qt-dataflow 0.2.3

A base for custom visual programming enviroments

This package tries to provide components for building your own visual programming environment. The authors aim is to make his data analysis tool available to his colleagues who don’t know programming or Python.

Because a standard gui is not very flexible, this projects tries to make visual canvas on which the dataflow can be defined and modified. Extensibility is given through simply adding or modifying Nodes.

This project is inspired by Orange - where i did not see an easy way to just use the canvas part (also: license differences). Also the design tries to be more flexible.

Requirements

It is made with Python 2.7. Not tested for lower versions or Python 3 (patches welcome). It should work with PySide and PyQt, but at the moment, the imports need to be manually changed.

The examples may have additional requirements:

• numpy

• matplotlib

• pyqtgraph

Examples

See example.py for an simple example using icons which react to double click. To make a connection draw from one nodes termial to another (only out-> in is allowed).

• example_widget uses widgets on the canvas directly, it also implements a simple callback. Note how the label updates after changing the SpinBox-value.

• example_pyqtgraph need also the pyqtgraph package. It plot directly on the canvas.

• example_matplotlib_on_canvas does the same, but uses matplotlib via a temporary file.

Code Example

To make custom nodes you need to subclass Node. It must return a NodeView via its ‘get_view’ method. The following example implements a Node which make a random number.

class RandomNumber(Node):
    """
    A test node which outputs a random number. Widget allow to set the number.
    """
    def __init__(self):
        super(DataGenNode, self).__init__()
        #Node type/name
        self.node_type = 'Random Array'
        #Icon_path is needed for the PixmapNodeView
        self.icon_path = 'icons/onebit_11.png'
        #The makes the node have an output terminal.
        self.generates_output = True

    def get_view(self):
        return PixmapNodeView(self)

    def get(self):
        #Method which can be called by other nodes. The name is just
        #a convention.
        num = [random.random() for i in range(self.num_points)]
        return num

    def show_widget(self):
        #Method called by double clicking on the icon.
        int, ok = Qt.QtGui.QInputDialog.getInteger(None, 'Input Dialog',
                                          'Number of Points', self.num_points)
        if ok:
            self.num_points = int

A node saves its connections in node.in_conn and node.out_conn. Also note, that each node view must be a child of a QGraphicsItem and NodeView.

Structure

In model the base Node- and Schema-classes are found. In view are some view available. gui contains some additional ready to use elements.

Todo

• add different icons (simple)

• nicer toolbar (drag-n-drop would be nice)

• test persistence, define a stable protocol if pickling does not work

• make an example with less requirements.

• checking and introducing a connection type

• move some logic for allowing or denying connections from SchemaView to the NodeView.

• checking and improving compatibility with different Python versions.

• better documentation

• make number of terminals variable.

• …

Coding Style

This projects tries to follow PEP8.

License

Example icons are from http://www.icojam.com/blog/?p=177 (Public Domain).

BSD - 3 clauses, see license.txt.