KeyError in NetworkX

Question

我正在挣扎于使用NetworkX的图形算法，并且我不断地得到以下错误消息。我知道这个问题以前已经回答过了(python and networkX keyerror)，但是似乎有一个不同的问题。我的图形结构如下所示，我创建了七个DiGraphs，但它们都有相同的结构，包含三个节点和三个依赖项：

## Network structure of Graph
for i in range(1):
    G_x = nx.DiGraph()
    G_x.add_node('x1', server = 5)
    G_x.add_node('x2', service = 3)
    G_x.add_node('x3', infrastructure = 2)
    G_x.add_edge('x1','x2', dependency = 0.6)
    G_x.add_edge('x1','x3', dependency = 0.6)
    G_x.add_edge('x2','x3', dependency = 0.3)

    G_a = nx.DiGraph()
    G_a.add_node('a1', server = 8)
    G_a.add_node('a2', service = 4)
    G_a.add_node('a3', infrastructure = 3)
    G_a.add_edge('a1','a2', dependency = 0.6)
    G_a.add_edge('a1','a3', dependency = 0.6)
    G_a.add_edge('a2','a3', dependency = 0.3)

    G_b = nx.DiGraph()
    G_b.add_node('b1', server = 12)
    G_b.add_node('b2', service = 5)
    G_b.add_node('b3', infrastructure = 3)
    G_b.add_edge('b1','b2', dependency = 0.6)
    G_b.add_edge('b1','b3', dependency = 0.6)
    G_b.add_edge('b2','b3', dependency = 0.3)

    G_c = nx.DiGraph()
    G_c.add_node('c1', server = 5)
    G_c.add_node('c2', service = 3)
    G_c.add_node('c3', infrastructure = 2)
    G_c.add_edge('c1','c2', dependency = 0.6)
    G_c.add_edge('c1','c3', dependency = 0.6)
    G_c.add_edge('c2','c3', dependency = 0.3)

    G_d = nx.DiGraph()
    G_d.add_node('d1', server = 9)
    G_d.add_node('d2', service = 4)
    G_d.add_node('d3', infrastructure = 2)
    G_d.add_edge('d1','d2', dependency = 0.6)
    G_d.add_edge('d1','d3', dependency = 0.6)
    G_d.add_edge('d2','d3', dependency = 0.3)

    G_e = nx.DiGraph()
    G_e.add_node('e1', server = 14)
    G_e.add_node('e2', service = 9)
    G_e.add_node('e3', infrastructure = 2)
    G_e.add_edge('e1','e2', dependency = 0.6)
    G_e.add_edge('e1','e3', dependency = 0.6)
    G_e.add_edge('e2','e3', dependency = 0.3)

    G_y = nx.DiGraph()
    G_y.add_node('y1', server = 3)
    G_y.add_node('y2', service = 1.5)
    G_y.add_node('y3', infrastructure = 0.5)
    G_y.add_edge('y1','y2', dependency = 0.6)
    G_y.add_edge('y1','y3', dependency = 0.6)
    G_y.add_edge('y2','y3', dependency = 0.3)

现在，我想遍历我的所有图，对一个随机节点进行示例，并在发生“失败”时更新它的值。此外，如果出现“扩散”，节点的邻居也会受到影响：

# Select random node based on probability == 0.05/5.0%
sum_partial_fail = 0
affected_nodes_list = []
affected_node_values_list = []
for graph in [G_x, G_a, G_b, G_c, G_d, G_e, G_y]:
    random_node = random.sample(graph.nodes, 1)[0]
    print("Random_node: ", random_node)
    percentage_chance_of_failure = random.uniform(0, 1)
    print("Percentage chance of failure: ", percentage_chance_of_failure)
    attr = list(graph.nodes.data()[random_node].keys())[0]
    if percentage_chance_of_failure <= 0.3:
        old_node_value = graph.nodes[random_node][attr]
        print("Old node value: ", old_node_value)
        new_node_value = graph.nodes[random_node][attr] - 1
        print("New node value: ", new_node_value)
        if new_node_value < 0:
            nx.set_node_attributes(graph, {(random_node): {attr: 0}})
        for j in graph.neighbors(random_node):
            percentage_chance_of_spread = random.uniform(0, 1)
            print("Percentage chance of spread: ", percentage_chance_of_spread)
            if percentage_chance_of_spread >= 0.5:
#error          new_value = (1 - (graph[random_node][j]['dependency'] * (new_node_value / old_node_value)))\
                            * (graph.nodes[j][attr])
                nx.set_node_attributes(graph, {(j): {attr: new_value}})
                print("Random node attribute chance >0.5: ", random_node[attr])

除为percentage_chance_of_spread为>= 0.5外，一切正常工作。Python没有返回我的“新值”，但给出了以下错误消息：“

##Returned values
Random_node:  b1
Percentage chance of failure:  0.02705497047686123
Old node value:  12
New node value:  11
Percentage chance of spread:  0.6987620126722746
##Returned error message
Traceback (most recent call last):
  File "C:/Users/XXX", line 139, in <module>
    * graph.nodes[j][attr]
KeyError: 'service'

我真的不知道这里有什么问题，也许有人能帮我解决这个问题？非常感谢！

Answer 1

在您的代码中，并不是每个节点都有属性'service'。但有些人会。

如果random_node具有该属性，那么在某个时候，您将得到attr='service'。然后循环for j in graph.neighbors(random_node):。

在这种情况下，如果是percentage...>0.5，您将尝试获得graph.nodes[j][attr]，但attr是'service'。我不认为有任何理由期望节点j具有该属性，仅仅因为random_node具有该属性。