Pandas DataFrame条件生成新变量md遇语法错误，求正确实现方案

样本数据 #

import pandas as pd

df = pd.DataFrame({'id':[1,2,3,4,5],
                   'cfs':['REFER', 'DECLINE','PASS','REFER','REFER'],
                   'ofs':['DECLINE','DECLINE', 'PASS','REFER','PASS'],
                   'ffs':['PASS','DECLINE','PASS','REFER','PASS']})

问题描述 #

尝试通过条件语句生成新列md时出现syntaxerror: invalid syntax，错误代码如下：

# If else condition approach
if all df['cfs'] == 'DECLINE' | df['ofs'] == 'DECLINE' | df['ffs'] == 'DECLINE':
       df['md'] =='DECLINE'
else:
    if df['cfs'] == 'PASS' & df['ofs'] == 'PASS' & df['ffs'] == 'PASS':
           df['md'] =='PASS'
else:
    if any df['cfs'] == 'REFER'
        if df['ofs'] == 'REFER'
            if df['ffs'] == 'REFER':
                df['md'] =='REFER'
                
else: df['md'] =='REFER'

需满足的条件：

• 若cfs、ofs或ffs为DECLINE，则md='DECLINE'
• 若cfs、ofs和ffs均为PASS，则md='PASS'
• 若cfs、ofs和ffs均为REFER，则md='REFER'
• 若cfs、ofs和ffs为PASS与REFER的组合，则md='REFER'

错误原因 #

1. 直接用Python原生if/else处理Pandas Series：df['cfs'] == 'DECLINE'返回布尔Series，而非单个布尔值，无法用于原生if判断
2. 逻辑运算符优先级错误：|/&优先级高于比较运算符，需给比较条件加括号
3. 赋值错误：使用==（比较）而非=（赋值）
4. all/any用法错误：未针对每行数据应用，而是直接处理整列

正确实现方法 #

方法一：使用numpy.select（推荐，向量化操作效率高） #

import numpy as np

# 定义条件列表与对应结果
conditions = [
    # 条件1：任意一列是DECLINE
    (df['cfs'] == 'DECLINE') | (df['ofs'] == 'DECLINE') | (df['ffs'] == 'DECLINE'),
    # 条件2：所有列都是PASS
    (df['cfs'] == 'PASS') & (df['ofs'] == 'PASS') & (df['ffs'] == 'PASS'),
    # 条件3：仅包含PASS和REFER的情况
    df[['cfs', 'ofs', 'ffs']].isin(['PASS', 'REFER']).all(axis=1)
]

values = ['DECLINE', 'PASS', 'REFER']

# 生成md列
df['md'] = np.select(conditions, values, default='REFER')

方法二：使用apply逐行处理（适合小数据集） #

def get_md_value(row):
    # 优先检查DECLINE
    if row['cfs'] == 'DECLINE' or row['ofs'] == 'DECLINE' or row['ffs'] == 'DECLINE':
        return 'DECLINE'
    # 检查全PASS
    elif row['cfs'] == 'PASS' and row['ofs'] == 'PASS' and row['ffs'] == 'PASS':
        return 'PASS'
    # 剩余情况均返回REFER
    else:
        return 'REFER'

df['md'] = df.apply(get_md_value, axis=1)

最终结果 #

处理后的数据框如下：

内容的提问来源于stack exchange，提问作者Ricky