How to merge two graphs

[gcorso]

I have two heterogeneous graphs (HeteroData) with disjoint node and edge classes that I need to merge into a single graph (to then create edges between the different type of nodes).

Right now I have tried the following code:

complex_graph = copy.deepcopy(receptor_graph)     # contains "receptor" node type
complex_graph['ligand'] = copy.deepcopy(ligand_graph['ligand'])
complex_graph['ligand', 'lig_bond', 'ligand'] = copy.deepcopy(ligand_graph['ligand', 'lig_bond', 'ligand'])

However this does not work because complex_graph['ligand'] results being of class dict instead of class torch_geometric.data.storage.NodeStorage as complex_graph['receptor']. I have also tried using get_node_store without success (same issue).

How can I properly merge the two graphs?
Thank you very much for the help!

[rusty1s]

We currently don't provide any dedicated function for this, so this gets a bit cumbersome. How about

new_graph = copy.copy(receptor_graph)
for type in ligrand_graph.node_types + ligrand_graph.edge_types:
    for key, value in ligrand_graph[type].items():
        new_graph[type][key] = value

[gcorso]

This worked!
Thank you very much for the answer and, as always, for the terrific work Matthias!

[rusty1s]

:) Also added this functionality into PyG directly, see #6313. You should now be able to do:

receptor_graph.update(ligrand_grad)