使用 @numpy.vectorize 在 pandas 中减去 2 个 datetime64 列

Question

我有一个带有 2 个日期列的 pandas DataFrame：已创建和已解决。

Created Resolved 0 2019-09-24 20:48:25 2019-10-31 22:07:36 1 2019-09-27 00:54:39 2019-11-18 17:24:13 2 2019-09-27 20:07:50 NaT 3 2019-09-27 20:17:10 2019-10-22 17:34:08 4 2019-09-27 22:01:29 2019-10-22 17:34:08 5 2019-09-30 17:41:02 NaT 6 2019-10-02 04:36:32 NaT 7 2019-10-03 17:42:15 2019-10-22 17:34:09 8 2019-10-03 18:34:29 NaT 10 2019-10-08 18:40:45 2019-10-22 17:34:09 这是信息。

class 'pandas.core.frame.DataFrame'> Int64Index: 10 entries, 0 to 10 Data columns (total 2 columns): Created 10 non-null datetime64[ns] Resolved 6 non-null datetime64[ns] dtypes: datetime64[ns](2)

我想查看解决需要多少天：

@numpy.vectorize def age(res, cr): if pd.isnull(res): return pd.to_datetime('today')-cr return res-cr

当我将列传递给函数df.diff=age(df.Resolved,df.Created) 时，会出现错误：

ValueError: Cannot add integral value to Timestamp without freq.

Answer 1

我相信您需要用Series.fillna 替换Resolved 列中的缺失值，然后减去Created 列，解决方案已经矢量化：

df['diff'] = df.Resolved.fillna(pd.to_datetime('today')) - df.Created
print (df)
               Created            Resolved                     diff
0  2019-09-24 20:48:25 2019-10-31 22:07:36         37 days 01:19:11
1  2019-09-27 00:54:39 2019-11-18 17:24:13         52 days 16:29:34
2  2019-09-27 20:07:50                 NaT 159 days 13:42:59.435281
3  2019-09-27 20:17:10 2019-10-22 17:34:08         24 days 21:16:58
4  2019-09-27 22:01:29 2019-10-22 17:34:08         24 days 19:32:39
5  2019-09-30 17:41:02                 NaT 156 days 16:09:47.435281
6  2019-10-02 04:36:32                 NaT 155 days 05:14:17.435281
7  2019-10-03 17:42:15 2019-10-22 17:34:09         18 days 23:51:54
8  2019-10-03 18:34:29                 NaT 153 days 15:16:20.435281
10 2019-10-08 18:40:45 2019-10-22 17:34:09         13 days 22:53:24