BUG #17357: When working with the window_function, the Where Clause does not work properly!

Started by PG Bug reporting formover 4 years ago2 messagesbugs
Jump to latest
#1PG Bug reporting form
noreply@postgresql.org

The following bug has been logged on the website:

Bug reference: 17357
Logged by: Muhammadxon Najimov
Email address: muhammadxonnajimov@outlook.com
PostgreSQL version: 14.1
Operating system: Fedora 35 (Workstation Edition)
Description:

Hi,

1. I created table like this:

create table elements (
id serial primary key,
priority smallint default 0
);
// CREATE TABLE [OK]

2. And insterted some rows:
insert into elements (priority) values
(1), (2), (3), (0), (0),
(3), (2), (1), (0), (0),
(0), (0), (0), (0), (0),
(1), (1), (2), (2), (3)
;
// INSERT 0 20 [OK]

3. Wrote Query
select
*,
row_number() over(
partition by priority order by priority desc
) as n
from elements
;

4. and the result was as follows:
// (20 rows) [OK]

All is well. then I took the result to a virtual table:
select *
from (
select
*,
row_number() over(
partition by priority order by priority desc
) as n
from elements
) as tmp
;

// (20 rows) [OK]

The result is the same, but here WHERE clause is not working properly:
select *
from (
select
*,
row_number() over(
partition by priority order by priority desc
) as n
from elements
) as tmp
WHERE n = 1 AND priority = 3
;

Is that logical?
if you look at the picture above, the id number of the record with n = 1 and
pr = 3 is not 3. Ok priority equals to 3 but n not equals 1.

#2Pantelis Theodosiou
ypercube@gmail.com
In reply to: PG Bug reporting form (#1)
Re: BUG #17357: When working with the window_function, the Where Clause does not work properly!

On Fri, Jan 7, 2022 at 11:06 AM PG Bug reporting form <
noreply@postgresql.org> wrote:

The following bug has been logged on the website:

Bug reference: 17357
Logged by: Muhammadxon Najimov
Email address: muhammadxonnajimov@outlook.com
PostgreSQL version: 14.1
Operating system: Fedora 35 (Workstation Edition)
Description:

Hi,

1. I created table like this:

create table elements (
id serial primary key,
priority smallint default 0
);
// CREATE TABLE [OK]

2. And insterted some rows:
insert into elements (priority) values
(1), (2), (3), (0), (0),
(3), (2), (1), (0), (0),
(0), (0), (0), (0), (0),
(1), (1), (2), (2), (3)
;
// INSERT 0 20 [OK]

3. Wrote Query
select
*,
row_number() over(
partition by priority order by priority desc
) as n
from elements
;

4. and the result was as follows:
// (20 rows) [OK]

All is well. then I took the result to a virtual table:
select *
from (
select
*,
row_number() over(
partition by priority order by priority desc
) as n
from elements
) as tmp
;

// (20 rows) [OK]

The result is the same, but here WHERE clause is not working properly:
select *
from (
select
*,
row_number() over(
partition by priority order by priority desc
) as n
from elements
) as tmp
WHERE n = 1 AND priority = 3
;

Is that logical?
if you look at the picture above, the id number of the record with n = 1
and
pr = 3 is not 3. Ok priority equals to 3 but n not equals 1.

This is not a bug.
Your ORDER BY inside the window function is not deterministic:

over(
partition by priority order by priority desc
)

The rows with same priority are ordered based on the (same) priority value
so the optimizer is free to choose any order to assign row numbers. In one
query it chooses to some order, in the second it chooses a different one.

Try using a deterministic ordering and you'll get consistent results, eg:

over(
partition by priority order by id desc
)

or

over(
partition by priority order by id
)

Best regards
Pantelis Theodosiou