For循环处理多Pandas DataFrame时concat操作未生效问题

尝试用for循环批量处理多个DataFrame（示例为2个，实际数量更多），循环中删除列的操作看似正常，但concat操作未生效，无法更新dfs列表中引用的原始DataFrame。循环内打印能得到预期结果，但外部打印原始DataFrame A时，并未更新为预期的Anew。

简化示例代码：

import numpy as np
import pandas as pd


data = [['Alex',10],['Bob',12],['Clarke',13]]
data2 = ['m','m','x']
A = pd.DataFrame(data, columns=['Name','Age'])
B = pd.DataFrame(data, columns=['Name','Age'])
C = pd.DataFrame(data2, columns=['Gender'])

#expected result for A:
Anew=pd.DataFrame([['Alex','m'],['Bob','m'],['Clarke','x']], columns=['Name', 'Gender'])

dfs = [A,B]

for k, v in enumerate(dfs):
    # The following line works as expected on A an B respectively, inplace is required to actually modify A,B as defined above
    dfs[k]=v.drop('Age',axis=1, inplace=True)
    # The following line doesn't do anything, I was expecting Anew (see above) 
    dfs[k] = pd.concat([v, C], axis=1)
    # The following line prints the expected result within the loop
    print(dfs[k])

# This just shows A, not Anew: To me tha tmeans A was never updated with dfs[k] as I thought it would. 
print(A)

核心问题分析 #

1. inplace=True的副作用：v.drop('Age', axis=1, inplace=True)会直接修改原始DataFrame v（即A和B），但该方法返回值是None。你将None赋值给dfs[k]后，后续pd.concat只是把列表对应位置替换成了新DataFrame，但这个新对象和原始的A、B没有引用关系。
2. Python引用赋值逻辑：列表dfs初始存储的是A、B的引用，但执行dfs[k] = 新对象时，仅修改了列表内的引用指向，并未改变原始的A、B本身。循环内打印的是新生成的对象，外部的A还是最初的那个实例。

解决方案 #

方案1：直接更新原始DataFrame变量 #

通过索引判断直接修改原始变量，适合数量较少的场景：

import numpy as np
import pandas as pd

data = [['Alex',10],['Bob',12],['Clarke',13]]
data2 = ['m','m','x']
A = pd.DataFrame(data, columns=['Name','Age'])
B = pd.DataFrame(data, columns=['Name','Age'])
C = pd.DataFrame(data2, columns=['Gender'])

dfs = [A,B]

for k, v in enumerate(dfs):
    # 先删除列（不用inplace，返回新对象），再拼接C
    updated_df = v.drop('Age', axis=1)
    updated_df = pd.concat([updated_df, C], axis=1)
    # 直接更新原始变量
    if k == 0:
        A = updated_df
    elif k == 1:
        B = updated_df
    dfs[k] = updated_df

print(A)  # 输出预期的Anew

方案2：用字典存储，批量更新更灵活 #

如果DataFrame数量多，用字典映射名称和对象，避免索引判断：

import numpy as np
import pandas as pd

data = [['Alex',10],['Bob',12],['Clarke',13]]
data2 = ['m','m','x']
df_dict = {
    'A': pd.DataFrame(data, columns=['Name','Age']),
    'B': pd.DataFrame(data, columns=['Name','Age'])
}
C = pd.DataFrame(data2, columns=['Gender'])

for name in df_dict:
    df = df_dict[name]
    updated_df = df.drop('Age', axis=1)
    df_dict[name] = pd.concat([updated_df, C], axis=1)

# 通过键名获取更新后的DataFrame
print(df_dict['A'])

方案3：生成新列表，不修改原始对象 #

如果不需要保留原始A、B，直接生成处理后的新列表即可：

import numpy as np
import pandas as pd

data = [['Alex',10],['Bob',12],['Clarke',13]]
data2 = ['m','m','x']
A = pd.DataFrame(data, columns=['Name','Age'])
B = pd.DataFrame(data, columns=['Name','Age'])
C = pd.DataFrame(data2, columns=['Gender'])

dfs = [A,B]
processed_dfs = []

for v in dfs:
    updated_df = v.drop('Age', axis=1)
    updated_df = pd.concat([updated_df, C], axis=1)
    processed_dfs.append(updated_df)

# 处理后的结果存储在processed_dfs中
print(processed_dfs[0])

内容的提问来源于stack exchange，提问作者Cattoaster