铅保持和案例

Question

*编辑

我有以下 table:

Table W

No    action_dt              type    status
1      5/6/2008 10:55:11 PM  CREATE  APPROVED
1      1/3/2013 9:52:01 AM   UPDATE  APPROVED
1      1/3/2013 9:57:26 AM   UPDATE  APPROVED
1      4/21/2015 2:19:04 PM  UPDATE  APPROVED
1    10/27/2016 10:40:19 AM  UPDATE  APPROVED
1    2/19/2018 4:18:18 PM    UPDATE  APPROVED
1    11/13/2018 4:13:38 PM   UPDATE  NEW
1    11/15/2018 9:44:55 AM   UPDATE  APPROVED
1    7/17/2019 2:21:52 PM    UPDATE  PENDING
1    7/17/2019 2:21:52 PM    UPDATE  NEW
1   7/17/2019 2:47:31 PM     UPDATE  APPROVED

预期结果：

No  Action_dt              type     status   start_dt               End_dt
1   5/6/2008 10:55:11 PM   CREATE   APPROVED 5/6/2008 10:55:11 PM   5/6/2008 10:55:11 PM
1   1/3/2013 9:52:01 AM    UPDATE   APPROVED 1/3/2013 9:52:01 AM    1/3/2013 9:52:01 AM
1   1/3/2013 9:57:26 AM    UPDATE   APPROVED 1/3/2013 9:57:26 AM    1/3/2013 9:57:26 AM
1   4/21/2015 2:19:04 PM   UPDATE   APPROVED 4/21/2015 2:19:04 PM   4/21/2015 2:19:04 PM
1   10/27/2016 10:40:19 AM UPDATE   APPROVED 10/27/2016 10:40:19 AM 10/27/2016 10:40:19 AM
1   2/19/2018 4:18:18 PM   UPDATE   APPROVED 2/19/2018 4:18:18 PM   2/19/2018 4:18:18 PM
1   11/13/2018 4:13:38 PM   UPDATE  NEW      11/13/2018 4:13:38 PM  
1   11/15/2018 9:44:55 AM   UPDATE  APPROVED 11/15/2018 9:44:55 AM  11/15/2018 9:44:55 AM
1   7/17/2019 2:21:52 PM    UPDATE  PENDING     
1   7/17/2019 2:21:52 PM    UPDATE  NEW      7/17/2019 2:21:52 PM   
1   7/17/2019 2:47:31 PM    UPDATE  APPROVED 7/17/2019 2:47:31 PM   7/17/2019 2:47:31 PM

现在我需要填写缺失的start_dt和end_dt

第 9 行更新挂起为 NULL 当请求类型 = 创建或更新且状态 = 新时，我需要获取上面的第一行

end_dt 第 7、9、10 行

当请求类型 = 创建或更新且状态代码 = 已批准时，采用以下第一个日期。

如何做到这一点？也许在 case 表达式中使用 Lead/Lag with Keep?

对于缺失的数据。预期输出将是（将其添加到上述预期结果中）

       Start_dt                   End_dt
row 7                             11/15/2018 9:44:55 AM
row 9  11/13/2018 4:13:38 PM       7/17/2019 2:47:31 PM
row 10                            7/17/2019 2:47:31 PM

SQL:

select *
,CASE
--WHEN W.REQUEST_TYPE IN ('CREATE','UPDATE') AND W.STATUS_CD IN ('APPROVED','DENIED') THEN W.ACTION_DT
WHEN W.REQUEST_TYPE IN ('CREATE','UPDATE') AND W.STATUS_CD IN ('NEW','APPROVED')  THEN W.ACTION_DT
--WHEN W.REQUEST_TYPE IN ('CREATE','UPDATE') AND W.STATUS_CD IN ('PENDING')  THEN W.ACTION_DT
ELSE NULL
END REQUEST_START_DT

,CASE
WHEN W.REQUEST_TYPE IN ('CREATE','UPDATE') AND W.STATUS_CD IN ('APPROVED','DENIED') THEN W.ACTION_DT
--WHEN W.REQUEST_TYPE IN ('CREATE','UPDATE') AND W.STATUS_CD IN ('PENDING') THEN LEAD((W.ACTION_DT) KEEP (DENSE_RANK ORDER BY W.STATUS_CD) OVER (PARTITION BY ORG_NO) end)
ELSE NULL
END REQUEST_END_DT
FROM APEXIM.CRD_ORG_WORK W
WHERE 1=1
AND W.NO = 1
from W

Answer 1

将 LAG( CASE WHEN <your matches> THEN action_dt END )（或 LEAD）与 IGNORE NULLS 选项一起使用可跳过与 CASE 语句不匹配的行。

Oracle 设置:

