Thread: Overlapping values (?) in multi-column partitioned tables

Hi,

I am clearly not understanding something.  Consider:

> xof=# create table t (pk bigint not null, ts timestamp not null) partition by range (ts, pk);
> CREATE TABLE
> xof=# create table t1 partition of t for values from ('2024-01-01'::timestamp, minvalue) to ('2024-02-01'::timestamp,
maxvalue);
> CREATE TABLE
> xof=# create table t2 partition of t for values from ('2024-02-01'::timestamp, minvalue) to ('2024-03-01'::timestamp,
maxvalue);
> ERROR:  partition "t2" would overlap partition "t1"
> LINE 1: ...on of t for values from ('2024-02-01'::timestamp, minvalue) ...
>                                                              ^
> xof=#

In what way do those partitions overlap?  In other words, there does not appear to be a value of (ts, pk) having
'2024-01-01'::timestamp<= ts < '2024-03-01'::timestamp for any pk where it would be ambiguous which partition that row
wouldgo into.  (I'm imagining it on a cartesian plane, and there isn't any overlap between the rectangles those
partitionboundaries define.) 

I'm obviously missing something, but... what?  Thank you!

The comparisons for partitions are lower <= value < upper, so this works fine:

> xof=# create table t (pk bigint not null, ts timestamp not null) partition by range (ts);
> CREATE TABLE
> xof=# create table t1 partition of t for values from ('2024-01-01'::timestamp) to ('2024-02-01'::timestamp);
> CREATE TABLE
> xof=# create table t2 partition of t for values from ('2024-02-01'::timestamp) to ('2024-03-01'::timestamp);
> CREATE TABLE

In both cases, it's unambiguous that a record with ('2024-02-01'::timestamp) would go into t2.

> On Sep 10, 2024, at 15:23, Brent Wood <Brent.Wood@niwa.co.nz> wrote:
>
> I'm not sure of the less than vs less than or equals in this one, so it may be my ignorance... but which partition
getsrecords with a date of '24-02-01", it seems that without a more explicit definition there is ambiguity on dates at
thepartition limits when those dates are common to multiple partitions? 
>
> To avoid this shouldn't the definition be:
>
>> xof=# create table t (pk bigint not null, ts timestamp not null) partition by range (ts, pk);
>> CREATE TABLE
>> xof=# create table t1 partition of t for values from ('2024-01-01'::timestamp, minvalue) to
('2024-01-31'::timestamp,maxvalue); 
>> CREATE TABLE
>> xof=# create table t2 partition of t for values from ('2024-02-01'::timestamp, minvalue) to
('2024-02-29'::timestamp,maxvalue); 
>
>
> Brent Wood
>
> Principal Technician, Fisheries
> NIWA
> DDI:  +64 (4) 3860529
> From: Christophe Pettus <xof@thebuild.com>
> Sent: Wednesday, September 11, 2024 10:13
> To: pgsql-general <pgsql-general@postgresql.org>
> Subject: Overlapping values (?) in multi-column partitioned tables
> Hi,
>
> I am clearly not understanding something.  Consider:
>
>> xof=# create table t (pk bigint not null, ts timestamp not null) partition by range (ts, pk);
>> CREATE TABLE
>> xof=# create table t1 partition of t for values from ('2024-01-01'::timestamp, minvalue) to
('2024-02-01'::timestamp,maxvalue); 
>> CREATE TABLE
>> xof=# create table t2 partition of t for values from ('2024-02-01'::timestamp, minvalue) to
('2024-03-01'::timestamp,maxvalue); 
>> ERROR:  partition "t2" would overlap partition "t1"
>> LINE 1: ...on of t for values from ('2024-02-01'::timestamp, minvalue) ...
>>                                                            ^
>> xof=#
>
> In what way do those partitions overlap?  In other words, there does not appear to be a value of (ts, pk) having
'2024-01-01'::timestamp<= ts < '2024-03-01'::timestamp for any pk where it would be ambiguous which partition that row
wouldgo into.  (I'm imagining it on a cartesian plane, and there isn't any overlap between the rectangles those
partitionboundaries define.) 
>
> I'm obviously missing something, but... what?  Thank you!
>
> Brent Wood
> Principal Technician - GIS and Spatial Data Management
> Programme Leader - Environmental Information Delivery
> +64-4-386-0529
>
> National Institute of Water & Atmospheric Research Ltd (NIWA)
> 301 Evans Bay Parade Hataitai Wellington New Zealand
> Connect with NIWA: niwa.co.nz Facebook LinkedIn Twitter Instagram YouTubeTo ensure compliance with legal requirements
andto maintain cyber security standards, NIWA's IT systems are subject to ongoing monitoring, activity logging and
auditing.This monitoring and auditing service may be provided by third parties. Such third parties can access
informationtransmitted to, processed by and stored on NIWA's IT systems. 
> Note: This email is intended solely for the use of the addressee and may contain information that is confidential or
subjectto legal professional privilege. If you receive this email in error please immediately notify the sender and
deletethe email. 

On Tue, 2024-09-10 at 15:13 -0700, Christophe Pettus wrote:
> I am clearly not understanding something.  Consider:
>
> > xof=# create table t (pk bigint not null, ts timestamp not null) partition by range (ts, pk);
> > CREATE TABLE
> > xof=# create table t1 partition of t for values from ('2024-01-01'::timestamp, minvalue) to
('2024-02-01'::timestamp,maxvalue); 
> > CREATE TABLE
> > xof=# create table t2 partition of t for values from ('2024-02-01'::timestamp, minvalue) to
('2024-03-01'::timestamp,maxvalue); 
> > ERROR:  partition "t2" would overlap partition "t1"
> > LINE 1: ...on of t for values from ('2024-02-01'::timestamp, minvalue) ...
> >                                                               ^
> > xof=#
>
> In what way do those partitions overlap?

In this way:

SELECT ROW('2024-02-01'::timestamp, '9223372036854775807'::bigint)
    <= ROW('2024-02-01'::timestamp, '-9223372036854775808'::bigint);

 ?column?
══════════
 f
(1 row)

So the upper limit of the first partition is strictly bigger than the lower end
of the second partition.

"record" types have the same lexicographical sorting order as ORDER BY clauses.

Perhaps your confusion is that you expect the first partition to only go up to
and including ('2023-12-31 23:59:59.999999', 9223372036854775806), but that's not
the case.  The biggest value you can store in the first partition is
('2024-02-01 00:00:00', 9223372036854775806).

You should specify the upper bound as ('2023-12-31 23:59:59.999999', MAXVALUE).

Yours,
Laurenz Albe

On Wed, 11 Sept 2024 at 10:57, Laurenz Albe <laurenz.albe@cybertec.at> wrote:
> You should specify the upper bound as ('2023-12-31 23:59:59.999999', MAXVALUE).

Or maybe do multi-level partitioning.

(it seems strange to always have MINVALUE and MAXVALUE as the range.
I'm guessing that was just an example.)

David

> On Sep 10, 2024, at 15:57, Laurenz Albe <laurenz.albe@cybertec.at> wrote:
> So the upper limit of the first partition is strictly bigger than the lower end
> of the second partition.
>
> "record" types have the same lexicographical sorting order as ORDER BY clauses.

Ah, OK, got it.  I was thinking of them as a points on a plane, which was misleading.  Thanks!

> On Sep 10, 2024, at 16:02, David Rowley <dgrowleyml@gmail.com> wrote:
> 
> (it seems strange to always have MINVALUE and MAXVALUE as the range.
> I'm guessing that was just an example.)

Correct.