oracel 去掉重复记录

时间:2021-01-07 06:26:38
 当设计表的时候没有建组合字段唯一约束,以后需要增加这一约束时,却发现表里已经有了很多重复记录了。

        请看看我用的去掉表里组合字段重复的记录方法:

        假设原始表名为source_table,字段名1为field_name1,字段名2为field_name2。

        (当然稍加修改也可以用到三个及以上组合字段重复的情况)

        第一步: 生成组合字段重复的临时表source_dup_simple
        create table source_dup_simple
        nologging
        pctfree 1 pctused 99
        as select field_name1,field_name2,count(0) as num from source_table
        group by field_name1,field_name2 having count(0)>1;

        第二步: 生成组合字段重复的主表里完整记录的临时表source_dup
        create table source_dup
        nologging
        pctfree 1 pctused 99
        as select t1.* from source_table t1,source_dup_simple t2
        where t1.field_name1=t2.field_name1 and t1.field_name2=t2.field_name2;

        第三步: 删去source_dup里的重复记录

        --可选择:保留rowid小的记录
        delete from source_dup a where rowid > (
        select min(rowid) from source_dup b where
        a.field_name1 = b.field_name1 and a.field_name2=b.field_name2);
        commit;

        --可选择:保留rowid大的记录
        delete from source_dup a where rowid < (
        select max(rowid) from source_dup b where
        a.field_name1 = b.field_name1 and a.field_name2=b.field_name2);
        commit;

        注意:如果操作一万条以上的记录最好在source_dup的field_name1和field_name2字段上建索引.

             如果想按别的删除规则,如保留日期最新的记录:

             --可选择:保留时间字段date_field大的记录

             delete from source_dup a where date_field < (
                select max(date_field) from source_dup b where
                a.field_name1 = b.field_name1 and a.field_name2=b.field_name2);
             commit;

             --可选择:保留时间字段date_field小的记录

             delete from source_dup a where date_field > (
                select min(date_field) from source_dup b where
                a.field_name1 = b.field_name1 and a.field_name2=b.field_name2);
             commit;

             如果时间字段上有重复,还需要再次根据rowid来删一次

             delete from source_dup a where rowid < (
                select max(rowid) from source_dup b where
                a.field_name1 = b.field_name1 and a.field_name2=b.field_name2);
             commit;

        第四步: 删去所有重复组合字段原始表里记录
        delete from source_table
        where field_name1||field_name2 in (select field_name1||field_name2 from source_dup_simple);
        commit;

        注意:如果操作一万条以上的记录最好在source_table的field_name1和field_name2字段上建索引.

        第五步: 把剩下的没有重复的记录插回原始表
        insert into source_table select * from source_dup;
        commit;