好吧，我要介绍的解决方案有点老套，但我愿意接受更多优化建议。在
首先，我们将创建一个用于测试的虚拟图import networkx as nx
G = nx.balanced_tree(2,4,create_using=nx.DiGraph())
接下来，我们将使用networkx的dfs_treeAPI(使用最新版本)，并使用depth_limit属性来提取深度为n和{}的树，其中n+1是用户输入的深度(因为它从1开始索引深度)
^{pr2}$
然后对深度n+1执行相同的操作T2 = nx.dfs_tree(G, source=0,depth_limit=4)
T2_edges =list(T2.edges())
#[(0, 1), (0, 2), (1, 3), (1, 4), (2, 5), (2, 6), (3, 8), (3, 7), (4, 9), (4, 10), (5, 11), (5, 12), (6, 13), (6, 14), (7, 16), (7, 15), (8, 17), (8, 18), (9, 19), (9, 20), (10, 21), (10, 22), (11, 24), (11, 23), (12, 25), (12, 26), (13, 27), (13, 28), (14, 29), (14, 30)]
现在来看看这两个列表的XORedges_left = list(set(T1_edges).symmetric_difference(T2_edges))
#[(14, 30), (11, 23), (10, 21), (7, 16), (11, 24), (7, 15), (10, 22), (9, 20), (12, 25), (13, 28), (8, 17), (14, 29), (12, 26), (13, 27), (8, 18), (9, 19)]
这些是3级的边。现在提取这些级别的节点nodes_at_level = set([x[0] for x in edges_left])
#{7, 8, 9, 10, 11, 12, 13, 14}
然后使用bfs_tree在这些节点提取树for n in nodes_at_level:
tree = nx.bfs_tree(G, n)
print tree.edges() #Do whatever you want with those subgraphs
#[(7, 16), (7, 15)]
#[(8, 17), (8, 18)]
#[(9, 19), (9, 20)]
#[(10, 21), (10, 22)]
#[(11, 24), (11, 23)]
#[(12, 25), (12, 26)]
#[(13, 27), (13, 28)]
#[(14, 29), (14, 30)]
发布者：全栈程序员-站长，转载请注明出处：https://javaforall.net/219454.html原文链接：https://javaforall.net
