Oracle 10g, 11g, MySQL, SQL SERVER (不支持 MATCH RECOGNIZE ) 如何将两个数据行警告关闭电源和电源合并为一个

Posted 2023-02-24

【中文标题】Oracle 10g, 11g, MySQL, SQL SERVER (不支持 MATCH RECOGNIZE ) 如何将两个数据行警告关闭电源和电源合并为一个

【英文标题】：How to merge two data row warning off power and on power into one in Oracle 10g, 11g , MySQL , SQL SERVER (no support MATCH RECOGNIZE )

【发布时间】：2019-03-23 21:25:17

【问题描述】：

https://drive.google.com/file/d/1yakmTD5M2MEINdBEg2ckUTb3vw7YPFTO/view?usp=sharing

大家好，

我已经说过链接链接的给定问题，并且可以通过支持模式匹配的 SQL 解决方案或 Wernfried Domscheit 的 ORACLE 12C 来解决该问题：

CREATE TABLE ALERT_EVENT (MEASUREMENT_POINT VARCHAR2(10), TIME_ALERT DATE, STATUS VARCHAR2(10));

INSERT INTO ALERT_EVENT VALUES('PE01', to_date('04/10/2018 00:00:00','dd/mm/yyyy hh24:mi:ss'), 'POWER ON');
INSERT INTO ALERT_EVENT VALUES('PE01', to_date('03/10/2018 00:00:00','dd/mm/yyyy hh24:mi:ss'), 'POWER OFF');
INSERT INTO ALERT_EVENT VALUES('PE01', to_date('02/10/2018 00:00:00','dd/mm/yyyy hh24:mi:ss'), 'POWER ON');
INSERT INTO ALERT_EVENT VALUES('PE01', to_date('01/10/2018 00:00:00','dd/mm/yyyy hh24:mi:ss'), 'POWER OFF');
INSERT INTO ALERT_EVENT VALUES('PE02',  to_date('03/10/2018 00:00:00','dd/mm/yyyy hh24:mi:ss'), 'POWER OFF'); 
INSERT INTO ALERT_EVENT VALUES('PE02',  to_date('02/10/2018 23:00:00','dd/mm/yyyy hh24:mi:ss'), 'POWER ON');
INSERT INTO ALERT_EVENT VALUES('PE02',  to_date('02/10/2018 22:00:00','dd/mm/yyyy hh24:mi:ss'), 'POWER OFF');
INSERT INTO ALERT_EVENT VALUES('PE03', to_date('04/10/2018 10:00:00','dd/mm/yyyy hh24:mi:ss'), 'POWER OFF');
INSERT INTO ALERT_EVENT VALUES('PE03',  to_date('02/10/2018 23:00:00','dd/mm/yyyy hh24:mi:ss'), 'POWER ON');
INSERT INTO ALERT_EVENT VALUES('PE03',  to_date('02/10/2018 22:00:00','dd/mm/yyyy hh24:mi:ss'), 'POWER OFF');
INSERT INTO ALERT_EVENT VALUES('PE03',  to_date('01/10/2018 00:00:00','dd/mm/yyyy hh24:mi:ss'), 'POWER ON');

查询结果使用 ORACLE 匹配模式 你很接近，试试这个：

SELECT * 
FROM ALERT_EVENT ml
MATCH_RECOGNIZE (
     PARTITION BY MEASUREMENT_POINT
     ORDER BY TIME_ALERT
     MEASURES  STRT.TIME_ALERT AS start_tstamp,
               LAST(END.TIME_ALERT) AS end_tstamp
     PATTERN (STRT END*)
     DEFINE
        STRT AS STRT.STATUS LIKE '%OFF%' ,
        END AS END.STATUS LIKE '%ON%' 
     ) MR
ORDER BY MEASUREMENT_POINT ,start_tstamp;

MEASUREMENT_POINT   START_TSTAMP           END_TSTAMP
==================  =====================  ====================
PE01                2018-10-01 00:00:00    2018-10-02 00:00:00
PE01                2018-10-03 00:00:00    2018-10-04 00:00:00
PE02                2018-10-02 22:00:00    2018-10-02 23:00:00
PE02                2018-10-03 00:00:00 
PE03                2018-10-02 22:00:00    2018-10-02 23:00:00
PE03                2018-10-04 10:00:00 

问题：但是执行到 Oracle 旧版本或 mysql 或 SQL Server，此功能不支持用于模式匹配的 SQL。哪个解决方案可以针对不使用 SQL 进行模式匹配的给定问题进行一般查询？

非常感谢！

【参考方案1】：

您可以在 Oracle 12c 之前的版本中执行此操作，方法是创建一个新列进行分组，然后为每个测量点和新分组列查找相关状态的最小/最大 time_alerts。

分组列是通过计算一个运行总和并在每次遇到POWER OFF状态时将其增加1来生成的。

这给了我们一个类似于以下内容的查询：

select measurement_point,
       min(case when status = 'POWER OFF' then time_alert end) start_tstamp,
       max(case when status = 'POWER ON' then time_alert end) end_tstamp
from   (select measurement_point,
               time_alert,
               status,
               sum(case when status = 'POWER OFF' then 1 else 0 end) over (partition by measurement_point order by time_alert) grp
        from   alert_event)
group by measurement_point, grp
having min(case when status = 'POWER OFF' then time_alert end) is not null
order by measurement_point, min(time_alert);

MEASUREMENT_POINT START_TSTAMP        END_TSTAMP
----------------- ------------------- -------------------
PE01              01/10/2018 00:00:00 02/10/2018 00:00:00
PE01              03/10/2018 00:00:00 04/10/2018 00:00:00
PE02              02/10/2018 22:00:00 02/10/2018 23:00:00
PE02              03/10/2018 00:00:00 
PE03              02/10/2018 22:00:00 02/10/2018 23:00:00
PE03              04/10/2018 10:00:00 

请注意，这会将状态为“POWER ON”的连续行分组为一组；如果这不是您想要的，您将不得不尝试如何生成 grp 列。