You are viewing an old version of this article. View the current version here.


CREATE [OR REPLACE] [TEMPORARY] TABLE [IF NOT EXISTS] tbl_name (create_definition,...) [table_options ]... [partition_options] CREATE [OR REPLACE] [TEMPORARY] TABLE [IF NOT EXISTS] tbl_name [(create_definition,...)] [table_options ]... [partition_options] select_statement CREATE [OR REPLACE] [TEMPORARY] TABLE [IF NOT EXISTS] tbl_name { LIKE old_table_name | (LIKE old_table_name) }

select_statement: [IGNORE | REPLACE] [AS] SELECT ... (Some legal select statement)


Use the CREATE TABLE statement to create a table with the given name.

In its most basic form, the CREATE TABLE statement provides a table name followed by a list of columns, indexes, and constraints. By default, the table is created in the default database. Specify a database with db_name.tbl_name. If you quote the table name, you must quote the database name and table name separately as `db_name`.`tbl_name`. This is particularly useful for CREATE TABLE ... SELECT, because it allows to create a table into a database, which contains data from other databases. See Identifier Qualifiers.

If a table with the same name exists, error 1050 results. Use IF NOT EXISTS to suppress this error and issue a note instead. Use SHOW WARNINGS to see notes.

The CREATE TABLE statement automatically commits the current transaction, except when using the TEMPORARY keyword.

For valid identifiers to use as table names, see Identifier Names.

Note: if the default_storage_engine is set to ColumnStore then it needs setting on all UMs. Otherwise when the tables using the default engine are replicated across UMs they will use the wrong engine. You should therefore not use this option as a session variable with ColumnStore.

Microsecond precision can be between 0-6. If no precision is specified it is assumed to be 0, for backward compatibility reasons.


Executing the CREATE TABLE statement requires the CREATE privilege for the table or the database.


If the OR REPLACE clause is used and the table already exists, then instead of returning an error, the server will drop the existing table and replace it with the newly defined table.

This syntax was originally added to make replication more robust if it has to rollback and repeat statements such as CREATE ... SELECT on replicas.

CREATE OR REPLACE TABLE table_name (a int);

is basically the same as:

CREATE TABLE table_name (a int);

with the following exceptions:

  • If table_name was locked with LOCK TABLES it will continue to be locked after the statement.
  • Temporary tables are only dropped if the TEMPORARY keyword was used. (With DROP TABLE, temporary tables are preferred to be dropped before normal tables).

Things to be Aware of With CREATE OR REPLACE

  • The table is dropped first (if it existed), after that the CREATE is done. Because of this, if the CREATE fails, then the table will not exist anymore after the statement. If the table was used with LOCK TABLES it will be unlocked.
  • One can't use OR REPLACE together with IF EXISTS.
  • Slaves in replication will by default use CREATE OR REPLACE when replicating CREATE statements that don''t use IF EXISTS. This can be changed by setting the variable slave-ddl-exec-mode to STRICT.


If the IF NOT EXISTS clause is used, then the table will only be created if a table with the same name does not already exist. If the table already exists, then a warning will be triggered by default.


Use the TEMPORARY keyword to create a temporary table that is only available to the current session. Temporary tables are dropped when the session ends. Temporary table names are specific to the session. They will not conflict with other temporary tables from other sessions even if they share the same name. They will shadow names of non-temporary tables or views, if they are identical. A temporary table can have the same name as a non-temporary table which is located in the same database. In that case, their name will reference the temporary table when used in SQL statements. You must have the CREATE TEMPORARY TABLES privilege on the database to create temporary tables. If no storage engine is specified, the default_tmp_storage_engine setting will determine the engine.

ROCKSDB temporary tables cannot be created by setting the default_tmp_storage_engine system variable, or using CREATE TEMPORARY TABLE LIKE. Before MariaDB 10.7, they could be specified, but would silently fail, and a MyISAM table would be created instead. From MariaDB 10.7 an error is returned. Explicitly creating a temporary table with ENGINE=ROCKSDB has never been permitted.


