CREATE TABLE

Description

This statement is used to create a table in a database.

Syntax

CREATE [GLOBAL TEMPORARY] TABLE table_name
      (table_definition_list) [table_option_list] [partition_option] [on_commit_option]
CREATE [GLOBAL TEMPORARY] TABLE table_name
      (table_definition_list) [table_option_list] [partition_option] [AS] select;

table_definition_list:
    table_definition [, table_definition ...]

table_definition:
      column_definition
     | [,
        | [CONSTRAINT [constraint_name]] { PRIMARY KEY|UNIQUE } (column_name) //Add a constraint after all columns are created.
        | [CONSTRAINT [constraint_name]] FOREIGN KEY (column_name, column_name ...) references_clause constraint_state
        | [CONSTRAINT [constraint_name]] CHECK(expression) constraint_state
       ]

column_definition_list:
    column_definition [, column_definition ...]

column_definition:
    column_name data_type
    [VISIBLE|INVISIBLE] [GENERATED BY DEFAULT AS IDENTITY | GENERATED ALWAYS AS IDENTITY]
    {
    [DEFAULT expression]
    [NULL | NOT NULL]
    [CONSTRAINT [constraint_name]] [ PRIMARY KEY|UNIQUE ]       //Add a constraint when the column is created.
    [CONSTRAINT [constraint_name] CHECK(expression) constraint_state]
    [CONSTRAINT [constraint_name] references_clause]
    |
    [GENERATED ALWAYS] AS (expression) [VIRTUAL]
    [NULL | NOT NULL] [UNIQUE KEY] [[PRIMARY] KEY] [UNIQUE LOWER_KEY]
    }

references_clause:
     REFERENCES table_name [ (column_name, column_name ...) ] [ON DELETE {SET NULL | CASCADE}]

constraint_state:
    [RELY | NORELY] [USING INDEX index_option_list] [ENABLE | DISABLE] [VALIDATE | NOVALIDATE]

index_option_list:
    index_option [ index_option ...]

index_option:
      [GLOBAL | LOCAL]
    | block_size
    | compression
    | STORING(column_name_list)

table_option_list:
    table_option [ table_option ...]

table_option:
      TABLEGROUP = tablegroup_name
    | block_size
    | compression
    | ENABLE ROW MOVEMENT
    | DISABLE ROW MOVEMENT
    | physical_attribute
    | parallel_clause
    | DUPLICATE_SCOPE [=] 'none|cluster'
    | TABLE_MODE [=] 'table_mode_value'

physical_attribute_list:
    physical_attribute [physical_attribute]

physical_attribute:
    PCTFREE [=] num
    | PCTUSED num
    | INITRANS num
    | MAXTRANS num
    | STORAGE(storage_option [storage_option] ...)
    | TABLESPACE tablespace

parallel_clause:
    {NOPARALLEL | PARALLEL integer}

table_mode_value:
    NORMAL
    | QUEUING
    | MODERATE
    | SUPER
    | EXTREME

compression:
    NOCOMPRESS
    | COMPRESS { BASIC | FOR OLTP | FOR QUERY [LOW | HIGH]  | FOR ARCHIVE [LOW | HIGH]}

storage_option:
    INITIAL num [K|M|G|T|P|E]
    | NEXT num [K|M|G|T|P|E]
    | MINEXTENTS num [K|M|G|T|P|E]
    | MAXEXTENTS num [K|M|G|T|P|E]

partition_option:
      PARTITION BY HASH(column_name_list)
      [subpartition_option] hash_partition_define
    | PARTITION BY RANGE (column_name_list)
      [subpartition_option] (range_partition_list)
    | PARTITION BY LIST (column_name_list)
      [subpartition_option] (list_partition_list)

/*Template-based subpartitioning*/
subpartition_option:
      SUBPARTITION BY HASH (column_name_list) hash_subpartition_define
    | SUBPARTITION BY RANGE (column_name_list) SUBPARTITION TEMPLATE
      (range_subpartition_list)
    | SUBPARTITION BY LIST (column_name_list) SUBPARTITION TEMPLATE
      (list_subpartition_list)

