使用graph_from_gdfs重建图后调用OSMnx simplify_graph抛TypeError

问题背景 #

工作流如下：
ox.graph_from_polygon => graph_to_gdfs => 修改gdf_edges => ox.graph_from_gdfs => ox.simplify_graph

执行到ox.simplify_graph时抛出TypeError: unhashable type: 'list'，错误位置在Osmnx的simplification.py第362行：

File .../osmnx/simplification.py:362, in simplify_graph(G, strict, edge_attrs_differ, endpoint_attrs, remove_rings, track_merged)
    359 if attr in attrs_to_sum:
    360     # if this attribute must be summed, sum it now
    361     path_attributes[attr] = sum(path_attributes[attr])
--> 362 elif len(set(path_attributes[attr])) == 1:
    363     # if there's only 1 unique value in this attribute list,

即使未修改数据框也会触发该错误，最小复现示例：

import shapely
import osmnx as ox

poly = shapely.from_wkt("POLYGON((9.493474960327147 51.228210969202365,"
                        "9.493346214294432 51.223359875812804,"
                        "9.498442411422728 51.22356145497176,"
                        "9.498571157455443 51.228036284963565,"
                        "9.493474960327147 51.228210969202365))")

G = ox.graph_from_polygon(poly, network_type="drive")

nodes_gdf, edges_gdf = ox.graph_to_gdfs(G.to_undirected())

G2 = ox.graph_from_gdfs(gdf_nodes=nodes_gdf, gdf_edges=edges_gdf)

nodes_gdf_2, edges_gdf_2 = ox.graph_to_gdfs(ox.simplify_graph(G2))

错误原因 #

问题源于无向图转换后的边属性类型异常：调用G.to_undirected()后，部分原本为单一值的边属性（如oneway）会被转为列表类型。而ox.simplify_graph在处理属性时，会尝试将属性值存入集合（set(path_attributes[attr])），但列表是不可哈希类型，直接存入集合就会触发该TypeError。

解决方法 #

方案1：放弃无向图转换（如果业务允许） #

直接去掉G.to_undirected()，保持原图的有向性，重建后即可正常简化：

import shapely
import osmnx as ox

poly = shapely.from_wkt("POLYGON((9.493474960327147 51.228210969202365,"
                        "9.493346214294432 51.223359875812804,"
                        "9.498442411422728 51.22356145497176,"
                        "9.498571157455443 51.228036284963565,"
                        "9.493474960327147 51.228210969202365))")

G = ox.graph_from_polygon(poly, network_type="drive")

nodes_gdf, edges_gdf = ox.graph_to_gdfs(G)  # 移除to_undirected()

G2 = ox.graph_from_gdfs(gdf_nodes=nodes_gdf, gdf_edges=edges_gdf)

nodes_gdf_2, edges_gdf_2 = ox.graph_to_gdfs(ox.simplify_graph(G2))

方案2：处理无向图的列表类型属性（如果必须使用无向图） #

在重建图之前，将边属性中的列表类型转换为可哈希的单一值。以oneway为例，取列表第一个元素（或根据业务逻辑合并）：

import shapely
import osmnx as ox

poly = shapely.from_wkt("POLYGON((9.493474960327147 51.228210969202365,"
                        "9.493346214294432 51.223359875812804,"
                        "9.498442411422728 51.22356145497176,"
                        "9.498571157455443 51.228036284963565,"
                        "9.493474960327147 51.228210969202365))")

G = ox.graph_from_polygon(poly, network_type="drive")

nodes_gdf, edges_gdf = ox.graph_to_gdfs(G.to_undirected())

# 处理列表类型的边属性
edges_gdf['oneway'] = edges_gdf['oneway'].apply(lambda x: x[0] if isinstance(x, list) else x)
# 其他可能转为列表的属性（如lanes、access等）需同理处理
# edges_gdf['lanes'] = edges_gdf['lanes'].apply(lambda x: x[0] if isinstance(x, list) else x)

G2 = ox.graph_from_gdfs(gdf_nodes=nodes_gdf, gdf_edges=edges_gdf)

nodes_gdf_2, edges_gdf_2 = ox.graph_to_gdfs(ox.simplify_graph(G2))

方案3：无向图转换时禁用属性列表化 #

调用to_undirected()时指定reciprocal=False，避免属性被转为列表，但这会改变无向化的逻辑（不会添加反向边），需根据需求选择：

nodes_gdf, edges_gdf = ox.graph_to_gdfs(G.to_undirected(reciprocal=False))

内容的提问来源于stack exchange，提问作者mcsoini