如何使用NetworkX绘制MultiDiGraph并解决绘图报错问题

报错核心原因 #

• 变量命名冲突：代码中直接将permutations赋值为排列结果列表，覆盖了从itertools导入的permutations方法，属于代码隐患。
• 多重图边属性格式不匹配：MultiDiGraph的每条边带有唯一key标识，nx.get_edge_attributes返回的边标签字典键为(起点, 终点, 边序号)三元组，旧版本nx.draw_networkx_edge_labels默认仅识别(起点, 终点)二元组键，直接传入会触发类型报错。
• 变量未定义隐患：当传入weight=False时，edge_labels不会被定义，后续调用绘制方法会触发变量未定义报错。
• 双向边重叠问题：默认绘制模式下双向边会完全重叠，无法区分方向和对应权重。

修改后完整可运行代码 #

import pandas as pd
import numpy as np
import networkx as nx
from itertools import combinations, permutations

# 修改变量名，避免覆盖导入的permutations方法
names = ['Mike','Bob','Phil']
perm_list = list(permutations(names, 2))

col1 = [val[0] for val in perm_list]
col2 = [val[1] for val in perm_list]
transactions = [1,3,5,3,9,2]

df = pd.DataFrame({'Sender':col1,'Recipient':col2,'Num_Transactions':transactions})

G = nx.MultiDiGraph()
G.add_nodes_from(names)

E = list(zip(df['Sender'].tolist(),df['Recipient'].tolist(),df['Num_Transactions'].tolist()))
G.add_weighted_edges_from(E)

def draw_graph(G, nodes_position, weight):
    # 添加connectionstyle参数让双向边呈弧形分开，避免重叠
    nx.draw(G, nodes_position, with_labels=True, font_size=15, node_size=400,
            edge_color='gray', arrowsize=10, connectionstyle='arc3, rad=0.1')
    edge_labels = {}
    if weight==True:
        # 处理多重图边标签，将三元组键转换为二元组
        for (u, v, key), w in nx.get_edge_attributes(G,'weight').items():
            # 若存在同起点同终点的多条平行边，可替换为下方注释代码拼接权重
            # edge_labels[(u, v)] = edge_labels.get((u, v), '') + f'/{w}'
            edge_labels[(u, v)] = w
    # 增加label_pos参数，让边标签显示在弧形边中间位置
    nx.draw_networkx_edge_labels(G, nodes_position, edge_labels=edge_labels, label_pos=0.3)

draw_graph(G, nx.circular_layout(G), True)

内容的提问来源于stack exchange，提问作者AI92