NetworkX Viewer

Introduction

NetworkX Viewer provides a basic interactive GUI to view networkx graphs. In addition to standard plotting and layout features as found natively in networkx, the GUI allows you to:

• Drag nodes around to tune the default layout
• Show and hide nodes
• Filter nodes
• Pan and zoom
• Display nodes only within a certain number of hops ("levels") of a "home node"
• Display and highlight the shortest path between two nodes. Nodes around the path can also be displayed within a settable number of levels
• Intelligently find and display nodes near displayed nodes using "Grow" and "Grow Until" functions
• Use attributes stored in the graph's node and edge dictionaries to customize the appearance of the node and edge tokens in the GUI
• Mark nodes and edges for reference
• Support for both nx.Graph and nx.MultiGraph

A typical usage might be:

import networkx as nx
from networkx_viewer import Viewer

G = nx.MultiGraph()
G.add_edge('a','b')
G.add_edge('b','c')
G.add_edge('c','a',0,{'fill':'green'})
G.add_edge('c','d')
G.add_edge('c','d',1,{'dash':(2,2)})
G.node['a']['outline'] = 'blue'
G.node['d']['label_fill'] = 'red'

app = Viewer(G)
app.mainloop()

The result will be:

Installation

NetworkX Viewer is hosted on PyPI and can be installed by simply doing the following.

pip install networkx_viewer

NetworkX Viewer requires NetworkX version 1.4 or greater.

Using the GUI

The default layout for the nodes is to use nx.spring_layout. While this layout is pretty good, it is not perfect, so the GUI supports standard features like rearranging the nodes, panning, and zooming.

By default, the viewer will display the entire graph on initialization. However, most of the power in the GUI comes in showing a subset of the graph. You can specify a subgraph to display using:

app = Viewer(G, home_node='a', levels=1)

Constructing a plot

On the right of the screen is a box to enter node(s) to graph.

• If you enter a single node, that node plus nodes upto "Neighbor Levels" levels will be ploted.
• If you enter a pair of nodes, the shortest path between the nodes will be found. Neighbors around the path upto "Neighbor Levels" will also be plotted
• If you enter three or more nodes, all those nodes will graphed out to "Neighbor Levels"

You may either "Build New" or "Add to Existing." If you choose to add to the existing plot, and a path exists between the new node and your existing display island, you will be prompted if you'd like the program to plot the intermediate nodes.

Right-click functionality

Several actions can be taken by right-clicking on nodes and edges, including

• Grow: Display all nodes connected to this node that may not be currently displayed. A node which does not have all of its neighbors currently displayed will have a grey label.

• Grow Until: This lets you find the path between this node and a node with a desired attribute. This is done by provided a lambda function which may accept the following arguments:

  • u - Name of the Node
  • d - The data dictionary for the node (ie, the contents of G.node[u])

  Say we had a graph which has nodes which are actors and edges which are the movie the two actors were both in. A classic example might be to use the lambda function:

  u=='Kevin Bacon'

  to find who the degrees of seperation are between the right-clicked actor and Kevin Bacon.

• Hide
• Hide Behind: Hide radial sections of the graph that are behind the edge formed by the node the cursor is currently over and the node in the menu. Note: if the graph is not radial behind the selected node, this item is greyed out in the dropdown.

You can also simply hover over a node and press the shortcut key ("G" for grow, "H" for hide, etc...) to activate the action.

Filtering

You can filter the nodes to display based on the attributes a node possess. This is done in a simmilar manner to how Grow Until works, as described above. You must write a lambda function which accepts the following paramaters:

• u - Name of the Node
• d - The data dictionary for the node (ie, the contents of G.node[u])

When this lambda function evaluates to False, the node is hidden, otherwise the node is displayed. Multiple Filters are ANDed together.

Node and Edge Attributes

The attributes (ie, the dictionary stored in G.node[u] and G.edge[u][v]) are displayed in the lower-right section of the screen.

At this time, neither the attributes not the graph's nodes/edges themselves are editable through the GUI. The GUI is read-only. You should programatically create/update the graph by doing the following:

G = app.canvas.dataG
# code to edit graph
app.canvas.refresh()

Using the Tk Pass-through

If the data dictionary stored in the graph for an edge or node contains a key that can be used by Tk, the token will be customized as such. Specifically,

• If a node contains a key used to configure Tkinter.Canvas.create_oval, it will be used to customize the node's marker (ie, the red circle).
• If a node contains a key prefixed with "label_" (for example, "label_font" or "label_fill") that can be used to configure Tkinter.Canvas.create_text, it will be used to customize the node's label.
• If an edge contains a key which can be used by Tkinter.Canvas.create_line, it will be used to customize the edge's display properties.

Expanding and Customizing the GUI

The core Tk widget that is implemented by networkx_viewer is the GraphCanvas widget. If you simply wish to use the GUI as presented as part of a larger application, you can just instantiate the canvas, passing it the graph to display as an argument and pack or grid it into your Tk application like any other canvas widget.

If you wish to change the tokens used for edges or nodes, subclass NodeToken or EdgeToken and pass as an argument into the GraphCanvas as such. For example:

import Tkinter as tk
import networkx as nx
from networkx_viewer import NodeToken, GraphCanvas

class CustomNodeToken(NodeToken):
    def render(self, data, node_name):
        """Example of custom Node Token
        Draw a circle if the node's data says we are a circle, otherwise
        draw us as a rectangle.  Also, if data contains a color key,
        use that as our color (default, red)
        """
        # For our convenience, the render method is called with the
        #  graph's data attributes and the name of the node in the graph

        # Note that NodeToken is a subclass of Tkinter.Canvas, so we
        #  simply draw on ourselves to create the apperance for the node.

        # Make us 50 pixles big
        self.config(width=50, height=50)

        # Set color and other options
        marker_options = {'fill':       data.get('color','red'),
                          'outline':    'black'}

        # Draw circle or square, depending on what the node said to do
        if data.get('circle', None):
            self.create_oval(0,0,50,50, **marker_options)
        else:
            self.create_rectangle(0,0,50,50, **marker_options)

class ExampleApp(tk.Tk):
    def __init__(self, graph, **kwargs):
        tk.Tk.__init__(self)

        self.canvas = GraphCanvas(graph, NodeTokenClass=CustomNodeToken,
            **kwargs)
        self.canvas.grid(row=0, column=0, sticky='NESW')

G = nx.path_graph(5)
G.node[2]['circle'] = True
G.node[3]['color'] = 'blue'

app = ExampleApp(G)
app.mainloop()

Development Status

As of September 2014, networkx_viewer is under active development. Bugs or feature requests should be submitted to the github issue tracker.