oracle中對insert into on duplicate key 語句的替代

insert into on duplicate key 是mysql中的語句，oracle中沒有直接對應。不過采用merge into語句可以起到相同效果。

比如有一個表TTT，我們在no,start_time,code三個字段上建立了唯一索引；希望插入新記錄時，如果在唯一索引上有重復的就更新，沒有就直接插入。

 

MERGE INTO TTT
USING 
  (select count(*) num from TTT where no = '7099' and start_time = to_date('20200817204243','yyyyMMddHH24miss') and code = '01060') q 
ON (q.num=1) 
WHEN MATCHED THEN
  update set end_time = to_date('20200817210000','yyyyMMddHH24miss'), state = 1 where  no = '7099' and start_time = to_date('20200817204243','yyyyMMddHH24miss') and code = '01060'
WHEN NOT MATCHED THEN 
  insert (id, no, sec, stor_time, o_time, start_time, end_time, state, frs, code) 
    values('5', '7099', '1', TO_DATE('2020-08-17 20:52:32', 'SYYYY-MM-DD HH24:MI:SS'), TO_DATE('2020-08-17 20:52:32', 'SYYYY-MM-DD HH24:MI:SS'),
           TO_DATE('2020-08-17 22:31:31', 'SYYYY-MM-DD HH24:MI:SS'), TO_DATE('2020-08-17 22:31:50', 'SYYYY-MM-DD HH24:MI:SS'),'0','145','01200');

后經人指點，發現可以改成下面這樣，效率提高：

MERGE INTO TTT
USING dual ON (no = '7099' and start_time = to_date('20200817204243','yyyyMMddHH24miss') and code = '01060')
WHEN MATCHED THEN
  update set end_time = to_date('20200817210000','yyyyMMddHH24miss'), state = 1 where  no = '7099' and start_time = to_date('20200817204243','yyyyMMddHH24miss') and code = '01060'
WHEN NOT MATCHED THEN 
  insert (id, no, sec, stor_time, o_time, start_time, end_time, state, frs, code) 
    values('5', '7099', '1', TO_DATE('2020-08-17 20:52:32', 'SYYYY-MM-DD HH24:MI:SS'), TO_DATE('2020-08-17 20:52:32', 'SYYYY-MM-DD HH24:MI:SS'),
           TO_DATE('2020-08-17 22:31:31', 'SYYYY-MM-DD HH24:MI:SS'), TO_DATE('2020-08-17 22:31:50', 'SYYYY-MM-DD HH24:MI:SS'),'0','145','01200');

新語句使用了dual表；避免了count(*)操作。update部分，需要加where條件，否則可能就變成了全表更新；insert部分實際上和條件之間沒有什么直接關聯，應該由外部程序（比如JDBC中的PreparedStatement設置參數時)來保持一致。