ValueError: Can't clean for JSON for networkx nodes that are objects.

Execute the following in a jupyter lab notebook and you’ll get ‘ValueError: Can’t clean for JSON’. I believe this should work, and is a bug. It works fine if I use nx.draw(G) instead of ipycytoscape.

import ipycytoscape
import networkx as nx

class Node:
    def __init__(self, name):
        self.name = name

n1 = Node("node 1")
n2 = Node("node 2")
        
G = nx.Graph()

G.add_node(n1)
G.add_node(n2)

G.add_edge(n1, n2)

undirected = ipycytoscape.CytoscapeWidget()
undirected.graph.add_graph_from_networkx(G)
undirected

[MridulS]

Is there any specific use case for having custom classes? (unless the idea here is to mirror exactly nx.draw). The str() argument will make it work with ipycytoscape but we will essentially lose all the information of the object. (which should be okay if we want to make ipycytoscape compatible with hashable node objects).

If the idea here is to pass information from the Node custom object to the ipycytoscape graph object then we can do something like

import ipycytoscape

class CustomNode(ipycytoscape.Node):
    def __init__(self, i, test):
        super().__init__()
        self.data['id'] = i
        self.data['test_var'] = test

ipyGraph = ipycytoscape.CytoscapeWidget()
ipyGraph.graph.add_node(CustomNode(1, 'boo'))

and we can access the newly created custom nodes by something like

>>> ipyGraph.graph.nodes
[CustomNode(data={'id': 1, 'test_var': 'boo'}, position={})]

[jaydonnell]

gotcha, sorry, I missed the call to str() in your’s the first time i looked at it.