Re: Partitions: \d vs \d+

Amit Langote <langote_amit_f8@lab.ntt.co.jp>

From: Amit Langote <Langote_Amit_f8@lab.ntt.co.jp>
To: Maksim Milyutin <milyutinma@gmail.com>, jesper.pedersen@redhat.com, pgsql-hackers@postgresql.org
Date: 2017-09-29T01:33:11Z
Lists: pgsql-hackers
On 2017/09/28 22:19, Maksim Milyutin wrote:
> I also noticed ambiguity in printing "No partition constraint" in
> non-verbose mode and "Partition constraint:..." in verbose one for
> partition tables regardless of the type of partition.
> Attached small patch removes any output about partition constraint in
> non-verbose mode.

Patch looks good.

So, we should be looking at partconstraintdef only when verbose is true,
because that's only when we set it to a valid value.  Now, if
partconstraintdef is NULL even after verbose is true, that means backend
returned that there exists no constraint for that partition, which I
thought would be true for a default partition (because the commit that
introduced default partitions also introduced "No partition constraint"),
but it's really not.

For example, \d and \d+ show contradictory outputs for a default partition.

create table p (a int) partition by list (a);
create table p1 partition of p for values in (1);
create table pd partition of p default;

\d pd
                 Table "public.pd"
 Column |  Type   | Collation | Nullable | Default
--------+---------+-----------+----------+---------
 a      | integer |           |          |
Partition of: p DEFAULT
No partition constraint

\d+ pd
                                    Table "public.pd"
 Column |  Type   | Collation | Nullable | Default | Storage | Stats
target | Description
--------+---------+-----------+----------+---------+---------+--------------+-------------
 a      | integer |           |          |         | plain   |              |
Partition of: p DEFAULT
Partition constraint: (NOT ((a IS NOT NULL) AND (a = ANY (ARRAY[1]))))


Perhaps, there is no case when "No partition constraint" should be output,
but I may be missing something.

Thanks,
Amit



Commits

  1. psql: Don't try to print a partition constraint we didn't fetch.

  2. Allow a partitioned table to have a default partition.

  3. Fix pgbench TAP tests to work with --disable-thread-safety.