CREATE TABLE w ( no, action_dt, request_type, status ) AS
SELECT 1, DATE '2008-05-06' + INTERVAL '22:55:11' HOUR TO SECOND, 'CREATE', 'APPROVED' FROM DUAL UNION ALL
SELECT 1, DATE '2013-01-03' + INTERVAL '09:52:01' HOUR TO SECOND, 'UPDATE', 'APPROVED' FROM DUAL UNION ALL
SELECT 1, DATE '2013-01-03' + INTERVAL '09:57:26' HOUR TO SECOND, 'UPDATE', 'APPROVED' FROM DUAL UNION ALL
SELECT 1, DATE '2015-04-21' + INTERVAL '14:19:04' HOUR TO SECOND, 'UPDATE', 'APPROVED' FROM DUAL UNION ALL
SELECT 1, DATE '2016-10-27' + INTERVAL '10:40:19' HOUR TO SECOND, 'UPDATE', 'APPROVED' FROM DUAL UNION ALL
SELECT 1, DATE '2018-02-19' + INTERVAL '16:18:18' HOUR TO SECOND, 'UPDATE', 'APPROVED' FROM DUAL UNION ALL
SELECT 1, DATE '2018-11-13' + INTERVAL '16:13:38' HOUR TO SECOND, 'UPDATE', 'NEW'      FROM DUAL UNION ALL
SELECT 1, DATE '2018-11-15' + INTERVAL '09:44:55' HOUR TO SECOND, 'UPDATE', 'APPROVED' FROM DUAL UNION ALL
SELECT 1, DATE '2019-07-17' + INTERVAL '14:21:52' HOUR TO SECOND, 'UPDATE', 'PENDING'  FROM DUAL UNION ALL
SELECT 1, DATE '2019-07-17' + INTERVAL '14:21:52' HOUR TO SECOND, 'UPDATE', 'NEW'      FROM DUAL UNION ALL
SELECT 1, DATE '2019-07-17' + INTERVAL '14:47:31' HOUR TO SECOND, 'UPDATE', 'APPROVED' FROM DUAL

查询:

SELECT w.*,
       CASE
       WHEN request_type IN ( 'CREATE', 'UPDATE' ) AND status IN ( 'NEW', 'APPROVED' )
       THEN action_dt
       WHEN request_type IN ( 'CREATE', 'UPDATE' ) AND status = 'PENDING'
       THEN LAG( CASE WHEN request_type IN ( 'CREATE', 'UPDATE' ) AND status = 'NEW' THEN action_dt END )
              IGNORE NULLS OVER ( PARTITION BY no ORDER BY action_dt )
       ELSE NULL
       END AS start_dt,
       CASE
       WHEN request_type IN ( 'CREATE', 'UPDATE' ) AND status IN ( 'NEW', 'APPROVED' )
       THEN action_dt
       WHEN request_type IN ( 'CREATE', 'UPDATE' ) AND status = 'PENDING'
       THEN LEAD( CASE WHEN request_type IN ( 'CREATE', 'UPDATE' ) AND status = 'APPROVED' THEN action_dt END )
              IGNORE NULLS OVER ( PARTITION BY no ORDER BY action_dt )
       ELSE NULL
       END AS end_dt
FROM   w

输出:

NO | ACTION_DT           | REQUEST_TYPE | STATUS   | START_DT            | END_DT             
-: | :------------------ | :----------- | :------- | :------------------ | :------------------
 1 | 2008-05-06 22:55:11 | CREATE       | APPROVED | 2008-05-06 22:55:11 | 2008-05-06 22:55:11
 1 | 2013-01-03 09:52:01 | UPDATE       | APPROVED | 2013-01-03 09:52:01 | 2013-01-03 09:52:01
 1 | 2013-01-03 09:57:26 | UPDATE       | APPROVED | 2013-01-03 09:57:26 | 2013-01-03 09:57:26
 1 | 2015-04-21 14:19:04 | UPDATE       | APPROVED | 2015-04-21 14:19:04 | 2015-04-21 14:19:04
 1 | 2016-10-27 10:40:19 | UPDATE       | APPROVED | 2016-10-27 10:40:19 | 2016-10-27 10:40:19
 1 | 2018-02-19 16:18:18 | UPDATE       | APPROVED | 2018-02-19 16:18:18 | 2018-02-19 16:18:18
 1 | 2018-11-13 16:13:38 | UPDATE       | NEW      | 2018-11-13 16:13:38 | 2018-11-13 16:13:38
 1 | 2018-11-15 09:44:55 | UPDATE       | APPROVED | 2018-11-15 09:44:55 | 2018-11-15 09:44:55
 1 | 2019-07-17 14:21:52 | UPDATE       | PENDING  | 2018-11-13 16:13:38 | 2019-07-17 14:47:31
 1 | 2019-07-17 14:21:52 | UPDATE       | NEW      | 2019-07-17 14:21:52 | 2019-07-17 14:21:52
 1 | 2019-07-17 14:47:31 | UPDATE       | APPROVED | 2019-07-17 14:47:31 | 2019-07-17 14:47:31

db<>fiddle here

铅保持和案例

lead keep and case

sql

oracle

case

lead