QDataflowCanvas

A Qt widget for displaying and editing a dataflow graph.

It is based on the Graphics View Framework and it uses the Model/View architecture.

Building a simple Dataflow application

Start by creating a UI with a QSpinBox and a QPushButton which will be used to send input to one of the nodes (the source object). Create also a read-only QLineEdit which will be used to display the data received by the sink object.

The dataflowMetaObject is a volatile field used to attach the dataflow logic to the graph.

We create a subclass of QDataflowMetaObject for the DFSource object. In the bool init(QStringList args) method we validate the object and set up the number of inlets/outlets according to object's functionality:

class DFSource : public QDataflowMetaObject
{
public:
    bool init(QStringList args)
    {
        Q_UNUSED(args);

        setInletCount(0);
        setOutletCount(1);

        return true;
    }
};

Similarily, we create a subclass of QDataflowMetaObject for the DFSink object. The void onDataReceved(int inlet, void *data) method will be called when data is received on an inlet. In the case of the sink object, we want to display the data in the QLineEdit object:

class DFSink : public QDataflowMetaObject
{
public:
    DFSink(QLineEdit *e) : e_(e) {}

    bool init(QStringList args)
    {
        Q_UNUSED(args);

        setInletCount(1);
        setOutletCount(0);

        return true;
    }

    void onDataReceved(int inlet, void *data)
    {
        if(inlet == 0)
        {
            e_->setText(QString::number(reinterpret_cast<long>(data)));
        }
    }

private:
    QLineEdit *e_;
};

Now we create an object to perform mathematical operations. Instead of creating many classes, we create just one subclass performing various math operations, depending on the value of the op field:

class DFMathBinOp : public QDataflowMetaObject
{
public:
    bool init(QStringList args)
    {
        s = 0;

        setInletCount(2);
        setOutletCount(1);

        op = args[0];

        if(args.length() > 1)
            s = args[1].toLong();

        return true;
    }

    void onDataReceved(int inlet, void *data)
    {
        if(inlet == 0)
        {
            int r = reinterpret_cast<long>(data);
            if(op == "add") r = r + s;
            if(op == "sub") r = r - s;
            if(op == "mul") r = r * s;
            if(op == "div") r = r / s;
            if(op == "pow") r = pow(r, s);
            sendData(0, reinterpret_cast<void*>(r));
        }
        else if(inlet == 1)
        {
            s = reinterpret_cast<long>(data);
        }
    }

private:
    QString op;
    int s;
};

The DFMathBinOp object has two inlets, because it implements binary mathematical operators. If we want to compute 2 + 3, we first send 3 to the right inlet, which will store 3 in its internal status variable, and then send 2 to the left inlet, which will trigger the computation and output the result on the first outlet.

This pattern is common in dataflow programming environments: the leftmost inlet (which will trigger the output) is the "hot" inlet, and the other inlets are "cold" inlets.

Now we need to attach the correct QDataflowMetaObject to the newly created instances of QDataflowModelNode. To do so, we create the void setupNode(QDataflowModelNode *node) method, which will parse (tokenize) the object creation arguments, assign dataflowMetaObject, and mark the object valid:

void MainWindow::setupNode(QDataflowModelNode *node)
{
    QStringList toks = node->text().split(QRegExp("(\\ |\\t)"));
    if(!classList.contains(toks[0]))
    {

classList is a QStringList with all the available dataflow object classes. It will be used also for completion.

        node->setValid(false);
        return;
    }
    if(toks[0] == "source")
    {
        sourceNode = node;

we track the last source node, so we have an object to call sendData() when the QPushButton is clicked.

        node->setDataflowMetaObject(new DFSource());
    }
    else if(toks[0] == "sink")
    {
        node->setDataflowMetaObject(new DFSink(result));
    }
    else
    {
        node->setDataflowMetaObject(new DFMathBinOp());
    }
    bool ok = node->dataflowMetaObject()->init(toks);
    if(!ok)
    {
        qDebug() << "initialization failed for:" << node->text();
    }
    node->setValid(ok);
}

The next thing we need is a method to feed data to the source object:

void MainWindow::processData()
{
    long x = input->value();
    sourceNode->dataflowMetaObject()->sendData(0, reinterpret_cast<void*>(x));
}

Also, a method for performing text completion:

void MainWindow::complete(QString txt, QStringList &completionList)
{
    foreach(QString className, classList)
        if(className.startsWith(txt))
            completionList << className;
}

And slots for when a node is added or modified in the model:

void MainWindow::onNodeAdded(QDataflowModelNode *node)
{
    setupNode(node);
}

void MainWindow::onNodeTextChanged(QDataflowModelNode *node, QString text)
{
    Q_UNUSED(text);
    setupNode(node);
}

Now proceed to the creation of the QDataflowCanvas:

QDataflowCanvas *canvas = new QDataflowCanvas(mainWindow);

initialize completion:

QStringList classList;
classList << "add" << "sub" << "mul" << "div" << "pow" << "source" << "sink";
canvas->setCompletion(this);

where this inherits also from QDataflowTextCompletion and implements the void complete(QString txt, QStringList &list) method.

Connect signals:

QObject::connect(sendButton, &QPushButton::clicked, this, &MainWindow::processData);
QObject::connect(model, &QDataflowModel::nodeTextChanged, this, &MainWindow::onNodeTextChanged);
QObject::connect(model, &QDataflowModel::nodeAdded, this, &MainWindow::onNodeAdded);

See mainwindow.ui/h/cpp for a complete example.

Note: in the widget, it is possible to create new objects by double clicking on an empty area, or edit existing objects by double clicking objects. Objects and connections can be removed by selecting them and hitting backspace. Connections are created by dragging from outlet to inlet.

Using the Model API

We can use the Model API to create or edit the dataflow graph programatically:

QDataflowModel *model = ...;
QDataflowModelNode *source = new QDataflowModelNode(model, QPoint(100, 50), "source", 0, 0);
QDataflowModelNode *add = new QDataflowModelNode(model, QPoint(100, 100), "add 5", 0, 0);
QDataflowModelNode *sink = new QDataflowModelNode(model, QPoint(100, 150), "sink", 0, 0);
model->addNode(source);
model->addNode(add);
model->addNode(sink);
model->addConnection(source->outlet(0), add->inlet(0));
model->addConnection(add->outlet(0), sink->inlet(0));

The model will emit signals for when a node/connection is added, removed, and also when a node change its validity status, position, text, inlet count, and outlet count.

Contribute

If you want to contribute with development, fork and make a pull requests. PRs are very welcome!

If you want to donate, use one of the following methods:

BTC: 34RqXSKCgPrVJ8Q266ahiWx3TfV2xAdNKv
BCH: bitcoincash:ppv36ls4fk82k4actzsl7unrfx5k35uzzu203rt2u9
ETH: 0x3a1Fe92C19609835De3975F64fca4c4aA9B0BDf3
LTC: LYTUr4xiMjwii88J8sZ44wWu8GCaCQVx3W
XMR: 4GdoN7NCTi8a5gZug7PrwZNKjvHFmKeV11L6pNJPgj5QNEHsN6eeX3DaAQFwZ1ufD4LYCZKArktt113W7QjWvQ7CWCSnRSymAvuG1vvpqi

qtdataflow / qdataflowcanvas Goto Github PK