使用NetworkX加载CSV构建图时遇类型错误与解包错误求助

操作流程 #

• 使用sknetwork.data.from_csv读取nodes.csv和edges.csv，得到稀疏矩阵格式的nodes和edges
• 转换为列表_nodes = [n for n in nodes]、_edges = [n for n in edges]
• 创建NetworkX的Graph实例，调用add_nodes_from(_nodes)和add_edges_from(_edges)添加节点与边

异常情况 #

• 调用add_nodes_from(_nodes)时，先触发TypeError: unhashable type: 'csr_matrix'，随后抛出ValueError: not enough values to unpack (expected 2, got 1)
• 调用add_edges_from(_edges)时，抛出"Unexpected exception formatting exception. Falling back to standard exception"

相关代码 #

import pandas as pd
import numpy as np
from scipy import sparse
import pandas as pd
import networkx as nx
from sknetwork.data import from_edge_list, from_adjacency_list, from_graphml, from_csv
from sknetwork.visualization import svg_graph, svg_bigraph
from sknetwork.utils import bipartite2undirected

edges = from_csv("edges.csv")
nodes = from_csv("nodes.csv")

_nodes = [n for n in nodes]
_edges = [n for n in edges]

G = nx.Graph() 
G.add_nodes_from(_nodes) # 触发错误
G.add_edges_from(_edges) # 触发错误
print(nx.info(G))

错误回溯（add_nodes_from时） #

TypeError                                 Traceback (most recent call last)
File ~/opt/anaconda3/lib/python3.9/site-packages/networkx/classes/graph.py:562, in Graph.add_nodes_from(self, nodes_for_adding, **attr)
    561 try:
--> 562     if n not in self._node:
    563         self._adj[n] = self.adjlist_inner_dict_factory()

TypeError: unhashable type: 'csr_matrix'

During handling of the above exception, another exception occurred:

ValueError                                Traceback (most recent call last)
Cell In[99], line 10
      7 _labels = [n for n in edges][1:]
      9 G = nx.Graph() 
--> 10 G.add_nodes_from(_nodes) 
     11 G.add_edges_from(_edges) 
     12 print(nx.info(G))

File ~/opt/anaconda3/lib/python3.9/site-packages/networkx/classes/graph.py:569, in Graph.add_nodes_from(self, nodes_for_adding, **attr)
    567         self._node[n].update(attr)
    568 except TypeError:
--> 569     nn, ndict = n
    570     if nn not in self._node:
    571         self._adj[nn] = self.adjlist_inner_dict_factory()

ValueError: not enough values to unpack (expected 2, got 1)

数据样本 #

• _nodes元素为csr_matrix类型：

[<1x5795 sparse matrix of type '<class 'numpy.int64'>' with 385 stored elements in Compressed Sparse Row format>, ...]

• _edges元素为csr_matrix类型：

[<1x5794 sparse matrix of type '<class 'numpy.int64'>' with 248 stored elements in Compressed Sparse Row format>, ...]

• 节点数据样本：

(0, 1265)   1
  (0, 1338) 1
  ...
  (5792, 3) 2
  (5792, 5) 1
  (5792, 29)    1

• 边数据样本：

(0, 682)    1
  (0, 683)  1
  ...
  (5793, 5766)  1
  (5793, 5767)  1
  (5793, 5768)  1
  (5793, 5769)  1

问题根源 #

sknetwork.data.from_csv返回的是稀疏矩阵对象，直接遍历得到的是矩阵的行（仍为csr_matrix类型），而NetworkX的add_nodes_from和add_edges_from需要可哈希的节点标识（如整数、字符串）和边的二元组（如(u, v)），csr_matrix无法满足要求。

修正步骤 #

1. 手动解析CSV数据 #

直接用Pandas读取CSV，提取NetworkX需要的格式：

import pandas as pd
import networkx as nx

# 读取节点CSV，提取节点ID（假设第一列为节点ID）
nodes_df = pd.read_csv("nodes.csv")
node_ids = nodes_df.iloc[:, 0].tolist()

# 读取边CSV，提取边的(u, v)二元组（假设前两列为起点、终点）
edges_df = pd.read_csv("edges.csv")
edge_list = edges_df.iloc[:, :2].values.tolist()

# 构建图
G = nx.Graph()
G.add_nodes_from(node_ids)
G.add_edges_from(edge_list)
print(nx.info(G))

2. 保留节点/边属性（可选） #

如果需要保留属性，可补充以下代码：

# 添加节点属性
for idx, row in nodes_df.iterrows():
    node_id = row.iloc[0]
    attrs = row.iloc[1:].to_dict()
    G.nodes[node_id].update(attrs)

# 添加边属性
for idx, row in edges_df.iterrows():
    u, v = row.iloc[:2]
    attrs = row.iloc[2:].to_dict()
    G.edges[u, v].update(attrs)

3. 利用sknetwork直接转NetworkX图 #

sknetwork提供内置转换方法，无需手动解析邻接关系：

from sknetwork.data import from_csv
from sknetwork.utils import to_networkx
import pandas as pd
import networkx as nx

# 读取边数据生成邻接矩阵，直接转为NetworkX图
adjacency = from_csv("edges.csv")
G = to_networkx(adjacency)

# 读取节点数据并添加属性
nodes_df = pd.read_csv("nodes.csv")
for idx, row in nodes_df.iterrows():
    node_id = row.iloc[0]
    attrs = row.iloc[1:].to_dict()
    G.nodes[node_id].update(attrs)

print(nx.info(G))

内容的提问来源于stack exchange，提问作者Data Science Analytics Manager