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;