python - 如何在 Pandas 中进行条件比较

Question

我在 Pandas 中有以下数据框

code     prod_a      prod_b     flag
123      MS          MS         to be checked
123      HS          MS         more than 1 prod
123      MS          HS         to be checked
123      HS          MS         more than 1 prod
123      MS          MS         to be checked

我只想在 flag = to be checked 和其他标志 more than 1 prod 保持原样的情况下比较 prod_a 和 prod_b。我想要的数据框如下

code     prod_a      prod_b     flag               final_flag
123      MS          MS         to be checked      matched
123      HS          MS         more than 1 prod   more than 1 prod   
123      MS          HS         to be checked      not matched
123      HS          MS         more than 1 prod   more than 1 prod
123      MS          MS         to be checked      matched

我怎样才能在 pandas 中做到这一点。

---

重新创建数据框:

import pandas as pd

data = '''\
code,prod_a,prod_b,flag
123,MS,MS,to be checked
123,HS,MS,more than 1 prod
123,MS,HS,to be checked
123,HS,MS,more than 1 prod
123,MS,MS,to be checked
'''

fileobj = pd.compat.StringIO(data)
df = pd.read_csv(fileobj, sep=',')

Answer 1

使用numpy.select通过 & 为按位 AND 和通过 ~ 反转的链条件:

m1 = df['flag'].eq('to be checked')
m2 = df.prod_a.eq(df.prod_b)

df['final_flag'] = np.select([m1 & m2, m1 & ~m2],['matched','not matched'],default=df['flag'])
print (df)
   code prod_a prod_b              flag        final_flag
0   123     MS     MS     to be checked           matched
1   123     HS     MS  more than 1 prod  more than 1 prod
2   123     MS     HS     to be checked       not matched
3   123     HS     MS  more than 1 prod  more than 1 prod
4   123     MS     MS     to be checked           matched

@Anton vBR 的解决方案:

m1 = df['flag'].eq('to be checked')
m2 = df.prod_a.eq(df.prod_b)

df['final_flag'] = df['flag']
df.loc[m1 & m2, 'final_flag'] = 'matched'
df.loc[m1 & ~m2, 'final_flag'] = 'not matched'
print (df)
   code prod_a prod_b              flag        final_flag
0   123     MS     MS     to be checked           matched
1   123     HS     MS  more than 1 prod  more than 1 prod
2   123     MS     HS     to be checked       not matched
3   123     HS     MS  more than 1 prod  more than 1 prod
4   123     MS     MS     to be checked           matched