如何使用networkx寻找加权图中的最短路径？

Question

我正在使用Python 2.7 Enthought distribution中的networkx包计算海港网络之间的最短路径。使用dijkstra_path_length计算距离是可行的，但是我还需要知道它使用dijkstra_path找到了什么路径（另外，我认为如果我先计算路径，然后计算路径的长度，而不是在同一数据上运行Dijkstra的算法两次，运行起来应该更快）。然而，path函数失败了，说list indices must be integers, not str。

这是产生错误的代码。有人能告诉我我做错了什么吗？

import networkx as nx

# Create graph
network_graph = nx.Graph()
f_routes = open('routes-list.txt', 'rb')
# Assign list items to variables
for line in f_routes:
    route_list = line.split(",")
    orig = route_list[0]
    dest = route_list[1]
    distance = float(route_list[2])
    # Add route as an edge to the graph
    network_graph.add_edge(orig, dest, distance=(distance))

# Loop through all destination and origin pairs
for destination in network_graph:
    for origin in network_graph:
        # This line works
        length = nx.dijkstra_path_length(network_graph, origin, destination, "distance")
        # This line fails
        path = nx.dijkstra_path(network_graph, origin, destination, "distance")

我在追踪中得到以下信息。

Traceback (most recent call last):
  File "C:\Users\jamie.bull\workspace\Shipping\src\shortest_path.py", line 67, in <module>
    path = nx.dijkstra_path(network_graph, origin, destination, "distance")
  File "C:\Enthought\Python27\lib\site-packages\networkx\algorithms\shortest_paths\weighted.py", line 74, in dijkstra_path
    return path[target]
TypeError: list indices must be integers, not str

Answer 1

经过一番试验，似乎nx.dijkstra_path在源节点和目标节点相同时引发了一个误导性异常：

>>> import networkx as nx
>>> g = nx.Graph()
>>> g.add_edge('a', 'b', distance=0.3)
>>> g.add_edge('a', 'c', distance=0.7)
>>> nx.dijkstra_path_length(g, 'b', 'c', 'distance')
1.0
>>> nx.dijkstra_path(g, 'b', 'c', 'distance')
['b', 'a', 'c']
>>> nx.dijkstra_path_length(g, 'b', 'b', 'distance')
0
>>> nx.dijkstra_path(g, 'b', 'b', 'distance')
Traceback (most recent call last):
  File "<pyshell#7>", line 1, in <module>
    nx.dijkstra_path(g, 'b', 'b', 'distance')
  File "C:\Users\barberm\AppData\Roaming\Python\Python27\site-packages\networkx\algorithms\shortest_paths\weighted.py", line 74, in dijkstra_path
    return path[target]
TypeError: list indices must be integers, not str

所以只需显式地测试destination和origin是否相同，并在它们相同时分别处理。