建立測試資料

create table nayi224_180824(col_1 varchar2(10), col_2 varchar2(10), col_3 varchar2(10));

insert into nayi224_180824
select 1, 2, 3 from dual union all
select 1, 2, 3 from dual union all
select 5, 2, 3 from dual union all
select 10, 20, 30 from dual ;

commit;

select*from nayi224_180824;

針對指定列，查出去重後的結果集

distinct

select distinct t1.* from nayi224_180824 t1;

方法侷限性很大，因為它只能對全部查詢的列做去重。如果我想對col_2,col3去重，那我的結果集中就只能有col_2,col_3列，而不能有col_1列。

select distinct t1.col_2, col_3 from nayi224_180824 t1

不過它也是最簡單易懂的寫法。

row_number()

select *
  from (select t1.*,
               row_number() over(partition by t1.col_2, t1.col_3 order by 1) rn
          from nayi224_180824 t1) t1
 where t1.rn = 1
;

寫法上要麻煩不少，但是有更大的靈活性。

針對指定列，查出所有重複的行

count having

select *
  from nayi224_180824 t
 where (t.col_2, t.col_3) in (select t1.col_2, t1.col_3
                                from nayi224_180824 t1
                               group by t1.col_2, t1.col_3
                              having count(1) > 1)

要查兩次表，效率會比較低。不推薦。

count over

select *
  from (select t1.*,
               count(1) over(partition by t1.col_2, t1.col_3) rn
          from nayi224_180824 t1) t1
 where t1.rn > 1
;

只需要查一次表，推薦。

刪除所有重複的行

delete from nayi224_180824 t
 where t.rowid in (
                   select rid
                     from (select t1.rowid rid,
                                   count(1) over(partition by t1.col_2, t1.col_3) rn
                              from nayi224_180824 t1) t1
                    where t1.rn > 1);

就是上面的語句稍作修改。

刪除重複資料並保留一條

分析函式法

delete from nayi224_180824 t
 where t.rowid in (select rid
                     from (select t1.rowid rid,
                                  row_number() over(partition by t1.col_2, t1.col_3 order by 1) rn
                             from nayi224_180824 t1) t1
                    where t1.rn > 1);

擁有分析函式一貫的靈活性高的特點。可以為所欲為的分組，並通過改變orderby從句來達到像”保留最大id“這樣的要求。

group by

delete from nayi224_180824 t
 where t.rowid not in
       (select max(rowid) from nayi224_180824 t1 group by t1.col_2, t1.col_3);

犧牲了一部分靈活性，換來了更高的效率。