如何将NetworkX生成的图对象导出为DIMACS文件格式？

Great question! Since NetworkX doesn’t include a built-in DIMACS format writer, you can easily build one yourself by following the DIMACS graph specification. Here’s a step-by-step solution tailored to your use case:

First, let’s recap the core rules of the DIMACS undirected graph format:

• The first line starts with p edge, followed by the total number of nodes and edges (e.g., p edge 5000 12000)
• Each subsequent line represents an edge, starting with e followed by two node IDs (DIMACS standard uses 1-based node numbering)

Step 1: Implement the Custom Export Function #

Here’s a reusable function that converts your NetworkX Graph object to valid DIMACS format:

import networkx as nx

def write_dimacs(graph, output_path):
    # Convert nodes to 1-based numbering (matches DIMACS convention)
    # Skip this line if your original graph already uses 1-based IDs
    standardized_graph = nx.convert_node_labels_to_integers(graph, first_label=1)
    
    node_count = standardized_graph.number_of_nodes()
    edge_count = standardized_graph.number_of_edges()
    
    with open(output_path, 'w') as dimacs_file:
        # Write the problem metadata line
        dimacs_file.write(f"p edge {node_count} {edge_count}\n")
        
        # Write each edge one line at a time
        for u, v in standardized_graph.edges():
            # For undirected graphs, we only write each edge once (no duplicates needed)
            dimacs_file.write(f"e {u} {v}\n")

# Use it with your Amazon graph
g = nx.read_edgelist('com-amazon.ungraph.txt', create_using=nx.Graph(), nodetype=int)
write_dimacs(g, 'com-amazon.dimacs')

Step 2: Key Details to Note #

• Node Numbering: The nx.convert_node_labels_to_integers call ensures your nodes start at 1, which is required for most DIMACS-compliant tools. If your original dataset already uses 1-based IDs, you can remove this line to keep your original node numbers intact.
• Efficiency: For large graphs like the Amazon dataset, this function writes directly to the file line-by-line, so it won’t overload your memory with all edges stored at once.
• Undirected Edges: Since you’re using nx.Graph(), we only write each edge once (DIMACS doesn’t require duplicate entries for undirected connections).

Quick Verification #

To confirm the output is correct, you can check the first few lines of the generated file with a command like:

head com-amazon.dimacs

You should see output similar to this:

p edge 334863 925872
e 1 2
e 1 3
e 1 4
...

内容的提问来源于stack exchange，提问作者Derrick Peh