Lattice does not support all rustworkx graph representations. about qiskit-nature HOT 3 OPEN

rustworkx graph cans actually store arbitrary data as payloads of their nodes and edges.
You cannot possibly cover all possible such payloads and expect the code to extract a weight from it.

That said, the payload which is supported (i.e. just a plain weight) should probably be documented properly. Right now, the only mention of this fact is here:

Therefore, this is either a documentation issue or a feature request to support more types although I would not be sure how this would work in a general case because the Hamiltonian generators which use the lattice would need to be adapted to interpret non-numeric payloads, too.

from qiskit-nature.

This is relevant because networkx graphs only store weights as values in the dictionary and if Lattice() is given a NetworkX graphs it does the following:

if not isinstance(graph, PyGraph):
            _optionals.HAS_NETWORKX.require_now("Lattice construction from networkx.Graph")
            graph = networkx_converter(graph)
            

The network_converter() will keep the format networkx uses into the rustworkx graph.

I stumbled into this issue trying to take a weighted graph that was built elsewhere with NetworkX into a Lattice(). I had to manually convert it to rustworkx to make sure it wasn’t using a dictionary for the weights before building the lattice.

from qiskit-nature.

Here is an example.

G = nx.Graph()
edges=[[1,2],[0,3],[2,3],[0,1]]
for edge in edges:
    edge_1 = edge[0]
    edge_2 = edge[1]
    weight = 1.0
    G.add_edge(edge_1, edge_2,weight=weight)

general_lattice = Lattice(G)

Gives ValueError: Unsupported weight {'weight': 1.0} on edge with index 0.

To fix it we have to manually convert it to an adequately formatted rustworkx object. This works:

G = nx.Graph()
edges=[[1,2],[0,3],[2,3],[0,1]]
for edge in edges:
    edge_1 = edge[0]
    edge_2 = edge[1]
    weight = 1.0
    G.add_edge(edge_1, edge_2,weight=weight)

rG = rx.PyGraph(multigraph=G.is_multigraph())
nodes = list(G.nodes)
node_indices = dict(zip(nodes, rG.add_nodes_from(nodes)))
rG.add_edges_from(
        [(node_indices[x[0]], node_indices[x[1]], x[2]['weight']) for x in G.edges(data=True)]
    ) 


general_lattice = Lattice(rG)

from qiskit-nature.