NetworkX - 创建反向图的深拷贝

Question

我在使用NetworkX创建一个反向有向图的深拷贝时遇到了问题，想要查看这个拷贝的PageRank（网页排名）。看起来问题不在于PageRank本身，而是在计算这个损坏的反向有向图的PageRank时抛出了异常。

首先，我用这段代码生成一个有向图，基本上就是创建一堆节点，随机选几个，然后从这些节点向另外50个左右随机选的节点画出有向边。最后，去掉所有连接数为0的节点。

def global_classroom(nodes):
    network = nx.DiGraph()
    network.add_nodes_from([User(i, "0.0.1") for i in range(nodes)])
    network_nodes = network.nodes()
    teachers = random.sample(network_nodes, nodes / 100)
    for teacher in teachers:
        for i in range(nodes / 20):
            student = network_nodes[random.randrange(nodes)]
            if teacher != student:
            network.add_edge(teacher, student)
    for node in network_nodes:
        if network.degree(node) == 0:
            network.remove_node(node)
    return network

现在，如果我调用这个：

G = global_classroom(1000)
G.reverse(False)
print nx.pagerank(G)

一切都运行得很好。NetworkX可以毫无问题地计算PageRank。

但是如果我调用这个：

G = global_classroom(1000)
G_page_rank = G.reverse()
nx.pagerank(G_page_rank)

就会抛出一个异常。

  Traceback (most recent call last):
  File "a.py", line 62, in <module>
  rank = nx.pagerank(L_page_rank)
  File "<string>", line 2, in pagerank
  File "/Users/PSnyder/Documents/Git/infection/networkx/utils/decorators.py", line 68, in _not_implemented_for
  return f(*args,**kwargs)
  File "/Users/PSnyder/Documents/Git/infection/networkx/algorithms/link_analysis/pagerank_alg.py", line 140, in pagerank
  dangling_nodes = [n for n in W if W.out_degree(n, weight=weight) == 0.0]
  File "/Users/PSnyder/Documents/Git/infection/networkx/classes/digraph.py", line 1023, in out_degree
  return next(self.out_degree_iter(nbunch,weight))[1]
  File "/Users/PSnyder/Documents/Git/infection/networkx/classes/digraph.py", line 941, in out_degree_iter
  for n,nbrs in nodes_nbrs:
  File "/Users/PSnyder/Documents/Git/infection/networkx/classes/digraph.py", line 934, in <genexpr>
  nodes_nbrs=((n,self.succ[n]) for n in self.nbunch_iter(nbunch))
  KeyError: <__main__.User object at 0x104618050>

其中<__main__.User object at 0x104618050>是我有向图中许多节点之一。

我希望能够简单地进行深拷贝，而不是需要再次反转我的图来恢复到原来的样子。

这里是构成我节点的User类的全部内容：

class User(object):
    def __init__(self, name, version, teachers = [], students = []):
        self.name = name
        self.version = version
        G.add_node(self)
        self.add_teachers(teachers)
        self.add_students(students)

    def add_students(self, students):
        for student in students:
            G.add_edge(self, student)

    def add_teachers(self, teachers):
        for teacher in teachers:
            G.add_edge(teacher, self)

Answer 1

也许你的用户对象出了问题？你没有提供这个信息，所以我无法判断这是不是networkx的bug。

这个是可以正常工作的（用整数代替User()）

import networkx as nx
import random

def global_classroom(nodes):
    network = nx.DiGraph()
    network.add_nodes_from([i for i in range(nodes)])
    network_nodes = network.nodes()
    teachers = random.sample(network_nodes, nodes / 100)
    for teacher in teachers:
        for i in range(nodes / 20):
            student = network_nodes[random.randrange(nodes)]
            if teacher != student:
                network.add_edge(teacher, student)
    for node in network_nodes:
        if network.degree(node) == 0:
            network.remove_node(node)
    return network


if __name__=='__main__':
    G = global_classroom(1000)
    G.reverse(False)
    print nx.pagerank(G).items()[0:5] # first 5