Python系列对象是可变地址解析

我试图在数据帧中创建一个新列，该列将从字符串中解析出地址。尝试执行此操作时，我收到以下错误：

("'Series' objects are mutable, thus they cannot be hashed", u'occurred at index pk')

我在这个网站上看到了其他类似的问题，但不太明白它是如何应用于我的代码的：

import usaddress, re, pyodbc
import pandas as pd

conn = pyodbc.connect("DSN=TEST;UID=test;PWD=test")

sql = "select top 10 pk, address from test..test"
df = pd.read_sql(sql,conn)

pattern = re.compile(".+\\b[0-9]{5}\\b")

def extract(pat):
    print pat
    test = pattern.findall(pat)
    return str(test[0])

i = 0

for i in df.iterrows():
    df[i]['cleansed_address'] = df.apply(lambda x: extract(df[i]['descrsched']))
    i+=1

MCVE 解决X/Y问题

---

df['descrsched'].应用（摘录）

？谢谢！比看上去简单多了。

df = pd.DataFrame([[1, 2,], [3, 4]])
df

# This is a tuple (index value, Series object that represents row)
#   |
#   v    
for i in df.iterrows():
    print(df[i])
#            ^
#            |
# This is you trying to tell Pandas to use a tuple
# in which the second element is a Series as a reference for a column name

df['cleansed_address'] = df['descrsched'].str.findall(pat).str[0]