之前看过networkx,igraph都说支持GML格式文件输入形式，直到今天想再Nepidemix load进文件进行分析，发现Nepidemix　对load输入文件要求是GML和gpickle，其余的不支持。所以查阅了下GML究竟是什么。

GML(Graph Modelling Language):
There are many different programs that work with graphs but almost all of them use their own file format. As a consequence, exchanging graphs between different programs is almost impossible. Simple tasks like exchange of data, externally reproducible results or a common benchmark suite are much harder than neccessary.
Therefore, we have developed a new file format for the Graphlet system: GML. GML supports attaching arbitrary information to graphs, nodes and edges, and is therefore able to emulate almost every other format.
原文链接 可以看出GML 是网络数据的一个统一标准，相当于度量衡的作用。GML支持关于图，点，边的任意数据的附加，因此可以仿真任意格式的数据。
networkx 提供了函数 可以读GML格式的数据 或者将你load进去的普通网络转换成GML格式的文件。链接请click

read_gml(path,[,encoding, relabel]) 读GML
write_gml(G,path)写网络G进GML文件
parse_gml(lines[,relael]) 从字符串中解析GML图
generate_gml(G)  Generate a single entry of the graph G in GML format

举个栗子：
我的原始数据是这样，存在一个GML.txt里
a 1
a 2
b 1
c 2
d 3
b 3
这样构成了一个二分网络，我想提取出第二列网络，即投影。然后将第二列数据的投影网络，用GML数据保
代码：

G = nx.Graph()   
f = open(\'GML.txt\')
i = 0
for line in f:
    i = i+1
    print i
    cell = line.split() 
    G.add_edge(cell[0],cell[1])

f.close()

NSet = nx.bipartite.sets(G) 

User = nx.project(G,NSet[1])
nx.write_gml(User,\'Project_Gml\')

结果：投影的GML文件是这样：

graph [
  node [
    id 0
    label \"1\"
  ]
  node [
    id 1
    label \"3\"
  ]
  node [
    id 2
    label \"2\"
  ]
  edge [
    source 0
    target 1
  ]
  edge [
    source 0
    target 2
  ]
]