QtConcurrent: Early stop a MappedReduced/FilteredReduced from inside

VTiTux

I am developping a "contains" like function, using the QtConcurrent MappedReduced function, to compute values from data stored in a list, then compare them to the value to search for.

To save time, I would like to stop the process, when I found the first match.

Is there a way to do that?

I tried to capture the QFuture, to cancel the process from the Reduce functor (that is call by only one thread at a time), but that crashes:

float compute(const QString &);
bool compare(const float &, const float &);
bool contains(const float &search, const QStringList &list)
{
    QFuture<bool> process;

    std::function<float(const QString &)> map = [](const QString &data) { return compute(data); };

    std::function<void(bool&, const float &)> reduce = [&search, &process](bool &found, const float &value)
    {
        if( compare(search, value) == true )
        {
            if( found == false )
                process.cancel();
            found = true;
            //throw found;
        }
    };

    process = QtConcurrent::mappedReduced<bool>(list, map, reduce, false);
    process.waitForFinished();
    return process.result();
}

Alternatively, I can share the result with the mapped functor, to skip the following computations, once a result has been found.

VTiTux

I found!
This was not a "multi threads" issue. The crash was from the return statement. This code works:

float compute(const QString &);
bool compare(const float &, const float &);
bool contains(const float &search, const QStringList &list)
{
    QFuture<bool> process;
    int count = 0;
    bool found = false;

    std::function<float(const QString &)> map = [](const QString &data) { return compute(data); };

    std::function<void(bool&, const float &)> reduce = [&found, &count, &search, &process](bool &, const float &value)
    {
        if( compare(search, value) == true )
        {
            if( found == false )
                process.cancel();
            found = true;
        }
        ++count;
    };

    process = QtConcurrent::mappedReduced<bool>(list, map, reduce);
    process.waitForFinished();
    qDebug() << count;
    return found;
}

(probably safer with a QFutureWatcher)

But I also read this old thread, and yeye_olive oriented me to this other solution:


float compute(const QString &);
bool compare(const float &, const float &);
bool contains(const float &search, const QStringList &list)
{
    int count = 0;
    bool found = false;
    QFutureWatcher<bool> watcher;
    QEventLoop loop;
    QObject::connect(&watcher, &QFutureWatcher<bool>::finished, &loop, &QEventLoop::quit);

    std::function<float(const QString &)> map = [](const QString &data) { return compute(data); };

    std::function<void(bool&, const float &)> reduce = [&found, &count, &search, &loop](bool &, const float &value)
    {
        if( compare(search, value) == true )
        {
            if( found == false )
                QMetaObject::invokeMethod(&loop, "quit", Qt::QueuedConnection);
            found = true;
        }
        ++count;
    };

    watcher.setFuture( QtConcurrent::mappedReduced<bool>(list, map, reduce) );
    loop.exec();
    watcher.cancel();
    watcher.waitForFinished();
    qDebug() << count;
    return found;
}

And finally, I implemented this solution, which save the reduce functor (the half of the threads):


class Result : public QObject
{
public:
    Result(QObject *parent=nullptr) : QObject(parent), mFound(false){}
signals:
    void stop();
public slots:
    void found(void)
    {
        mFound = true;
        emit stop();
    }
public:
    bool mFound;
    Q_OBJECT
};

float compute(const QString &);
bool compare(const float &, const float &);
bool contains(const float &search, const QStringList &list)
{
    int count = 0;
    Result result;
    QFutureWatcher<bool> watcher;
    QEventLoop loop;
    QObject::connect(&watcher, &QFutureWatcher<bool>::finished, &loop, &QEventLoop::quit);
    QObject::connect(&result, &Result::stop, &loop, &QEventLoop::quit);

    std::function<float(const QString &)> map = [&result, &count, &search](const QString &data)
    {
        ++count;
        if( compare(search, compute(data)) == true )
            QMetaObject::invokeMethod(&result, "found", Qt::QueuedConnection);
        return false;    // need to return a value
    };

    watcher.setFuture( QtConcurrent::mapped(list, map) );
    loop.exec();
    watcher.cancel();
    watcher.waitForFinished();
    qDebug() << count;
    return result.mFound;
}

Are these solutions totally "thread-safe"? I don't really have experience in "multi-threading" coding.

SGaist

Hi,

Out of curiosity, why implement it that way ? You are in fact blocking your application because you are waiting on the result.

VTiTux

No, it's not about a performance issue.
I was just not satisfied with leaving this function running until the end, since I already had my result and the rest will not bring me anything more.
If I need to write it in a simple 'foreach' loop, I would write something like

for( value : list )
    if( value == search )
        return true;
return false;

And I wanted to keep the same spirit with 'mappedReduced' (as long as this doesn't violate any "thread-rules" and without having to "hack" QtConcurrent).

But yes, in fact I'm blocking my application untile I receive the results. My other methods that need to perform an action on all the items use 'blockingMapped', 'blockingMappedReduced' and 'blockingFiltered'. It's for a library that I am writing. And the applications that will use it are procedurale, not a GUI (or, maybe later...who knows?). But they may have to handle very long lists, and the 'compute' function are heavy. So QtConcurrent is a welcome help.

I think I'll keep the first solution, the most elegant one in my eyes, if I don't get any warning.