相同代码在.ipynb正常运行但在.py文件中报错的问题求助

问题本质 #

Windows命令行环境下，Python默认采用cp1252编码打开文件，但Geofabrik下载的.osm文件是UTF-8编码，导致ElementTree解析时出现解码失败。而Jupyter Notebook会自动以UTF-8编码处理文件，因此无问题。

快速解决方案：设置Python默认编码为UTF-8 #

在运行脚本前，在命令行设置环境变量：

set PYTHONUTF8=1
python your_script.py

或者在Python脚本开头添加以下代码，全局设置默认编码：

import os
os.environ['PYTHONUTF8'] = '1'

# 后续正常调用OSMnx
import osmnx as ox

network = 'network_file.osm'
G = ox.graph_from_xml(network)

替代方案：手动处理文件编码（无需全局设置） #

如果不想修改全局编码，可以重写OSMnx内部的文件解析逻辑，强制用UTF-8读取文件：

import xml.etree.ElementTree as ET
import osmnx as ox
from osmnx import settings, osm_xml

def _custom_parse_osm_xml(filepath, node_attrs=None, way_attrs=None, rel_attrs=None,
                   node_tags=None, way_tags=None, rel_tags=None):
    node_attrs = node_attrs or settings.osm_xml_node_attrs
    way_attrs = way_attrs or settings.osm_xml_way_attrs
    rel_attrs = rel_attrs or settings.osm_xml_rel_attrs
    node_tags = node_tags or settings.osm_xml_node_tags
    way_tags = way_tags or settings.osm_xml_way_tags
    rel_tags = rel_tags or settings.osm_xml_rel_tags

    nodes = {}
    ways = {}
    relations = {}

    # 强制用UTF-8编码打开文件
    with open(filepath, 'r', encoding='utf-8') as f:
        for event, elem in ET.iterparse(f):
            if elem.tag == 'node':
                node = {'id': elem.attrib['id']}
                for attr in node_attrs:
                    if attr in elem.attrib:
                        node[attr] = elem.attrib[attr]
                node['tags'] = {tag.attrib['k']: tag.attrib['v'] for tag in elem.findall('tag') if tag.attrib['k'] in node_tags}
                nodes[node['id']] = node

            elif elem.tag == 'way':
                way = {'id': elem.attrib['id']}
                for attr in way_attrs:
                    if attr in elem.attrib:
                        way[attr] = elem.attrib[attr]
                way['nodes'] = [nd.attrib['ref'] for nd in elem.findall('nd')]
                way['tags'] = {tag.attrib['k']: tag.attrib['v'] for tag in elem.findall('tag') if tag.attrib['k'] in way_tags}
                ways[way['id']] = way

            elif elem.tag == 'relation':
                relation = {'id': elem.attrib['id']}
                for attr in rel_attrs:
                    if attr in elem.attrib:
                        relation[attr] = elem.attrib[attr]
                relation['members'] = [{k: member.attrib[k] for k in ['type', 'ref', 'role']} for member in elem.findall('member')]
                relation['tags'] = {tag.attrib['k']: tag.attrib['v'] for tag in elem.findall('tag') if tag.attrib['k'] in rel_tags}
                relations[relation['id']] = relation

            elem.clear()

    return nodes, ways, relations

def custom_overpass_json_from_file(filepath):
    # 强制用UTF-8编码解析XML根节点
    with open(filepath, 'r', encoding='utf-8') as f:
        root = ET.parse(f).getroot()
    root_attrs = root.attrib
    nodes, ways, relations = _custom_parse_osm_xml(filepath)
    response_json = osm_xml._create_overpass_json(nodes, ways, relations, root_attrs)
    return response_json

# 替换OSMnx原有的解析函数
osm_xml._overpass_json_from_file = custom_overpass_json_from_file

# 正常加载.osm文件
network = 'network_file.osm'
G = ox.graph_from_xml(network)

为什么之前的ChatGPT方案失败？ #

ox.graph_from_xml只接受文件路径字符串或字节流路径，不接受已经打开的TextIOWrapper对象，因此直接传入文件句柄会触发TypeError。

内容的提问来源于stack exchange，提问作者David Micallef