如何叠加两个DataFrame并为历史差异值添加前缀？

错误原因 #

你的代码里更新字典的逻辑有误：dict_now.update({dict_now[mother_key][son_key]: '(history) ' + value}) 这行是把当前dict_now[mother_key][son_key]的值（也就是nan）作为外层字典的键添加新条目，而非修改内层字典中son_key对应的值。这就是为什么结果里多出一堆nan: '(history) ...'的无效条目，目标内层字典的空值却没被替换。

方法一：修正字典操作逻辑 #

直接修改内层字典的对应键值，替代错误的外层update操作：

import pandas as pd
from io import StringIO
import math

csvfile_now = StringIO(
"""Name Project_A   Project_B   Project_C   Project_D
Mike    2           8
Jane        7       
Kate    17          
""")

csvfile_history = StringIO(
"""Name Project_A   Project_B   Project_C   Project_D
Mike        7       
Jane        8   2   6
Kate    11      12  1
""")

df_now = pd.read_csv(csvfile_now, sep='\t', engine='python')
df_history = pd.read_csv(csvfile_history, sep='\t', engine='python')

df_now = df_now.set_index('Name')
df_history = df_history.set_index('Name')

dict_now = df_now.to_dict('index')
dict_history = df_history.to_dict('index')

# 遍历并更新字典
for name, projects in dict_now.items():
    for proj, val in projects.items():
        if math.isnan(val) and not math.isnan(dict_history[name][proj]):
            # 直接修改内层字典的目标项目值
            dict_now[name][proj] = f'(history) {dict_history[name][proj]}'

# 转换回DataFrame
result_df = pd.DataFrame.from_dict(dict_now, orient='index').reset_index().rename(columns={'index': 'Name'})
print(result_df)

运行输出：

Name Project_A     Project_B     Project_C     Project_D
0   Mike       2.0  (history) 7.0           NaN           8.0
1   Jane       NaN           7.0  (history) 2.0  (history) 6.0
2   Kate      17.0           NaN  (history) 12.0 (history) 1.0

方法二：Pandas原生向量化操作（推荐） #

无需转字典，直接利用Pandas的批量处理能力，代码更简洁高效：

import pandas as pd
from io import StringIO

csvfile_now = StringIO(
"""Name Project_A   Project_B   Project_C   Project_D
Mike    2           8
Jane        7       
Kate    17          
""")

csvfile_history = StringIO(
"""Name Project_A   Project_B   Project_C   Project_D
Mike        7       
Jane        8   2   6
Kate    11      12  1
""")

df_now = pd.read_csv(csvfile_now, sep='\t', engine='python').set_index('Name')
df_history = pd.read_csv(csvfile_history, sep='\t', engine='python').set_index('Name')

# 定位需要填充的单元格：df_now为空且df_history有值
mask = df_now.isna() & df_history.notna()

# 批量替换目标单元格，其余保留原df_now的值
result_df = df_now.mask(mask, '(history) ' + df_history.astype(str)).reset_index()

print(result_df)

该方法通过mask函数精准定位待替换区域，实现批量处理，完全符合Pandas的最佳实践，输出结果与方法一一致。

内容的提问来源于stack exchange，提问作者Mark K