/*Non-template-based subpartitioning*/
subpartition_option:
      SUBPARTITION BY HASH (column_name_list)
    | SUBPARTITION BY RANGE (column_name_list)
    | SUBPARTITION BY LIST (column_name_list)

subpartition_list:
      (hash_subpartition_list)
    | (range_subpartition_list)
    | (list_subpartition_list)

hash_partition_define:
      PARTITIONS partition_count [TABLESPACE tablespace] [compression]
    | (hash_partition_list)

hash_partition_list:
    hash_partition [, hash_partition ...]

hash_partition:
    partition [partition_name] [subpartition_list/*Only applicable to non-template-based subpartitioning*/]

hash_subpartition_define:
      SUBPARTITIONS subpartition_count
    | SUBPARTITION TEMPLATE (hash_subpartition_list)

hash_subpartition_list:
    hash_subpartition [, hash_subpartition ...]

hash_subpartition:
    subpartition [subpartition_name]

range_partition_list:
    range_partition [, range_partition ...]

range_partition:
    PARTITION [partition_name]
    VALUES LESS THAN {(expression_list) | (MAXVALUE)}
    [subpartition_list/*Only applicable to non-template-based subpartitioning*/]
    [ID = num] [physical_attribute_list] [compression]

range_subpartition_list:
    range_subpartition [, range_subpartition ...]

range_subpartition:
    SUBPARTITION subpartition_name
    VALUES LESS THAN {(expression_list) | MAXVALUE} [physical_attribute_list]

list_partition_list:
    list_partition [, list_partition] ...

list_partition:
    PARTITION [partition_name]
    VALUES (DEFAULT | expression_list)
    [subpartition_list /*Only applicable to non-template-based subpartitioning*/]
    [ID num] [physical_attribute_list] [compression]

list_subpartition_list:
    list_subpartition [, list_subpartition] ...

list_subpartition:
    SUBPARTITION [partition_name] VALUES (DEFAULT | expression_list) [physical_attribute_list]

expression_list:
    expression [, expression ...]

column_name_list:
    column_name [, column_name ...]

partition_name_list:
    partition_name [, partition_name ...]

partition_count | subpartition_count:
    INT_VALUE

on_commit_option:
    ON COMMIT DELETE ROWS
  | ON COMMIT PRESERVE ROWS

Parameter description

table_mode_value

• NORMAL: Normal. The default value. The probability of triggering a merge after data is dumped is very low.
• QUEUING: The probability of triggering a merge after data is dumped is low.
• MODERATE: Moderate. The probability of triggering a merge after data is dumped is moderate.
• SUPER: Super. The probability of triggering a merge after data is dumped is high.
• EXTREME: Extreme. The probability of triggering a merge after data is dumped is relatively high.

For more information about merges, see Adaptive major compaction.

Examples

• Create a table named tbl1.

  obclient> CREATE TABLE tbl1 (col1 INT PRIMARY KEY, col2 VARCHAR(50)) ;
  Query OK, 0 rows affected
  

• Create a table partitioned by HASH with 8 partitions.

  obclient> CREATE TABLE tbl4 (col1 INT PRIMARY KEY, col2 INT) PARTITION BY HASH(col1) PARTITIONS 8;
  Query OK, 0 rows affected
  

• Create a table partitioned by RANGE and subpartitioned by HASH.

  obclient> CREATE TABLE tbl5 (col1 INT, col2 INT, col3 INT)
                   PARTITION BY RANGE(col1)
                   SUBPARTITION BY HASH(col2)  SUBPARTITIONS  5
                   (PARTITION p0 VALUES LESS THAN(0),
                    PARTITION p1 VALUES LESS THAN(100));
  Query OK, 0 rows affected
  

• Enable encoding, use zstd as the compression algorithm, and set the macroblock reservation percentage to 5%.

  obclient> CREATE TABLE tbl6 (col1 INT, col2 INT, col3 VARCHAR(64))
                   COMPRESS FOR ARCHIVE
                   PCTFREE 5;
  Query OK, 0 rows affected
  

• Create a transaction-level temporary table.

  obclient> CREATE GLOBAL TEMPORARY TABLE tbl7(col1 INT) ON COMMIT DELETE ROWS;
  Query OK, 0 rows affected
  

• Create a table with constraints.

  obclient> CREATE TABLE tbl8 (col1 INT, col2 INT, col3 INT,CONSTRAINT equal_check1 CHECK(col2 = col3 * 2) ENABLE VALIDATE);
  Query OK, 0 rows affected
  

