python - 如何计算任何 datetime64 列的第二天分钟差？

Question

Index   DateTimeColumn
5517    2015-04-20 15:27:30
5674    2015-04-20 15:37:30
5675    2015-04-20 15:37:30
5917    2015-04-20 15:49:30
5919    2015-04-20 15:49:30
6619    2015-04-20 16:18:12
6652    2015-04-20 16:36:30
6696    2015-04-20 16:21:42

我想生成一些时间范围，比如从 15:27:30 到 20 分钟。或者找到第二天......有人可以为我推荐这个任务的代码片段吗？

Answer 1

我想你可以使用 pd.offsets像 Day 或 Minute:

print df['DateTimeColumn'] + pd.offsets.Minute(20)
0   2015-04-20 15:47:30
1   2015-04-20 15:57:30
2   2015-04-20 15:57:30
3   2015-04-20 16:09:30
4   2015-04-20 16:09:30
5   2015-04-20 16:38:12
6   2015-04-20 16:56:30
7   2015-04-20 16:41:42
Name: DateTimeColumn, dtype: datetime64[ns]

print df['DateTimeColumn'] + pd.offsets.Day()
0   2015-04-21 15:27:30
1   2015-04-21 15:37:30
2   2015-04-21 15:37:30
3   2015-04-21 15:49:30
4   2015-04-21 15:49:30
5   2015-04-21 16:18:12
6   2015-04-21 16:36:30
7   2015-04-21 16:21:42
Name: DateTimeColumn, dtype: datetime64[ns]

您可以创建新的列:

df['Minutes20'] = df['DateTimeColumn'] + pd.offsets.Minute(20)
df['NextDay'] = df['DateTimeColumn'] + pd.offsets.Day()
print df
   Index      DateTimeColumn           Minutes20             NextDay
0   5517 2015-04-20 15:27:30 2015-04-20 15:47:30 2015-04-21 15:27:30
1   5674 2015-04-20 15:37:30 2015-04-20 15:57:30 2015-04-21 15:37:30
2   5675 2015-04-20 15:37:30 2015-04-20 15:57:30 2015-04-21 15:37:30
3   5917 2015-04-20 15:49:30 2015-04-20 16:09:30 2015-04-21 15:49:30
4   5919 2015-04-20 15:49:30 2015-04-20 16:09:30 2015-04-21 15:49:30
5   6619 2015-04-20 16:18:12 2015-04-20 16:38:12 2015-04-21 16:18:12
6   6652 2015-04-20 16:36:30 2015-04-20 16:56:30 2015-04-21 16:36:30
7   6696 2015-04-20 16:21:42 2015-04-20 16:41:42 2015-04-21 16:21:42

您可以通过 normalize 找到下一个 Day 的差异:

#df['Minutes20'] = df['DateTimeColumn'] + pd.offsets.Minute(20)
df['NextDay'] = df['DateTimeColumn'] + pd.offsets.Day()
df['NextDayNorm'] = df['NextDay'].dt.normalize() 
df['NextDayDiff'] = df['NextDay'].dt.normalize() - df['DateTimeColumn']
print df
   Index      DateTimeColumn             NextDay NextDayNorm  NextDayDiff
0   5517 2015-04-20 15:27:30 2015-04-21 15:27:30  2015-04-21     08:32:30
1   5674 2015-04-20 15:37:30 2015-04-21 15:37:30  2015-04-21     08:22:30
2   5675 2015-04-20 15:37:30 2015-04-21 15:37:30  2015-04-21     08:22:30
3   5917 2015-04-20 15:49:30 2015-04-21 15:49:30  2015-04-21     08:10:30
4   5919 2015-04-20 15:49:30 2015-04-21 15:49:30  2015-04-21     08:10:30
5   6619 2015-04-20 16:18:12 2015-04-21 16:18:12  2015-04-21     07:41:48
6   6652 2015-04-20 16:36:30 2015-04-21 16:36:30  2015-04-21     07:23:30
7   6696 2015-04-20 16:21:42 2015-04-21 16:21:42  2015-04-21     07:38:18

如果要按时间范围选择，首先将列 DateTimeColumn 设置为索引，然后设置 indexer_between_time :

print df
   Index      DateTimeColumn
0   5517 2015-04-20 15:27:30
1   5674 2015-04-20 15:37:30
2   5675 2015-04-20 15:37:30
3   5917 2015-04-20 15:49:30
4   5675 2015-04-21 15:37:30
5   5917 2015-04-22 15:49:30
6   5919 2015-04-23 15:49:30
7   6619 2015-04-20 16:18:12
8   6652 2015-04-20 16:36:30
9   6696 2015-04-20 16:21:42

start = pd.to_datetime('15:27:30').time()
end = pd.to_datetime('15:47:30').time()
print start
15:27:30
print end
15:47:30

df.index = df['DateTimeColumn']
print df.ix[df.index.indexer_between_time(start, end)]
                     Index      DateTimeColumn
DateTimeColumn                                
2015-04-20 15:27:30   5517 2015-04-20 15:27:30
2015-04-20 15:37:30   5674 2015-04-20 15:37:30
2015-04-20 15:37:30   5675 2015-04-20 15:37:30
2015-04-21 15:37:30   5675 2015-04-21 15:37:30

您可以通过偏移创建开始和结束时间:

startdate = pd.to_datetime('15:27:30')
enddate = startdate + pd.offsets.Minute(20)

start = startdate.time()
end = enddate.time()
print start
15:27:30
print end
15:47:30