Use the LIKE clause instead of a full table definition to create an empty table with the same definition as another table, including columns, indexes, and table options. Foreign key definitions, as well as any DATA DIRECTORY or INDEX DIRECTORY table options specified on the original table, will not be created.

LIKE does not preserve the TEMPORARY status of the original table. To make the new table TEMPORARY as well, use CREATE TEMPORARY TABLE ... LIKE.

LIKE does not work with views, only base tables. Attempting to use it on a view will result in an error:

CREATE VIEW v (mycol) AS SELECT 'abc';

ERROR 1347 (HY000): 'test.v' is not of type 'BASE TABLE'

The same version of the table storage format as found in the original table is used for the new table.

CREATE TABLE ... LIKE performs the same checks as CREATE TABLE. So a statement may fail if a change in the SQL_MODE renders it invalid. For example:



ERROR 1067 (42000): Invalid default value for 'd'


You can create a table containing data from other tables using the CREATE ... SELECT statement. Columns will be created in the table for each field returned by the SELECT query.

You can also define some columns normally and add other columns from a SELECT. You can also create columns in the normal way and assign them some values using the query, this is done to force a certain type or other field characteristics. The columns that are not named in the query will be placed before the others. For example:

    SELECT 5 AS b, c, d FROM another_table;

Remember that the query just returns data. If you want to use the same indexes, or the same columns attributes ([NOT] NULL, DEFAULT, AUTO_INCREMENT) in the new table, you need to specify them manually. Types and sizes are not automatically preserved if no data returned by the SELECT requires the full size, and VARCHAR could be converted into CHAR. The CAST() function can be used to forcee the new table to use certain types.

Aliases (AS) are taken into account, and they should always be used when you SELECT an expression (function, arithmetical operation, etc).

If an error occurs during the query, the table will not be created at all.

If the new table has a primary key or UNIQUE indexes, you can use the IGNORE or REPLACE keywords to handle duplicate key errors during the query. IGNORE means that the newer values must not be inserted an identical value exists in the index. REPLACE means that older values must be overwritten.

If the columns in the new table are more than the rows returned by the query, the columns populated by the query will be placed after other columns. Note that if the strict SQL_MODE is on, and the columns that are not names in the query do not have a DEFAULT value, an error will raise and no rows will be copied.

Concurrent inserts are not used during the execution of a CREATE ... SELECT.

If the table already exists, an error similar to the following will be returned:

ERROR 1050 (42S01): Table 't' already exists

If the IF NOT EXISTS clause is used and the table exists, a note will be produced instead of an error.

To insert rows from a query into an existing table, INSERT ... SELECT can be used.

Column Definitions

create_definition: { col_name column_definition | index_definition | period_definition | CHECK (expr) }

column_definition: data_type [NOT NULL | NULL] [DEFAULT default_value | (expression)] [ON UPDATE [NOW | CURRENT_TIMESTAMP] [(precision)]] [AUTO_INCREMENT] [ZEROFILL] [UNIQUE [KEY] | [PRIMARY] KEY] [INVISIBLE] [{WITH|WITHOUT} SYSTEM VERSIONING] [COMMENT 'string'] [REF_SYSTEM_ID = value] [reference_definition] | data_type [GENERATED ALWAYS] AS [ ROW {START|END} [NOT NULL ENABLE] [[PRIMARY] KEY] | (expression) [VIRTUAL | PERSISTENT | STORED] ] [INVISIBLE] [UNIQUE [KEY]] [COMMENT 'string']

constraint_definition: CONSTRAINT [constraint_name] CHECK (expression)

Note: Until MariaDB 10.4, MariaDB accepts the shortcut format with a REFERENCES clause only in ALTER TABLE and CREATE TABLE statements, but that syntax does nothing. For example:

CREATE TABLE b(for_key INT REFERENCES a(not_key));

MariaDB simply parses it without returning any error or warning, for compatibility with other DBMS's. Before MariaDB 10.2.1 this was also true for CHECK constraints. However, only the syntax described below creates foreign keys.

