sql - 根据 Hive 中的组类别移动日期列的行

Question

我有这些带有位置和日期的列。我想使用 Hive HQL 将每个位置中的日期列向下移动，而不是整个列。我尝试使用 LAG() 但不支持 Hive，也不支持 datediff() 但它不会给我带来移动日期，只会带来日期差异.

Location    Dates
0000076-0   01/01/2020
0000076-0   28/01/2020
0000076-0   28/02/2020
0000076-0   01/04/2020
0000076-0   28/04/2020
0000076-0   28/05/2020
0000076-0   01/07/2020
000478-2    01/01/2020
000478-2    01/02/2020
000478-2    26/02/2020
000478-2    01/04/2020
000478-2    26/04/2020
000478-2    26/05/2020
000478-2    01/07/2020
0005655-3   01/01/2020
0005655-3   07/02/2020
0005655-3   09/03/2020
0005655-3   07/04/2020
0005655-3   07/05/2020
0005655-3   08/06/2020
0005655-3   01/07/2020
0010828-6   01/01/2020
0010828-6   31/01/2020
0010828-6   29/02/2020
0010828-6   01/04/2020
0010828-6   01/05/2020

所需输出

Location    Dates       Dates_shifted_down_per_location
0000076-0   01/01/2020  null
0000076-0   28/01/2020  01/01/2020     
0000076-0   28/02/2020  28/01/2020     
0000076-0   01/04/2020  28/02/2020     
0000076-0   28/04/2020  01/04/2020     
0000076-0   28/05/2020  28/04/2020     
0000076-0   01/07/2020  28/05/2020     
000478-2    01/01/2020  null     
000478-2    01/02/2020  01/01/2020     
000478-2    26/02/2020  01/02/2020     
000478-2    01/04/2020  26/02/2020     
000478-2    26/04/2020  01/04/2020     
000478-2    26/05/2020  26/04/2020     
000478-2    01/07/2020  26/05/2020     
0005655-3   01/01/2020  null     
0005655-3   07/02/2020  01/01/2020     
0005655-3   09/03/2020  07/02/2020     
0005655-3   07/04/2020  09/03/2020     
0005655-3   07/05/2020  07/04/2020     
0005655-3   08/06/2020  07/05/2020     
0005655-3   01/07/2020  08/06/2020     
0010828-6   01/01/2020  null     
0010828-6   31/01/2020  01/01/2020     
0010828-6   29/02/2020  31/01/2020     
0010828-6   01/05/2020  29/02/2020      
0010828-6   01/06/2020  01/05/2020

Answer 1

Hive 支持 lag() 。您想要的代码是:

select t.*,
       lag(date) over (partition by location order by date) as prev_date
from t;