Pandas使用update合并DataFrame返回None，致to_csv报错的解决方法

问题描述 #

合并足球赛事数据的archive和df_new两个DataFrame时，执行df_merge = df_merge.update(df_results)后，df_merge变为NoneType，调用to_csv时触发如下错误：

AttributeError: 'NoneType' object has no attribute 'to_csv'

核心原因 #

Pandas的DataFrame.update()方法是原地修改当前DataFrame，不会返回新的DataFrame对象，它的返回值是None。把这个返回值赋值给df_merge，会直接覆盖原有的DataFrame对象，导致后续调用to_csv时出错。

解决方案 #

1. 修正update的调用方式 #

删除赋值操作，直接原地修改：

# 错误写法：df_merge = df_merge.update(df_results)
df_merge.update(df_results)  # 正确写法，原地修改，不返回新对象

2. 优化合并逻辑（可选） #

原合并代码存在冗余，可简化为基于唯一标识的合并逻辑，确保数据匹配更准确：

# 替换原合并代码
df_merge = df_csv.merge(dt, on=['match_date', 'competition', 'team_home', 'team_away', 'match'], how='left')
# 合并后处理列名冲突，优先保留新数据填充原空值
for col in ['home_goals', 'away_goals', 'score']:
    df_merge[col] = df_merge[f'{col}_y'].fillna(df_merge[f'{col}_x'])
# 保留原csv的列顺序
df_merge = df_merge[df_csv.columns]

3. 修复数据类型潜在问题 #

get_result中直接转换int(x['home_goals'])可能因NaN报错，先处理空值：

def get_result(df):
    df = df[(df['score'].notnull()) & (df['result'].isnull())].copy()  # 加copy避免SettingWithCopyWarning
    # 先填充空值为0，再转int
    df['home_goals'] = df['home_goals'].fillna(0).astype(int)
    df['away_goals'] = df['away_goals'].fillna(0).astype(int)
    df['result'] = df.apply(lambda x: market_result(x['team_home'], x['team_away'], x['tip'], x['home_goals'], x['away_goals']), axis=1)
    return df

修改后的完整代码 #

import pandas as pd

def market_result(home,away,mkt,hg,ag):
    if (mkt == f'{home} To Win') and (hg > ag):
        return 'GREEN'
    if (mkt == f'{home} To Win') and (hg <= ag):
        return 'RED'
    if (mkt == f'{away} To Win') and (ag > hg):
        return 'GREEN'
    if (mkt == f'{away} To Win') and (ag <= hg):
        return 'RED'
    if (mkt == 'Both Teams To Score') and (hg > 0) and (ag > 0):
        return 'GREEN'
    if (mkt == 'Both Teams To Score') and ((hg == 0) or (ag == 0)):
        return 'RED'
    if (mkt == 'Both Teams To Score - No') and ((hg == 0) or (ag == 0)):
        return 'GREEN'
    if (mkt == 'Both Teams To Score - No') and ((hg > 0) or (ag > 0)):
        return 'RED'
    if (mkt == 'Under 2.5 Goals') and (hg+ag < 2.5):
        return 'GREEN'
    if (mkt == 'Under 2.5 Goals') and (hg+ag >= 2.5):
        return 'RED'
    if (mkt == 'Over 2.5 Goals') and (hg+ag > 2.5):
        return 'GREEN'
    if (mkt == 'Over 2.5 Goals') and (hg+ag <= 2.5):
        return 'RED'
    if (mkt == 'Under 3.5 Goals') and (hg+ag < 3.5):
        return 'GREEN'
    if (mkt == 'Under 3.5 Goals') and (hg+ag >= 3.5):
        return 'RED'
    if (mkt == 'Over 3.5 Goals') and (hg+ag > 3.5):
        return 'GREEN'
    if (mkt == 'Over 3.5 Goals') and (hg+ag <= 3.5):
        return 'RED'

def get_result(df):
    df = df[(df['score'].notnull()) & (df['result'].isnull())].copy()
    df['home_goals'] = df['home_goals'].fillna(0).astype(int)
    df['away_goals'] = df['away_goals'].fillna(0).astype(int)
    df['result'] = df.apply(lambda x: market_result(x['team_home'], x['team_away'], x['tip'], x['home_goals'], x['away_goals']), axis=1)
    return df

def append_matches(archive,df_new):
    df_csv = pd.read_csv(archive)
    # 处理空字符串为NaN
    df_csv = df_csv.replace('', pd.NA)
    dt = pd.read_csv(df_new)
    
    # 简化合并逻辑
    df_merge = df_csv.merge(dt, on=['match_date', 'competition', 'team_home', 'team_away', 'match'], how='left')
    # 合并后填充数据
    for col in ['home_goals', 'away_goals', 'score']:
        df_merge[col] = df_merge[f'{col}_y'].fillna(df_merge[f'{col}_x'])
    # 保留原csv的列顺序
    df_merge = df_merge[df_csv.columns]
    
    df_results = get_result(df_merge)
    # 直接原地更新，不赋值
    df_merge.update(df_results)
    df_merge.to_csv(archive, index=False)

def main():
    append_matches('archive.csv','df_new.csv')

if __name__ == '__main__':
    main()

内容的提问来源于stack exchange，提问作者Digital Farmer