From MariaDB 10.5, MariaDB will attempt to apply the constraint. See Foreign Keys examples.

Each definition either creates a column in the table or specifies and index or constraint on one or more columns. See Indexes below for details on creating indexes.

Create a column by specifying a column name and a data type, optionally followed by column options. See Data Types for a full list of data types allowed in MariaDB.


Use the NULL or NOT NULL options to specify that values in the column may or may not be NULL, respectively. By default, values may be NULL. See also NULL Values in MariaDB.

DEFAULT Column Option

Specify a default value using the DEFAULT clause. If you don't specify DEFAULT then the following rules apply:

  • If the column is not defined with NOT NULL, AUTO_INCREMENT or TIMESTAMP, an explicit DEFAULT NULL will be added. Note that in MySQL and in MariaDB before 10.1.6, you may get an explicit DEFAULT for primary key parts, if not specified with NOT NULL.

The default value will be used if you INSERT a row without specifying a value for that column, or if you specify DEFAULT for that column. Before MariaDB 10.2.1 you couldn't usually provide an expression or function to evaluate at insertion time. You had to provide a constant default value instead. The one exception is that you may use CURRENT_TIMESTAMP as the default value for a TIMESTAMP column to use the current timestamp at insertion time.

CURRENT_TIMESTAMP may also be used as the default value for a DATETIME

You can use most functions in DEFAULT. Expressions should have parentheses around them. If you use a non deterministic function in DEFAULT then all inserts to the table will be replicated in row mode. You can even refer to earlier columns in the DEFAULT expression (excluding AUTO_INCREMENT columns):

CREATE TABLE t1 (a int DEFAULT (1+1), b int DEFAULT (a+1));
CREATE TABLE t2 (a bigint primary key DEFAULT UUID_SHORT());

The DEFAULT clause cannot contain any stored functions or subqueries, and a column used in the clause must already have been defined earlier in the statement.

It is possible to assign BLOB or TEXT columns a DEFAULT value. In versions prior to MariaDB 10.2.1, assigning a default to these columns was not possible.

You can also use DEFAULT (NEXT VALUE FOR sequence)


Use AUTO_INCREMENT to create a column whose value can can be set automatically from a simple counter. You can only use AUTO_INCREMENT on a column with an integer type. The column must be a key, and there can only be one AUTO_INCREMENT column in a table. If you insert a row without specifying a value for that column (or if you specify 0, NULL, or DEFAULT as the value), the actual value will be taken from the counter, with each insertion incrementing the counter by one. You can still insert a value explicitly. If you insert a value that is greater than the current counter value, the counter is set based on the new value. An AUTO_INCREMENT column is implicitly NOT NULL. Use LAST_INSERT_ID to get the AUTO_INCREMENT value most recently used by an INSERT statement.

ZEROFILL Column Option

If the ZEROFILL column option is specified for a column using a numeric data type, then the column will be set to UNSIGNED and the spaces used by default to pad the field are replaced with zeros. ZEROFILL is ignored in expressions or as part of a UNION. ZEROFILL is a non-standard MySQL and MariaDB enhancement.

PRIMARY KEY Column Option

Use PRIMARY KEY to make a column a primary key. A primary key is a special type of a unique key. There can be at most one primary key per table, and it is implicitly NOT NULL.

Specifying a column as a unique key creates a unique index on that column. See the Index Definitions section below for more information.

UNIQUE KEY Column Option

Use UNIQUE KEY (or just UNIQUE) to specify that all values in the column must be distinct from each other. Unless the column is NOT NULL, there may be multiple rows with NULL in the column.

Specifying a column as a unique key creates a unique index on that column.

See the Index Definitions section below for more information.

COMMENT Column Option

You can provide a comment for each column using the COMMENT clause. The maximum length is 1024 characters. Use the SHOW FULL COLUMNS statement to see column comments.


