在具有日期时间索引的行之间执行计算并将它们放入新列中

我有一个数据框，由这段代码生成：

import pandas as pd
import datetime as dt

datarange=pd.date_range('01-05-2018 00:00:00', periods=50, freq="4H")
range_series_1=pd.Series(np.random.randint(-5,3,size=50).astype(float), index=datarange)
range_series_2=pd.Series(np.random.randint(5,9,size=50).astype(float), index=datarange)
frame=pd.DataFrame({'value1':range_series_1, 'value2':range_series_2})
frame.index.name='datetime'

我需要两个额外的列（例如“列 1”和“列 2”），它们承载以下计算：

第 1 列 = value1[每天 @ 4:00] - value1[同一天 @ 12:00] --> 该值应放置在当天的 00:00

第 2 列 = value1[每天 @ 4:00] - value2[同一天 @ 12:00] --> 该值应放置在当天的 00:00

第 1 列和第 2 列中的其余行应为 NaN 或 0。

请问有什么熊猫想法吗？

澄清的附加信息：我可能有一个数据框，它不会以如此严格的方式呈现此处的日期时间，所以pd.shift()可能不是一个全球性的解决方案。

所需输出的示例如下（带有更清晰的注释）:

                     value1  value2  column1   column2
datetime                           
2018-01-05 00:00:00    -3.0     8.0      2.0      -8.0  #2.0=-3.0-(-5.0) and -8.0=-3.0-(5.0)   
2018-01-05 04:00:00    -3.0     8.0      NaN       NaN
2018-01-05 08:00:00     1.0     8.0      NaN       NaN
2018-01-05 12:00:00    -5.0     5.0      NaN       NaN
2018-01-05 16:00:00     0.0     5.0      NaN       NaN
2018-01-05 20:00:00    -4.0     8.0      NaN       NaN
2018-01-06 00:00:00    -5.0     8.0     -1.0     -11.0  #-1.0=-5.0-(-4.0) and -11.0=-5.0-(6.0)
2018-01-06 04:00:00    -5.0     5.0      NaN       NaN
2018-01-06 08:00:00    -2.0     5.0      NaN       NaN
2018-01-06 12:00:00    -4.0     6.0      NaN       NaN
2018-01-06 16:00:00    -1.0     6.0      NaN       NaN
2018-01-06 20:00:00     1.0     7.0      NaN       NaN
2018-01-07 00:00:00    -1.0     5.0      4.0      -4.0  #4.0=2.0-(-2.0) and -4.0=2.0-(6.0)
2018-01-07 04:00:00     2.0     8.0      NaN       NaN
2018-01-07 08:00:00    -1.0     8.0      NaN       NaN
2018-01-07 12:00:00    -2.0     6.0      NaN       NaN
2018-01-07 16:00:00     2.0     6.0      NaN       NaN
2018-01-07 20:00:00    -4.0     6.0      NaN       NaN

我想弄清楚这个答案是否可以帮助我，但我还不太熟悉迭代行（在逐行迭代时更新 pandas 中的数据框 https://stackoverflow.com/questions/23330654/update-a-dataframe-in-pandas-while-iterating-row-by-row?source=post_page-----a7b87c3748ea----------------------).

Use:

from datetime import time

np.random.seed(2019)

datarange=pd.date_range('01-05-2018 00:00:00', periods=50, freq="4H")
range_series_1=pd.Series(np.random.randint(-5,3,size=50).astype(float), index=datarange)
range_series_2=pd.Series(np.random.randint(5,9,size=50).astype(float), index=datarange)
frame=pd.DataFrame({'value1':range_series_1, 'value2':range_series_2})
frame.index.name='datetime'
#print (frame)

想法是比较索引和使用的时间DatetimeIndex.floor http://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DatetimeIndex.floor.html用于删除时间（创建默认值00:00:00) 次Series and DataFrame:

s = frame.loc[frame.index.time == time(4, 0), 'value1']
s.index = s.index.floor('d')
print (s)
datetime
2018-01-05   -3.0
2018-01-06   -5.0
2018-01-07   -5.0
2018-01-08   -5.0
2018-01-09   -1.0
2018-01-10   -4.0
2018-01-11   -2.0
2018-01-12    0.0
2018-01-13    1.0
Name: value1, dtype: float64

df1 = frame.loc[frame.index.time == time(12, 0), ['value1', 'value2']]
df1.index = df1.index.floor('d')
print (df1)
            value1  value2
datetime                  
2018-01-05     0.0     6.0
2018-01-06     2.0     8.0
2018-01-07    -5.0     7.0
2018-01-08     2.0     7.0
2018-01-09    -1.0     5.0
2018-01-10     1.0     7.0
2018-01-11     2.0     7.0
2018-01-12    -2.0     6.0

然后从右侧减去DataFrame.rsub http://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.rsub.html，向新列添加一些前缀并连接到原始列：

frame = frame.join(df1.rsub(s, axis=0).add_prefix('new_'))
print (frame.head(15))
                     value1  value2  new_value1  new_value2
datetime                                                   
2018-01-05 00:00:00    -5.0     6.0        -3.0        -9.0
2018-01-05 04:00:00    -3.0     5.0         NaN         NaN
2018-01-05 08:00:00     2.0     7.0         NaN         NaN
2018-01-05 12:00:00     0.0     6.0         NaN         NaN
2018-01-05 16:00:00    -5.0     7.0         NaN         NaN
2018-01-05 20:00:00     1.0     6.0         NaN         NaN
2018-01-06 00:00:00     1.0     5.0        -7.0       -13.0
2018-01-06 04:00:00    -5.0     8.0         NaN         NaN
2018-01-06 08:00:00     0.0     6.0         NaN         NaN
2018-01-06 12:00:00     2.0     8.0         NaN         NaN
2018-01-06 16:00:00    -1.0     8.0         NaN         NaN
2018-01-06 20:00:00    -3.0     8.0         NaN         NaN
2018-01-07 00:00:00    -5.0     5.0         0.0       -12.0
2018-01-07 04:00:00    -5.0     8.0         NaN         NaN
2018-01-07 08:00:00     2.0     5.0         NaN         NaN