Question

我有一个数据帧df，其中包括一个标识符列和4个日期列。请参见下面的数据框标题。

   Identifier                    A \
0       12253  1989-11-09 00:00:00   
1       11413  1990-09-03 00:00:00   
2       12337  1977-09-07 00:00:00   
3       10610  1994-08-24 00:00:00   
4        7493  1993-08-22 00:00:00   

                      B                              C  \
0   2012-03-01 00:00:00            2015-04-01 00:00:00   
1   2009-08-16 00:00:00            2015-05-18 00:00:00   
2   1977-09-07 00:00:00            1977-09-07 00:00:00   
3   2009-01-09 00:00:00            2015-03-01 00:00:00   
4   2002-06-03 00:00:00            2015-02-16 00:00:00   

                      D  \
0   2012-01-03 00:00:00   
1   2015-05-18 00:00:00   
2   1977-09-07 00:00:00   
3   2015-03-01 00:00:00   
4   2015-02-16 00:00:00

我正在尝试将日期列彼此进行比较，如果A> = B> = C> = D，则在新列中检查True应该返回True，否则返回False。

因此，结果数据帧如下所示：

   Identifier                    A \
0       12253  1989-11-09 00:00:00   
1       11413  1990-09-03 00:00:00   
2       12337  1977-09-07 00:00:00   
3       10610  1994-08-24 00:00:00   
4        7493  1993-08-22 00:00:00   

                      B                              C  \
0   2012-03-01 00:00:00            2015-04-01 00:00:00   
1   2009-08-16 00:00:00            2015-05-18 00:00:00   
2   1977-09-07 00:00:00            1977-09-07 00:00:00   
3   2009-01-09 00:00:00            2015-03-01 00:00:00   
4   2002-06-03 00:00:00            2015-02-16 00:00:00   

                      D       Check\
0   2012-01-03 00:00:00       False
1   2015-05-18 00:00:00       True
2   1977-09-07 00:00:00       True 
3   2015-03-01 00:00:00       True
4   2015-02-16 00:00:00       True

我尝试过

df['Check'] = np.where(df['A'] >= df['B']>= df['C']>= df['D'], 'True', 'False')

但是无法使它正常工作。

任何帮助将不胜感激。

谢谢

Answer 1

使用DataFrame.eval：

index

Answer 2

使用类似于np.diff()的df.diff（）并检查是否有任何元素低于pd.Timedelta（0）。

df['check'] = ~df.diff(axis=1).lt(pd.Timedelta(0)).any(1)

完整示例：

import pandas as pd
import numpy

np.random.seed(333)

# Random dates from:  https://stackoverflow.com/questions/50559078/
def pp(start, end, n):
    start_u = start.value//10**9
    end_u = end.value//10**9
    return pd.DatetimeIndex((10**9*np.random.randint(start_u, end_u, n)).view('M8[ns]'))

n = 10

df = pd.DataFrame({
    'A': pp(pd.Timestamp('2018'), pd.Timestamp('2019'), n),
    'B': pp(pd.Timestamp('2018'), pd.Timestamp('2019'), n),
    'C': pp(pd.Timestamp('2018'), pd.Timestamp('2019'), n),
    'D': pp(pd.Timestamp('2018'), pd.Timestamp('2019'), n)
})

df['check'] = ~df.diff(axis=1).lt(pd.Timedelta(0)).any(1)
print(df)

返回：

                    A                   B                   C  \
0 2018-07-30 04:54:04 2018-03-13 00:28:13 2018-08-24 11:01:29   
1 2018-12-26 21:22:20 2018-09-23 14:25:11 2018-08-19 07:21:59   
2 2018-04-29 17:15:57 2018-05-28 12:35:35 2018-10-16 00:19:11   
3 2018-12-11 06:56:35 2018-08-15 00:12:12 2018-08-05 23:47:08   
4 2018-03-04 11:00:03 2018-07-03 07:22:30 2018-09-09 01:45:09   
5 2018-08-22 03:24:30 2018-12-17 17:38:34 2018-01-29 13:02:29   
6 2018-04-21 01:10:14 2018-06-09 20:37:08 2018-04-30 12:30:00   
7 2018-06-27 18:40:46 2018-09-15 10:26:06 2018-05-13 03:51:36   
8 2018-03-18 06:31:24 2018-11-10 06:24:12 2018-02-25 02:58:15   
9 2018-11-08 17:52:19 2018-03-27 01:02:12 2018-03-06 00:10:02   

                    D  check  
0 2018-07-30 16:16:03  False  
1 2018-07-21 23:38:59  False  
2 2018-10-25 03:46:37   True  
3 2018-12-01 07:43:53  False  
4 2018-12-07 16:11:31   True  
5 2018-09-17 14:58:20  False  
6 2018-07-02 09:36:35  False  
7 2018-03-16 23:21:27  False  
8 2018-10-30 11:24:01  False  
9 2018-04-03 12:17:52  False

基准化

%timeit ~df.diff(axis=1).lt(pd.Timedelta(0)).any(1)
%timeit df.eval('A <= B <= C <= D')

10.000行：

#1000 loops, best of 3: 1.58 ms per loop
#100 loops, best of 3: 3.31 ms per loop

10.000.000行：

#1 loop, best of 3: 2.27 s per loop
#1 loop, best of 3: 243 ms per loop

Answer 3

也许有更聪明的方法，但这应该可行

df['Check'] = (df['A'] <= df['B']) & (df['B']<= df['C']) & (df['C'] <= df['D'])

比较数据框日期列并返回新列

3 个答案:

基准化