REF_SYSTEM_ID can be used to specify Spatial Reference System IDs for spatial data type columns. For example:


Generated Columns

A generated column is a column in a table that cannot explicitly be set to a specific value in a DML query. Instead, its value is automatically generated based on an expression. This expression might generate the value based on the values of other columns in the table, or it might generate the value by calling built-in functions or user-defined functions (UDFs).

There are two types of generated columns:

  • PERSISTENT or STORED: This type's value is actually stored in the table.
  • VIRTUAL: This type's value is not stored at all. Instead, the value is generated dynamically when the table is queried. This type is the default.

Generated columns are also sometimes called computed columns or virtual columns.

For a complete description about generated columns and their limitations, see Generated (Virtual and Persistent/Stored) Columns.


Certain columns may be compressed. See Storage-Engine Independent Column Compression.


Columns may be made invisible, and hidden in certain contexts. See Invisible Columns.


Columns may be explicitly marked as included from system versioning. See System-versioned tables for details.


Columns may be explicitly marked as excluded from system versioning. See System-versioned tables for details.

Index Definitions

    {INDEX|KEY} [index_name] [index_type] (index_col_name,...) [index_option] ...
  {{{|}}} {FULLTEXT|SPATIAL} [INDEX|KEY] [index_name] (index_col_name,...) [index_option] ...
  {{{|}}} [CONSTRAINT [symbol]] PRIMARY KEY [index_type] (index_col_name,...) [index_option] ...
  {{{|}}} [CONSTRAINT [symbol]] UNIQUE [INDEX|KEY] [index_name] [index_type] (index_col_name,...) [index_option] ...
  {{{|}}} [CONSTRAINT [symbol]] FOREIGN KEY [index_name] (index_col_name,...) reference_definition

    col_name [(length)] [ASC | DESC]


    [ KEY_BLOCK_SIZE [=] value
  {{{|}}} index_type
  {{{|}}} WITH PARSER parser_name
  {{{|}}} COMMENT 'string'
  {{{|}}} CLUSTERING={YES| NO} ]

    REFERENCES tbl_name (index_col_name,...)
      [ON DELETE reference_option]
      [ON UPDATE reference_option]


INDEX and KEY are synonyms.

Index names are optional, if not specified an automatic name will be assigned. Index name are needed to drop indexes and appear in error messages when a constraint is violated.

For limits on InnoDB indexes, see InnoDB Limitations.

Index Categories

Plain Indexes

Plain indexes are regular indexes that are not unique, and are not acting as a primary key or a foreign key. They are also not the "specialized" FULLTEXT or SPATIAL indexes.

See Getting Started with Indexes: Plain Indexes for more information.


For PRIMARY KEY indexes, you can specify a name for the index, but it is ignored, and the name of the index is always PRIMARY. From MariaDB 10.3.18 and MariaDB 10.4.8, a warning is explicitly issued if a name is specified. Before then, the name was silently ignored.

See Getting Started with Indexes: Primary Key for more information.


The UNIQUE keyword means that the index will not accept duplicated values, except for NULLs. An error will raise if you try to insert duplicate values in a UNIQUE index.

For UNIQUE indexes, you can specify a name for the constraint, using the CONSTRAINT keyword. That name will be used in error messages.

MariaDB starting with 10.5

Unique, if index type is not specified, is normally a BTREE index that can also be used by the optimizer to find rows. If the key is longer than the max key length for the used storage engine, a HASH key will be created. This enables MariaDB to enforce uniqueness for any type or number of columns.

See Getting Started with Indexes: Unique Index for more information.


For FOREIGN KEY indexes, a reference definition must be provided.

For FOREIGN KEY indexes, you can specify a name for the constraint, using the CONSTRAINT keyword. That name will be used in error messages.

