Oracle删除重复记录的几种方式

Oracle删除重复记录的几种方式

如果把一个文件多次导入数据库，可能会引入重复记录，那么有哪些方法可以删除重复记录呢？

 REATE TABLE tbl_test(
  SER_NO NUMBER,
  FST_NM VARCHAR2(30),
  DEPTID NUMBER,
  CMNT  VARCHAR2(30));
 
 INSERT INTO tbl_test VALUES(1, 'aaaaa', 2004, 'xxx');
 INSERT INTO tbl_test VALUES(2, 'bbbbb', 2005, 'yyy');
 INSERT INTO tbl_test VALUES(1, 'aaaaa', 2004, 'xxx');
 INSERT INTO tbl_test VALUES(1, 'aaaaa', 2004, 'xxx');
 INSERT INTO tbl_test VALUES(3, 'ccccc', 2005, 'zzz');
 INSERT INTO tbl_test VALUES(2, 'bbbbb', 2005, 'yyy');
 
1.Using MIN(rowid) 最常用的方法，但是数据量大的话执行会很长时间
 
 DELETE FROM tbl_test
    WHERE ROWID NOT IN (SELECT  MIN (ROWID)
          FROM tbl_test
        GROUP BY ser_no, fst_nm, deptid, cmnt);
       
2.Using MIN(rowid) & Join 跟第一条差不多

 DELETE FROM tbl_test t
    WHERE t.ROWID NOT IN (SELECT MIN (b.ROWID)
          FROM tbl_test b
        WHERE b.ser_no = t.ser_no
          AND b.fst_nm = t.fst_nm
          AND b.deptid = t.deptid
          AND b.cmnt  = t.cmnt);
         
3.Using Subquery

 DELETE FROM tbl_test
 WHERE ser_no IN (SELECT ser_no FROM tbl_test GROUP BY ser_no, fst_nm, deptid, cmnt HAVING COUNT (*) > 1)
 AND fst_nm IN (SELECT fst_nm FROM tbl_test GROUP BY ser_no, fst_nm, deptid, cmnt HAVING COUNT (*) > 1)
 AND deptid IN (SELECT deptid FROM tbl_test GROUP BY ser_no, fst_nm, deptid, cmnt HAVING COUNT (*) > 1)
 AND cmnt  IN (SELECT cmnt  FROM tbl_test GROUP BY ser_no, fst_nm, deptid, cmnt HAVING COUNT (*) > 1)
 AND ROWID NOT IN (SELECT  MIN (ROWID)
 FROM tbl_test
    GROUP BY ser_no, fst_nm, deptid, cmnt
 HAVING COUNT (*) > 1)         
       
4. Using Nested Subqueries     

 DELETE FROM tbl_test a WHERE (a.ser_no, a.fst_nm, a.deptid, a.cmnt) IN (SELECT b.ser_no, b.fst_nm, b.deptid, b.cmnt
 FROM tbl_test b WHERE a.ser_no = b.ser_no AND a.fst_nm = b.fst_nm AND a.deptid = b.deptid AND a.cmnt  = b.cmnt AND
 a.ROWID  > b.ROWID);
               
5. Using Analytic Fucntions: 对于大表这是最有效的方法

 DELETE FROM tbl_test WHERE ROWID IN (SELECT rid FROM (SELECT ROWID rid,
 ROW_NUMBER () OVER (PARTITION BY ser_no, fst_nm, deptid, cmnt ORDER BY ROWID) rn FROM tbl_test)WHERE rn <> 1);             
     
6. CREATE-DROP-RENAME 对资源使用比较合理，特别对于大表。但是如果需要回滚则会产生大量undo日志信息。
 
 CREATE  TABLE tbl_test1 NOLOGGING AS SELECT tbl_test .*
 FROM tbl_test tbl_test WHERE ROWID IN (SELECT rid
 FROM (SELECT ROWID rid, ROW_NUMBER() OVER (PARTITION BY ser_no, fst_nm, deptid, cmnt ORDER BY ROWID) rn
 FROM tbl_test) WHERE rn=1);
     
 DROP TABLE tbl_test; --drop the original table with lots of duplicate 
 
 RENAME tbl_test1 TO tbl_test; -- your original table without duplicates.