v2.23 Preview Preview release - not for production v2024.1 STS Stable release with standard-term support v2.20 LTS Stable release with long-term support v2.18 STS Stable release with standard-term support v2.14 LTS Stable release with long-term support Unsupported versions

CREATE SEQUENCE

Synopsis

Use the CREATE SEQUENCE statement to create a sequence in the current schema.

Syntax

create_sequence ::= CREATE [ TEMPORARY | TEMP ] SEQUENCE 
                    [ IF NOT EXISTS ] sequence_name sequence_options

sequence_name ::= qualified_name

sequence_options ::= [ AS seq_data_type ]  
                     [ INCREMENT [ BY ] int_literal ]  
                     [ MINVALUE int_literal | NO MINVALUE ]  
                     [ MAXVALUE int_literal | NO MAXVALUE ]  
                     [ START [ WITH ] int_literal ]  
                     [ CACHE positive_int_literal ] [ [ NO ] CYCLE ]

create_sequence

CREATETEMPORARYTEMPSEQUENCEIFNOTEXISTSsequence_namesequence_options

sequence_name

qualified_name

sequence_options

ASseq_data_typeINCREMENTBYint_literalMINVALUEint_literalNOMINVALUEMAXVALUEint_literalNOMAXVALUESTARTWITHint_literalCACHEpositive_int_literalNOCYCLE

Semantics

Specify the name of the sequence (sequence_name). An error is raised if a sequence with that name already exists in the current schema and IF NOT EXISTS is not specified.

The sequence name must be distinct from any other sequences, tables, indexes, views, or foreign tables in the same schema.

sequence_options

TEMPORARY or TEMP

Using this qualifier will create a temporary sequence. Temporary sequences are visible only in the current client session in which they are created and are automatically dropped at the end of the session. See the section Creating and using temporary schema-objects.

AS seq_data_type

Specifies the data type of the sequence. This automatically determines the minimum and maximum values of the sequence. Valid types are smallint, integer, and bigint. Default is bigint.

INCREMENT BY int_literal

Specify the increment value to add to the current sequence value to create a new value. The default value is 1. A positive number

MINVALUE int_literal | NO MINVALUE

Specify the minimum value allowed in the sequence. If this value is reached (in a sequence with a negative increment), nextval() will return an error. If NO MINVALUE is specified, the default value will be used. Default is 1.

MAXVALUE int_literal | NO MAXVALUE

Specify the maximum value allowed in the sequence. If this value is reached, nextval() will return an error. If NO MAXVALUE is specified, the default will be used. Default is 2⁶³-1.

START WITH int_literal

Specify the first value in the sequence. start cannot be less than minvalue. Default is 1.

CACHE int_literal

Specify how many numbers from the sequence to cache in the client. Default is 100.

When YB-TServer ysql_sequence_cache_minval configuration flag is not explicitly turned off (set to 0), the maximum value of the flag and the cache clause will be used.

[ NO ] CYCLE

If CYCLE is specified, the sequence will wrap around once it has reached minvalue or maxvalue. If maxvalue was reached, minvalue will be the next number in the sequence. If minvalue was reached (for a descending sequence), maxvalue will be the next number in a sequence. NO CYCLE is the default.

Cache

In YSQL as in PostgreSQL, the sequence's data is stored in a persistent system table. In YSQL this table has one row per sequence and it stores the sequence data in two values:

last_val

Stores the last value used or the next value to be used.

is_called

Stores whether last_val has been used. If false, last_val is the next value in the sequence. Otherwise, last_val + INCREMENT is the next one.

By default (when INCREMENT is 1), each call to nextval() updates last_val for that sequence. In YSQL, the table holding the sequence's data is replicated as opposed to being in the local file system. Each update to this table requires two RPCs (and will be optimized to one RPC in the future), In any case, the latency experienced by a call to nextval() in YSQL will be significantly higher than the same operation in Postgres. To avoid such performance degradation, Yugabyte recommends using a cache value with a value large enough. Cached values are stored in the memory of the local node, and retrieving such values avoids any RPCs, so the latency of one cache allocation can be amortized over all the numbers allocated for the cache.

SERIAL types create a sequence with a cache with default value of 1. So SERIAL types should be avoided, and their equivalent statement should be used. Instead of creating a table with a SERIAL type like this:

CREATE TABLE t(k SERIAL)

You should create a sequence with a large enough cache first, and then set the column that you want to have a serial type to DEFAULT to nextval() of the sequence.

CREATE SEQUENCE t_k_seq CACHE 10000;
CREATE TABLE t(k integer NOT NULL DEFAULT nextval('t_k_seq'));

Examples

Create a simple sequence that increments by 1 every time nextval() is called.

yugabyte=# CREATE SEQUENCE s;

CREATE SEQUENCE

Call nextval().

yugabyte=# SELECT nextval('s');

 nextval
---------
       1
(1 row)

Create a sequence with a cache of 10,000 values.

yugabyte=# CREATE SEQUENCE s2 CACHE 10000;

CREATE SEQUENCE

In the same session, select nextval().

SELECT nextval('s2');

 nextval
---------
       1
(1 row)

In a different session, select nextval().

SELECT nextval('s2');

nextval
---------
   10001
(1 row)

Create a sequence that starts at 0. MINVALUE also has to be changed from its default 1 to something less than or equal to 0.

CREATE SEQUENCE s3 START 0 MINVALUE 0;

CREATE SEQUENCE

SELECT nextval('s3');

nextval
---------
       0
(1 row)

See also