First, you have to specify the name of the target (parent) table and a column or a column list which must be indexed and whose values must match to the foreign key's values. The MATCH clause is accepted to improve the compatibility with other DBMS's, but has no meaning in MariaDB. The ON DELETE and ON UPDATE clauses specify what must be done when a DELETE (or a REPLACE) statements attempts to delete a referenced row from the parent table, and when an UPDATE statement attempts to modify the referenced foreign key columns in a parent table row, respectively. The following options are allowed:

  • RESTRICT: The delete/update operation is not performed. The statement terminates with a 1451 error (SQLSTATE '2300').
  • NO ACTION: Synonym for RESTRICT.
  • CASCADE: The delete/update operation is performed in both tables.
  • SET NULL: The update or delete goes ahead in the parent table, and the corresponding foreign key fields in the child table are set to NULL. (They must not be defined as NOT NULL for this to succeed).
  • SET DEFAULT: This option is currently implemented only for the PBXT storage engine, which is disabled by default and no longer maintained. It sets the child table's foreign key fields to their DEFAULT values when the referenced parent table key entries are updated or deleted.

If either clause is omitted, the default behavior for the omitted clause is RESTRICT.

See Foreign Keys for more information.


Use the FULLTEXT keyword to create full-text indexes.

See Full-Text Indexes for more information.


Use the SPATIAL keyword to create geometric indexes.

See SPATIAL INDEX for more information.

Index Options


The KEY_BLOCK_SIZE index option is similar to the KEY_BLOCK_SIZE table option.

With the InnoDB storage engine, if you specify a non-zero value for the KEY_BLOCK_SIZE table option for the whole table, then the table will implicitly be created with the ROW_FORMAT table option set to COMPRESSED. However, this does not happen if you just set the KEY_BLOCK_SIZE index option for one or more indexes in the table. The InnoDB storage engine ignores the KEY_BLOCK_SIZE index option. However, the SHOW CREATE TABLE statement may still report it for the index.

For information about the KEY_BLOCK_SIZE index option, see the KEY_BLOCK_SIZE table option below.

Index Types

Each storage engine supports some or all index types. See Storage Engine Index Types for details on permitted index types for each storage engine.

Different index types are optimized for different kind of operations:

  • BTREE is the default type, and normally is the best choice. It is supported by all storage engines. It can be used to compare a column's value with a value using the =, >, >=, <, <=, BETWEEN, and LIKE operators. BTREE can also be used to find NULL values. Searches against an index prefix are possible.
  • HASH is only supported by the MEMORY storage engine. HASH indexes can only be used for =, <=, and >= comparisons. It can not be used for the ORDER BY clause. Searches against an index prefix are not possible.
  • RTREE is the default for SPATIAL indexes, but if the storage engine does not support it BTREE can be used.

Index columns names are listed between parenthesis. After each column, a prefix length can be specified. If no length is specified, the whole column will be indexed. ASC and DESC can be specified. Prior to MariaDB 10.8, this was only for compatibility with other DBMSs, but had no meaning in MariaDB. From MariaDB 10.8, individual columns in the index can now be explicitly sorted in ascending or descending order. This can be useful for optimizing certain ORDER BY cases (MDEV-13756, MDEV-26938, MDEV-26939, MDEV-26996). From MariaDB 11.4.0, not only ascending, but also descending, indexes can now be used to optimize MIN() and MAX() (MDEV-27576).

WITH PARSER Index Option

The WITH PARSER index option only applies to FULLTEXT indexes and contains the fulltext parser name. The fulltext parser must be an installed plugin.

COMMENT Index Option

A comment of up to 1024 characters is permitted with the COMMENT index option.

The COMMENT index option allows you to specify a comment with user-readable text describing what the index is for. This information is not used by the server itself.


The CLUSTERING index option is only valid for tables using the TokuDB storage engine.


MariaDB starting with 10.6.0

From MariaDB 10.6.0, indexes can be specified to be ignored by the optimizer. See Ignored Indexes.


period_definition: PERIOD FOR [time_period_name | SYSTEM_TIME] (start_column_name, end_column_name)

MariaDB supports System-versioned tables, Application-time-period tables or Bitemporal Tables.

