Re: Why does the OID jump by 3 when creating tables?

[Guillaume Lelarge <guillaume@xxxxxxxxxxxx>]

Hi,

Le sam. 30 oct. 2021 à 10:55, Daniel Westermann (DWE) <daniel.westermann@xxxxxxxxxxxxxxxx> a écrit :

Hi all,

as I could not find the reason in the source code, can someone tell me why the OID counter jumps by 3 between two create table statements?

postgres=# create table t1 ( a int );
CREATE TABLE
postgres=# create table t2 ( a int );
CREATE TABLE
postgres=# select oid,relname from pg_class where relname in ('t1','t2');
  oid  | relname
-------+---------
 16453 | t1
 16456 | t2
(2 rows)

These seems not to happen with other objects, e.g. namespaces:

postgres=# create schema a;
CREATE SCHEMA
postgres=# create schema b;
CREATE SCHEMA
postgres=# select oid,nspname from pg_namespace where nspname in ('a','b');
  oid  | nspname
-------+---------
 16459 | a
 16460 | b
(2 rows)

... or indexes:

postgres=# select oid,relname from pg_class where relname in ('i1','i2');
  oid  | relname
-------+---------
 16461 | i1
 16462 | i2

When you create a table, it also creates two data types: tablename and _tablename. For example, for your table t1, you should have a t1 type and a _t1 type. Both have OIDs. On my cluster, your example gives me:

# select oid,relname from pg_class where relname in ('t1','t2');
┌───────┬─────────┐
│  oid  │ relname │
├───────┼─────────┤
│ 24635 │ t1      │
│ 24638 │ t2      │
└───────┴─────────┘
(2 rows)

Time: 0.507 ms

# select oid, typname from pg_type where typname like '%t1' or typname like '%t2' and oid>24000 order by oid;
┌───────┬─────────┐
│  oid  │ typname │
├───────┼─────────┤
│ 24636 │ _t1     │
│ 24637 │ t1      │
│ 24639 │ _t2     │
│ 24640 │ t2      │
└───────┴─────────┘
(4 rows)

Time: 1.203 ms

The jump between t1 OID (24635) and t2 OID (24638) is the _t1 data type OID (24636) and the t1 data type OID (24637).

--

Guillaume.