[HACKERS] Multi column range partition table - Mailing list pgsql-hackers

From amul sul
Subject [HACKERS] Multi column range partition table
Date
Msg-id CAAJ_b947mowpLdxL3jo3YLKngRjrq9+Ej4ymduQTfYR+8=YAYQ@mail.gmail.com
Whole thread Raw
Responses Re: [HACKERS] Multi column range partition table  (Amit Langote <Langote_Amit_f8@lab.ntt.co.jp>)
List pgsql-hackers
Hi,

While working on the another patch, I came across the case where
I need an auto generated partition for a mutil-column range partitioned
table having following range bound:

PARTITION p1 FROM  (UNBOUNDED, UNBOUNDED) TO (10, 10)
PARTITION p2 FROM  (10, 10)  TO (10, UNBOUNDED)
PARTITION p3 FROM  (10, UNBOUNDED) TO (20, 10)
PARTITION p4 FROM (20, 10) TO (20, UNBOUNDED)
PARTITION p5 FROM (20, UNBOUNDED) TO (UNBOUNDED, UNBOUNDED)

In this, a lower bound of the partition is an upper bound of the
previous partition.

While trying to create p3 partition with (10, UNBOUNDED) to (20, 10) bound,
got an overlap partition error.

Here is the SQL to reproduced this error:

CREATE TABLE range_parted ( i1 int,  i2 int ) PARTITION BY RANGE (i1, i2);
CREATE TABLE p1 PARTITION OF range_parted FOR VALUES FROM (UNBOUNDED,
UNBOUNDED) TO (10, 10);
CREATE TABLE p2 PARTITION OF range_parted FOR VALUES FROM (10, 10) TO
(10, UNBOUNDED);
CREATE TABLE p3   PARTITION OF tab1 FOR VALUES FROM (10, UNBOUNDED) TO (20, 10);

ERROR:  partition "p3" would overlap partition "tab1_p_10_10"

This happened because of UNBOUNDED handling, where it is a negative infinite
if it is in FROM clause.  Wondering can't we explicitly treat this as
a positive infinite value, can we?

Thoughts/Comments?

Regards,
Amul



pgsql-hackers by date:

Previous
From: "Daniel Verite"
Date:
Subject: Re: [HACKERS] PATCH: Batch/pipelining support for libpq
Next
From: Michael Paquier
Date:
Subject: Re: [HACKERS] SQL MERGE patches for PostgreSQL Versions