Constraint Expressions

Note: Before MariaDB 10.2.1, constraint expressions were accepted in the syntax but ignored.

MariaDB 10.2.1 introduced two ways to define a constraint:

  • CHECK(expression) given as part of a column definition.
  • CONSTRAINT [constraint_name] CHECK (expression)

Before a row is inserted or updated, all constraints are evaluated in the order they are defined. If any constraints fails, then the row will not be updated. One can use most deterministic functions in a constraint, including UDFs.

create table t1 (a int check(a>0) ,b int check (b> 0), constraint abc check (a>b));

If you use the second format and you don't give a name to the constraint, then the constraint will get a auto generated name. This is done so that you can later delete the constraint with ALTER TABLE DROP constraint_name.

One can disable all constraint expression checks by setting the variable check_constraint_checks to OFF. This is useful for example when loading a table that violates some constraints that you want to later find and fix in SQL.

See CONSTRAINT for more information.

Table Options

For each individual table you create (or alter), you can set some table options. The general syntax for setting options is:

<OPTION_NAME> = <option_value>, [<OPTION_NAME> = <option_value> ...]

The equal sign is optional.

Some options are supported by the server and can be used for all tables, no matter what storage engine they use; other options can be specified for all storage engines, but have a meaning only for some engines. Also, engines can extend CREATE TABLE with new options.

If the IGNORE_BAD_TABLE_OPTIONS SQL_MODE is enabled, wrong table options generate a warning; otherwise, they generate an error.

table_option: [STORAGE] ENGINE [=] engine_name | AUTO_INCREMENT [=] number | AVG_ROW_LENGTH [=] number | [DEFAULT] CHARACTER SET [=] charset_name | CHECKSUM [=] {0 | 1} | [DEFAULT] COLLATE [=] collation_name | COMMENT [=] 'string' | CONNECTION [=] 'connect_string' | DATA DIRECTORY [=] 'absolute path to directory' | DELAY_KEY_WRITE [=] {0 | 1} | ENCRYPTED [=] {YES | NO} | ENCRYPTION_KEY_ID [=] number | IETF_QUOTES [=] {YES | NO} | INDEX DIRECTORY [=] 'absolute path to directory' | INSERT_METHOD [=] { NO | FIRST | LAST } | KEY_BLOCK_SIZE [=] number | MAX_ROWS [=] number | MIN_ROWS [=] number | PACK_KEYS [=] {0 | 1 | DEFAULT} | PAGE_CHECKSUM [=] {0 | 1} | PAGE_COMPRESSED [=] {0 | 1} | PAGE_COMPRESSION_LEVEL [=] {0 .. 9} | PASSWORD [=] 'string' | ROW_FORMAT [=] {DEFAULT|DYNAMIC|FIXED|COMPRESSED|REDUNDANT|COMPACT|PAGE} | SEQUENCE [=] {0|1} | STATS_AUTO_RECALC [=] {DEFAULT|0|1} | STATS_PERSISTENT [=] {DEFAULT|0|1} | STATS_SAMPLE_PAGES [=] {DEFAULT|number} | TABLESPACE tablespace_name | TRANSACTIONAL [=] {0 | 1} | UNION [=] (tbl_name[,tbl_name]...) | WITH SYSTEM VERSIONING


[STORAGE] ENGINE specifies a storage engine for the table. If this option is not used, the default storage engine is used instead. That is, the default_storage_engine session option value if it is set, or the value specified for the --default-storage-engine mariadbd startup option, or the default storage engine, InnoDB. If the specified storage engine is not installed and active, the default value will be used, unless the NO_ENGINE_SUBSTITUTION SQL MODE is set (default). This is only true for CREATE TABLE, not for ALTER TABLE. For a list of storage engines that are present in your server, issue a SHOW ENGINES.


