2009/9/2 Tom Lane <tgl@sss.pgh.pa.us>:
> Pavel Stehule <pavel.stehule@gmail.com> writes:
>> create table x1 (a integer);
>> insert into x1 values(2),(2),(3),(3),(4),(4),(5),(5),(6),(6),(6),(8),(9),(9),(10),(10);
>
>> postgres=# select row_number() over (order by a), row_number() over
>> (order by a desc) from x1;
>> row_number | row_number
>> ------------+------------
>> 16 | 1
>> 15 | 2
>> 14 | 3
>> 11 | 4
>> 13 | 5
>> 12 | 6
>> 9 | 7
>> 10 | 8
>> 7 | 9
>> 8 | 10
>> 5 | 11
>> 6 | 12
>> 4 | 13
>> 3 | 14
>> 1 | 15
>> 2 | 16
>> (16 rows)
>
>> I am not sure, is this correct?
>
> I don't see any grounds for arguing that it's wrong. The results for
> rows with equal "a" values are indeterminate.
I can understand it. So I found Joe Celko's bug :)
regards
Pavel Stehule
>
> regards, tom lane
>
