Python中替换JSON文件内同名多字段的问题

我尝试修改JSON文件，但仅成功修改了第二个同名字段，第一个字段完全没有变化。

我的代码如下：

with open(os.path.join(model_folder, 'config.json'), 'r') as f:
    config = json.load(f)

config['speakers_file'] = os.path.join('model', 'speaker_ids', 'speakers.pth')
config['embeddings_file'] = [os.path.join('embeddings', 'corpus_speakers.pth'), os.path.join('embeddings', 'train_speakers.pth')]


with open(os.path.join(model_folder, 'config.json'), 'w') as f:
    json.dump(config, f, indent=4)

工具自动生成的原始config内容：

"speakers_file": "/home/username/pathinmymachine/modelrunfolder/speakers.pth",
    "embeddings_file": [
        "/home/username/pathinmymachine/corpusfolder/speakers.pth",
        "speakers.pth"
"speakers_file": "/home/username/pathinmymachine/modelrunfolder/speakers.pth",
"embeddings_file": [
    "/home/username/pathinmymachine/corpusfolder/speakers.pth",
    "speakers.pth"

我期望修改后的结果：

"speakers_file": "model/speaker_ids/speakers.pth",
    "embeddings_file": [
        "embeddings/corpus_speakers.pth",
        "embeddings/train_speakers.pth"
"speakers_file": "model/speaker_ids/speakers.pth",
"embeddings_file": [
    "embeddings/corpus_speakers.pth",
    "embeddings/train_speakers.pth"

实际得到的结果：

"speakers_file": "/home/username/pathinmymachine/modelrunfolder/speakers.pth",
    "embeddings_file": [
        "/home/username/pathinmymachine/corpusfolder/speakers.pth",
        "speakers.pth"
"speakers_file": "model/speaker_ids/speakers.pth",
"embeddings_file": [
    "embeddings/corpus_speakers.pth",
    "embeddings/train_speakers.pth"

核心原因 #

JSON规范不允许对象中存在重复键，Python标准库的json.load()在解析这种不规范JSON时，会自动保留最后一个重复键的值，前面的键值对会被直接覆盖。因此你读取到的config字典里只有最后一组speakers_file和embeddings_file，修改自然只影响最后一个。

解决方法 #

方法一：直接文本替换（适合结构固定的场景） #

跳过JSON解析，直接读取文件内容为字符串，进行精准替换，避免解析时丢失重复键。

import os

file_path = os.path.join(model_folder, 'config.json')

# 读取原始文件内容
with open(file_path, 'r') as f:
    content = f.read()

# 替换speakers_file字段值
content = content.replace(
    '"speakers_file": "/home/username/pathinmymachine/modelrunfolder/speakers.pth"',
    '"speakers_file": "model/speaker_ids/speakers.pth"'
)

# 替换embeddings_file字段值
content = content.replace(
    '"embeddings_file": [\n        "/home/username/pathinmymachine/corpusfolder/speakers.pth",\n        "speakers.pth"',
    '"embeddings_file": [\n        "embeddings/corpus_speakers.pth",\n        "embeddings/train_speakers.pth"'
)

# 写入修改后的内容
with open(file_path, 'w') as f:
    f.write(content)

注意：这种方法要求原始文本的格式（空格、换行、引号）完全匹配，否则替换会失效。

方法二：使用支持重复键的JSON解析库 #

使用第三方库demjson，它可以解析包含重复键的JSON，并保留所有键值对的顺序。

1. 先安装库：

pip install demjson

2. 修改代码：

import os
import demjson

file_path = os.path.join(model_folder, 'config.json')

# 解析JSON，保留重复键和顺序
with open(file_path, 'r') as f:
    config = demjson.decode(f.read(), preserve_order=True)

# 遍历所有键，修改所有匹配的字段
for key in list(config.keys()):
    if key == 'speakers_file':
        config[key] = 'model/speaker_ids/speakers.pth'
    elif key == 'embeddings_file':
        config[key] = ['embeddings/corpus_speakers.pth', 'embeddings/train_speakers.pth']

# 写回文件，保持格式缩进
with open(file_path, 'w') as f:
    demjson.encode_to_file(f, config, indent=4)

内容的提问来源于stack exchange，提问作者badner