AUTO_INCREMENT specifies the initial value for the AUTO_INCREMENT primary key. This works for MyISAM, Aria, InnoDB, MEMORY, and ARCHIVE tables. You can change this option with ALTER TABLE, but in that case the new value must be higher than the highest value which is present in the AUTO_INCREMENT column. If the storage engine does not support this option, you can insert (and then delete) a row having the wanted value - 1 in the AUTO_INCREMENT column.


AVG_ROW_LENGTH is the average rows size. It only applies to tables using MyISAM and Aria storage engines that have the ROW_FORMAT table option set to FIXED format.

MyISAM uses MAX_ROWS and AVG_ROW_LENGTH to decide the maximum size of a table (default: 256TB, or the maximum file size allowed by the system).


[DEFAULT] CHARACTER SET (or [DEFAULT] CHARSET) is used to set a default character set for the table. This is the character set used for all columns where an explicit character set is not specified. If this option is omitted or DEFAULT is specified, database's default character set will be used. See Setting Character Sets and Collations for details on setting the character sets.


CHECKSUM (or TABLE_CHECKSUM) can be set to 1 to maintain a live checksum for all table's rows. This makes write operations slower, but CHECKSUM TABLE will be very fast. This option is only supported for MyISAM and Aria tables.


[DEFAULT] COLLATE is used to set a default collation for the table. This is the collation used for all columns where an explicit character set is not specified. If this option is omitted or DEFAULT is specified, database's default option will be used. See Setting Character Sets and Collations for details on setting the collations


COMMENT is a comment for the table. The maximum length is 2048 characters. Also used to define table parameters when creating a Spider table.


CONNECTION is used to specify a server name or a connection string for a Spider, CONNECT, Federated or FederatedX table.


DATA DIRECTORY and INDEX DIRECTORY are supported for MyISAM and Aria, and DATA DIRECTORY is also supported by InnoDB if the innodb_file_per_table server system variable is enabled, but only in CREATE TABLE, not in ALTER TABLE. So, carefully choose a path for InnoDB tables at creation time, because it cannot be changed without dropping and re-creating the table. These options specify the paths for data files and index files, respectively. If these options are omitted, the database's directory will be used to store data files and index files. Note that these table options do not work for partitioned tables (use the partition options instead), or if the server has been invoked with the --skip-symbolic-links startup option. To avoid the overwriting of old files with the same name that could be present in the directories, you can use the --keep_files_on_create option (an error will be issued if files already exist). These options are ignored if the NO_DIR_IN_CREATE SQL_MODE is enabled (useful for replication slaves). Also note that symbolic links cannot be used for InnoDB tables.

DATA DIRECTORY works by creating symlinks from where the table would normally have been (inside the datadir) to where the option specifies. For security reasons, to avoid bypassing the privilege system, the server does not permit symlinks inside the datadir. Therefore, DATA DIRECTORY cannot be used to specify a location inside the datadir. An attempt to do so will result in an error 1210 (HY000) Incorrect arguments to DATA DIRECTORY.


DELAY_KEY_WRITE is supported by MyISAM and Aria, and can be set to 1 to speed up write operations. In that case, when data are modified, the indexes are not updated until the table is closed. Writing the changes to the index file altogether can be much faster. However, note that this option is applied only if the delay_key_write server variable is set to 'ON'. If it is 'OFF' the delayed index writes are always disabled, and if it is 'ALL' the delayed index writes are always used, disregarding the value of DELAY_KEY_WRITE.


The ENCRYPTED table option can be used to manually set the encryption status of an InnoDB table. See InnoDB Encryption for more information.

Aria does not support the ENCRYPTED table option. See MDEV-18049.

See Data-at-Rest Encryption for more information.


The ENCRYPTION_KEY_ID table option can be used to manually set the encryption key of an InnoDB table. See InnoDB Encryption for more information.

Aria does not support the ENCRYPTION_KEY_ID table option. See MDEV-18049.

See Data-at-Rest Encryption for more information.


For the CSV storage engine, the IETF_QUOTES option, when set to YES, enables IETF-compatible parsing of embedded quote and comma characters. Enabling this option for a table improves compatibility with other tools that use CSV, but is not compatible with MySQL CSV tables, or MariaDB CSV tables created without this option. Disabled by default.


