Re: speed this up?

Just for the record, another approach is to use an analytic function to avoid the self join.

select column1 from (

    select row_number() over (partition by column_name) rn,

           column1
    from table1
    where column2 > value
)
where rn = 2

Using rn=2 means that single rows get skipped, and only column1 values with more than 1 row will make it through (once each).

Relative performance - depends... If there are not many duplicates or worse, this could be the best way.

Regards Nigel

2008/10/22 Guang Mei <guangmei_at_yahoo.com>

> Oracle 10.2.0.1.0
>
> I have a sql that is slow to run:
>
> create table tmp_table nologging as
> (select column1 from table1 where column2 > some_value
> having count(*) > 1 group by column1);
>
> table1 is large, column2 has index on it. The count(*) value in the above
> query is quite big (typically in thousands) . Since I only need to check if
> the count(*) is large than 1, is there a way (say to use rownum somehow ) so
> that when count(*) gets up to 2, it stop counting (using stopkey), therefore
> I can have the above query run faster?
>
> Thanks.
> Guang
>
>
>
> --
> http://www.freelists.org/webpage/oracle-l
>
>
>

--
http://www.freelists.org/webpage/oracle-l