like:

struct Functor {
    void operator()(QPromise<int> &) { }
    void operator()(QPromise<double> &) { }
};

Functor f;
run<double>(f); // this will select the 2nd overload
// run(f);      // error, both candidate overloads potentially match