pandas如何仅删除文本中超过2位的数字（保留1-2位数字）

使用pandas开展文本数据处理时，需要删除文本中所有位数超过2位的连续数字，同时保留1位、2位长度的数字。
初始测试数据集如下：

customerId                            text
0           1  Hello you should call 46232348
1           2                      What is 42
2           3       Is this a number or 23213
3           4               1 person is there
4           5                    It is 4x4 cm

初始实现代码使用\d+匹配所有长度的连续数字做替换，代码如下：

import pandas as pd
d = {
    "customerId": [1, 2, 3, 4, 5],
    "text": ["Hello you should call 46232348",
             "What is 42",
             "Is this a number or 23213",
             '1 person is there',
             'It is 4x4 cm'],
}
df = pd.DataFrame(data=d)
print(df)
df['text_without_number'] = df['text'].str.replace('\d+', '')

print(df)

代码运行后所有数字都被删除，结果不符合预期：

customerId                            text     text_without_number
0           1  Hello you should call 46232348  Hello you should call 
1           2                      What is 42                What is 
2           3       Is this a number or 23213    Is this a number or 
3           4               1 person is there         person is there
4           5                    It is 4x4 cm              It is x cm

预期的正确处理结果需要保留2位及以下长度的数字：

customerId                            text     text_without_number
0           1  Hello you should call 46232348  Hello you should call 
1           2                      What is 42             What is 42  
2           3       Is this a number or 23213    Is this a number or 
3           4               1 person is there      1 person is there
4           5                    It is 4x4 cm           It is 4x4 cm

原代码的问题在于正则规则\d+会匹配所有长度≥1的连续数字，因此会无差别删除所有数字。要实现仅删除3位及以上数字的需求，只需要将正则匹配规则修改为\d{3,}，匹配所有连续3个及以上的数字串替换为空即可。
修正后的代码如下：

import pandas as pd
d = {
    "customerId": [1, 2, 3, 4, 5],
    "text": ["Hello you should call 46232348",
             "What is 42",
             "Is this a number or 23213",
             '1 person is there',
             'It is 4x4 cm'],
}
df = pd.DataFrame(data=d)
# 仅替换3位及以上的连续数字，显式声明regex=True避免pandas版本差异导致规则失效
df['text_without_number'] = df['text'].str.replace(r'\d{3,}', '', regex=True)
print(df)

运行后即可得到符合预期的结果：

customerId                            text     text_without_number
0           1  Hello you should call 46232348  Hello you should call 
1           2                      What is 42               What is 42
2           3       Is this a number or 23213    Is this a number or 
3           4               1 person is there       1 person is there
4           5                    It is 4x4 cm            It is 4x4 cm

规则说明 #

• \d{3,}是正则表达式的量词写法，代表匹配连续出现至少3次的数字字符，刚好对应「位数超过2位的数字」的匹配需求
• 正则字符串前加r声明为Python原生字符串，避免转义字符干扰正则规则解析
• 显式传入regex=True参数，兼容不同pandas版本的replace默认行为，保证代码跨版本稳定运行

内容的提问来源于stack exchange，提问作者Pyyyyt