• Add a foreign key to the ref_t2 table. When a DELETE operation affects a matching row in a subtable, the value of the primary key in the parent table is set to NULL.

  obclient> CREATE TABLE ref_t1(c1 INT PRIMARY KEY,C2 INT);
  Query OK, 0 rows affected
  
  obclient> CREATE TABLE ref_t2(c1 INT PRIMARY KEY,C2 INT,FOREIGN KEY(c2) REFERENCES ref_t1(c1) ON DELETE SET NULL);
  Query OK, 0 rows affected
  

• Create a table partitioned by RANGE and subpartitioned by RANGE.

  obclient> CREATE TABLE tbl9 (col1 INT, col2 INT, col3 INT)
                   PARTITION BY RANGE(col1)
                   SUBPARTITION BY RANGE(col2)
  (
    PARTITION p0 VALUES LESS THAN(100)
    (
      SUBPARTITION p0_r1 VALUES LESS THAN(2019),
      SUBPARTITION p0_r2 VALUES LESS THAN(2020),
      SUBPARTITION p0_r3 VALUES LESS THAN(2021)
    ),
    PARTITION p1 VALUES LESS THAN(200)
    (
      SUBPARTITION p1_r1 VALUES LESS THAN(2019),
      SUBPARTITION p1_r2 VALUES LESS THAN(2020),
      SUBPARTITION p1_r3 VALUES LESS THAN(2021)
    ),
    PARTITION p2 VALUES LESS THAN(300)
    (
      SUBPARTITION p2_r1 VALUES LESS THAN(2019),
      SUBPARTITION p2_r2 VALUES LESS THAN(2020),
      SUBPARTITION p2_r3 VALUES LESS THAN(2021)
    )
  );
  Query OK, 0 rows affected
  

• Create the tbl10 table with a DOP of 3.

  obclient> CREATE TABLE tbl10(col1 INT PRIMARY KEY, col2 INT) PARALLEL 3;
  Query OK, 0 rows affected
  

• Set the default value of a column based on a function.

  obclient> CREATE SEQUENCE SEQ_PERSONIPTVSEQ START WITH 1 MINVALUE 1 MAXVALUE 10 INCREMENT BY 2 NOCYCLE NOORDER CACHE 30;
  Query OK, 0 rows affected
  
  obclient> SELECT LPAD(SEQ_PERSONIPTVSEQ.NEXTVAL,18,TO_CHAR(SYSDATE,'YYYY-MM-DD HH24:MI:SS')) FROM DUAL;
  +----------------------------------------------------------------------------+
  | LPAD(SEQ_PERSONIPTVSEQ.NEXTVAL,18,TO_CHAR(SYSDATE,'YYYY-MM-DDHH24:MI:SS')) |
  +----------------------------------------------------------------------------+
  | 2022-07-05 11:34:1                                                         |
  +----------------------------------------------------------------------------+
  1 row in set
  
  obclient> CREATE TABLE FUNC_DEFAULT_TEST ( OID NUMBER(20,0) DEFAULT LPAD(SEQ_PERSONIPTVSEQ.NEXTVAL,18,TO_CHAR(SYSDATE,'YYYY-MM-DD HH24:MI:SS')) NOT NULL);
  Query OK, 0 rows affected
  

• Create a replica-level table named dup_t1 and read from and write to the replica-level table as you would with a normal table. For a read request, if the Proxy is used, the read request can be routed to any OBServer node; if direct connection is used, the read request is processed on the OBServer node where the local replica is located.

  obclient> CREATE TABLE dup_t1(c1 int) DUPLICATE_SCOPE = 'cluster';
  Query OK, 0 rows affected
  
  obclient> INSERT INTO dup_t1 VALUES(1);
  Query OK, 1 row affected
  
  obclient> SELECT * FROM dup_t1;
  +------+
  | c1   |
  +------+
  |    1 |
  +------+
  1 row in set
  

• Create a table named test_tbl1, set the col1 column to an auto-increment column, and set the col1 column as the primary key.

  CREATE TABLE test_tbl1 (
    col1 INT GENERATED BY DEFAULT AS IDENTITY,
    col2 VARCHAR2(50),
    PRIMARY KEY (col1)
  );  
  

Limitations on global temporary tables in Oracle compatibility mode

• Temporary tables are used in various business scenarios in Oracle compatibility mode and have basic accuracy and functionality.
• Generally, temporary tables are used for compatibility and to reduce business modifications. If the business scenarios are limited and the performance requirements for temporary tables are not high, you can use temporary tables. It is better to convert the business scenarios to use regular tables.

