pandas字符串列多格式日期提取正则表达式调试求助

架构老林

2026-10-2

正则问题诊断

不必要的捕获组过多：原正则共写了7个捕获组，pandas.Series.str.extract会按顺序返回所有捕获组的匹配结果，你没有对非必要分组加?:标记为非捕获组，导致最终拿到的不是完整的日期匹配结果，这也是Jan 27, 1983这类格式只能拿到部分内容的核心原因
开头数字匹配逻辑错误：\d{0,2}\d等价于\d{1,3}，没有明确区分「月/日前置的数字格式」和「月份英文前置的格式」，匹配优先级混乱，容易出现部分匹配就停止的情况
分隔符匹配逻辑冗余：[, -./]{0,2}和[, -./]{1,2}的范围没有和前后的日期部分绑定，遇到带逗号的月日年格式时，分隔符匹配后没有正确关联后面的年份
独立年份的匹配和前面的完整日期匹配是或关系，优先级处理不当容易导致长日期被截断匹配

优化后的正则及用法

将不同格式的日期拆分为独立匹配分支，按「长格式优先、短格式在后」排序，仅保留1个捕获组用于返回完整日期，兼容你列出的所有日期格式：

import pandas as pd
import re

# 样本数据定义
sample_list = ['.Got back to U.S. Jan 27, 1983.\n',
       '.On 21 Oct 1983 patient was discharged from Scroder Hospital after EIGHT DAY ADMISSION\n',
       '4-13-89 Communication with referring physician?: Not Done\n',
       '7intake for follow up treatment at Anson General Hospital on 10 Feb 1983 @ 12 AM\n',
       '.  Pt diagnosed in Apr 1976 after he presented with 2 month history of headaches and gait instability. MRI demonstrated 4 cm L cereballar mass in the paravermian region. He was admitted to PRM and underwent resection complicated by post-op delirium. Post-op sequelas include left palatal myoclonus and ataxia on the left upper and lower extremities which has progressively improved. Pt has not had any evidence of tumor recurrence.\n',
       '1-14-81 Communication with referring physician?: Done\n',
       '. Went to Emerson, in Newfane Alaska. Started in 2002 at CNM. Generally likes job, does not have time to do what she needs to do. Feels she is working more than should be.\n',
       '09/14/2000 CPT Code: 90792: With medical services\n',
       '. Sep 2015- Transferred to Memorial Hospital from above.  Discharged to MH Partial Hospital on Zoloft, Trazadone and Neurontin but unclear if she followed up.\n',
       'Born and raised in Fowlerville, IN.  Parents divorced when she was young, states that it was a "bad" divorce.  Received her college degree from Allegheny College in 2003.  Past verbal, emotional, physical, sexual abuse: No\n']
sample_series = pd.Series(sample_list)

# 优化后正则，用注释区分不同匹配分支便于维护
my_pattern = r"""
(
(?:\d{1,3}[/-]\d{1,2}[/-]\d{2,4})        # 兼容数字格式：月/日/年，包括011/14/83这类前导零多写的格式
|(?:\d{1,2}[/-]\d{2,4})                   # 兼容数字格式：月/年，如6/2008
|(?:(?:Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec)[a-z]*[., -]+\d{1,2}[dhnst]*[., -]+\d{2,4}) # 月英文前置+日+年，如Jan 27, 1983、Mar 20th, 2009
|(?:\d{1,2}[dhnst]*[., -]+(?:Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec)[a-z]*[., -]+\d{2,4}) # 日前置+月英文+年，如20 Mar 2009
|(?:(?:Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec)[a-z]*[., -]+\d{2,4}) # 月英文+年，如Apr 1976、Sep 2015
|(?:\d{4})                                # 单独年份，如2002、2010
)
"""
# 提取日期，flags参数启用正则注释识别和大小写忽略
result = sample_series.str.extract(my_pattern, expand=False, flags=re.VERBOSE | re.IGNORECASE)
print(result)

正则调试优化方法

复杂场景优先按格式拆分为独立分支，不要揉成一坨模糊匹配，分支优先级按「长格式在前、短格式在后」排序，避免长日期被短规则截断
所有不需要提取的分组统一加?:标记为非捕获组，确保str.extract仅返回你需要的完整日期内容
单条样本调试可以用re.search逐行测试，打印匹配对象的group()和groups()结果，快速定位哪部分规则没有命中
对模糊匹配的分隔符、后缀（比如序数词的st/nd/rd/th），明确匹配范围，不要用太宽泛的元字符避免误匹配

内容的提问来源于stack exchange，提问作者Aung Seth Pai