Generating a connected graph instead of disjoint method graphs?

[jsdelfino]

Is there a way to generate a single connected graph for a codebase like Joern's export.graphml, instead or in addition to the disjoint graphs that are currently generated by Atom for each method?

Alternatively, if that is not supported by Atom --export... can you recommend an approach to post-process the generated disjoint graphs and reconnect them once they're loaded in networkx for example? I already understand how to connect nodes in networkx, but I'm not really sure which source nodes should be matched with which target nodes in the CPG, and what attributes on these nodes should be used to match them.

Thanks!

[jsdelfino]

I can, of course, load the disjoint GraphML graphs into a single NetworkX graph and manually add edges from the CALL nodes to the METHOD nodes with matching names, but that approach is probably too naive to establish all the correct connections, and it feels a bit hacky...

# Load the generated graphs into a single directed graph
graph = nx.DiGraph()
files = glob.glob("graph/atom/export/*.graphml")
for file in files:
    print(file)
    g = read_graphml(file)
    for node in g.nodes:
        g.nodes[node]["source_file"] = file
    graph.update(g)

print(f"Graph has {graph.number_of_nodes()} nodes and {graph.number_of_edges()} edges.")

# Build an index of methods
methods = {}
for node, attrs in graph.nodes(data=True):
    if attrs.get("labelV") == "METHOD":
        method_name = attrs.get("FULL_NAME")
        methods[method_name] = node

# Print sorted list of method names
print("Loaded methods:")
for method in sorted(list(methods.keys())):
    print(method)
                                                                                                                                                                                              
# Connect call nodes to their target methods                                                                                                                                                  
for node, attrs in graph.nodes(data=True):
    if attrs.get("labelV") == "CALL":
        method_name = attrs.get("METHOD_FULL_NAME")
        target = methods.get(method_name)
        if target is not None:
            graph.add_edge(node, target)

Is there a better way to produce the fully connected graph?