Performance and stability

• The SQL execution efficiency of temporary tables is similar to that of normal tables, without particular advantages.
• The system needs to perform additional work, such as data cleanup, when a transaction ends or a session disconnects. This causes additional overheads.
• The data check and cleanup performed on login may put pressure on the login thread, which can increase the login time. In severe cases, it may even prevent login.

Create a temporary table

By default, the system will rewrite the table creation statement:

• Add the SYS_SESSION_ID column as the primary key.
• Add the SYS_SESS_CREATE_TIME column as a normal column.
• Create a hash-partitioned table with 16 partitions based on the SYS_SESSION_ID column as the partitioning key.

For example:

CREATE GLOBAL TEMPORARY TABLE t1(
  c1 INT,
  c2 INT,
  PRIMARY KEY(c1)
);

will be rewritten into the following statement.

CREATE GLOBAL TEMPORARY TABLE t1(
  SYS_SESSION_ID INT,
  SYS_SESS_CREATE_TIME INT,
  c1 INT,
  c2 INT,
  PRIMARY KEY(SYS_SESSION_ID, c1)
)
PARTITION BY HASH(SYS_SESSION_ID) PARTITIONS 16;

DML/Query statements for temporary tables

When you execute the INSERT statement, the system will default insert the session ID and session creation time of the current session into the SYS_SESSION_ID and SYS_SESS_CREATE_TIME columns.

When you execute the UPDATE, DELETE, or SELECT statement, the system will add the filter condition "SYS_SESSION_ID = session ID of the current session" to the SQL statement through SQL rewriting. This condition allows the SQL optimizer to perform partition pruning and extract the query range.

Data cleanup for temporary tables

• For temporary tables with the ON COMMIT DELETE ROWS option (transaction temporary tables, which is also the default option), when a transaction ends, the system will start a new transaction and execute the DELETE statement to delete the data in the temporary table.
• For temporary tables with the ON COMMIT PRESERVE ROWS option (session temporary tables), when a session disconnects, the system will execute the DELETE statement to delete the data in the temporary table.
• Since session IDs can be reused, in OceanBase Database V3.2.4 BP4 and earlier, when you log in, the system will check the data of the current session ID. If data that needs additional cleanup is found, the system will perform additional cleanup.
• Data check and cleanup based on non-unique session IDs may cause login failures (cluster unavailable).

Routing of temporary tables

• Transaction temporary tables (ON COMMIT DELETE ROWS) Only the node where the transaction is initiated can access the temporary table during the transaction.

• Session temporary tables (ON COMMIT PRESERVE ROWS) After a session accesses a temporary table, the OBServer node will notify the Proxy, which will restrict subsequent requests from the Proxy to the current session.

Drop a temporary table

You can successfully execute the DROP statement while executing an DML statement, to delete a temporary table and all its data. This behavior differs from that in Oracle, where you must wait until all sessions no longer hold resources of the temporary table before you can execute the DROP statement.

Cross-version support

Workarounds for severe issues

Unable to log in

1. Stop the business related to temporary tables, delete temporary tables, or perform a major compaction. Generally, the issue can be resolved. If not, proceed to step 2.
2. Restart the server where the login fails.

PL cache bloat due to plan reusability

For example, a procedure definition statement that contains a temporary table:

CREATE PROCEDUCE p1 () (
  INSERT INTO temp_table VALUES (xxx);
  SELECT * FROM temp_table WHERE ROWNUM = 1 INTO var1;
);

Different sessions cannot share a plan for accessing the temporary table. Therefore, each session needs to compile the procedure p1 to generate a cache for it. This may lead to stability issues. You can rewrite the SQL statement as follows:

EXCUTE IMMEDIATE 'SELECT * FROM temp_table WHERE ROWNUM = 1 INTO var1;'

You can use dynamic SQL to bypass this issue.

Data not cleared after a failure

A failure may leave residual data. At present, there is no automatic cleanup method. Generally, this does not affect the use of temporary tables. If a large amount of residual data is found, you can drop the temporary table and rebuild it.