INSERT_METHOD is only used with MERGE tables. This option determines in which underlying table the new rows should be inserted. If you set it to 'NO' (which is the default) no new rows can be added to the table (but you will still be able to perform INSERTs directly against the underlying tables). FIRST means that the rows are inserted into the first table, and LAST means that thet are inserted into the last table.


KEY_BLOCK_SIZE is used to determine the size of key blocks, in bytes or kilobytes. However, this value is just a hint, and the storage engine could modify or ignore it. If KEY_BLOCK_SIZE is set to 0, the storage engine's default value will be used.

With the InnoDB storage engine, if you specify a non-zero value for the KEY_BLOCK_SIZE table option for the whole table, then the table will implicitly be created with the ROW_FORMAT table option set to COMPRESSED.


MIN_ROWS and MAX_ROWS let the storage engine know how many rows you are planning to store as a minimum and as a maximum. These values will not be used as real limits, but they help the storage engine to optimize the table. MIN_ROWS is only used by MEMORY storage engine to decide the minimum memory that is always allocated. MAX_ROWS is used to decide the minimum size for indexes.


PACK_KEYS can be used to determine whether the indexes will be compressed. Set it to 1 to compress all keys. With a value of 0, compression will not be used. With the DEFAULT value, only long strings will be compressed. Uncompressed keys are faster.


PAGE_CHECKSUM is only applicable to Aria tables, and determines whether indexes and data should use page checksums for extra safety.


PAGE_COMPRESSED is used to enable InnoDB page compression for InnoDB tables.


PAGE_COMPRESSION_LEVEL is used to set the compression level for InnoDB page compression for InnoDB tables. The table must also have the PAGE_COMPRESSED table option set to 1.

Valid values for PAGE_COMPRESSION_LEVEL are 1 (the best speed) through 9 (the best compression), .


PASSWORD is unused.


RAID_TYPE is an obsolete option, as the raid support has been disabled since MySQL 5.0.


The ROW_FORMAT table option specifies the row format for the data file. Possible values are engine-dependent.

Supported MyISAM Row Formats

For MyISAM, the supported row formats are:


The COMPRESSED row format can only be set by the myisampack command line tool.

See MyISAM Storage Formats for more information.

Supported Aria Row Formats

For Aria, the supported row formats are:

  • PAGE

See Aria Storage Formats for more information.

Supported InnoDB Row Formats

For InnoDB, the supported row formats are:


If the ROW_FORMAT table option is set to FIXED for an InnoDB table, then the server will either return an error or a warning depending on the value of the innodb_strict_mode system variable. If the innodb_strict_mode system variable is set to OFF, then a warning is issued, and MariaDB will create the table using the default row format for the specific MariaDB server version. If the innodb_strict_mode system variable is set to ON, then an error will be raised.

See InnoDB Storage Formats for more information.

Other Storage Engines and ROW_FORMAT

Other storage engines do not support the ROW_FORMAT table option.


If the table is a sequence, then it will have the SEQUENCE set to 1.


STATS_AUTO_RECALC indicates whether to automatically recalculate persistent statistics (see STATS_PERSISTENT, below) for an InnoDB table. If set to 1, statistics will be recalculated when more than 10% of the data has changed. When set to 0, stats will be recalculated only when an ANALYZE TABLE is run. If set to DEFAULT, or left out, the value set by the innodb_stats_auto_recalc system variable applies. See InnoDB Persistent Statistics.


STATS_PERSISTENT indicates whether the InnoDB statistics created by ANALYZE TABLE will remain on disk or not. It can be set to 1 (on disk), 0 (not on disk, the pre-MariaDB 10 behavior), or DEFAULT (the same as leaving out the option), in which case the value set by the innodb_stats_persistent system variable will apply. Persistent statistics stored on disk allow the statistics to survive server restarts, and provide better query plan stability. See InnoDB Persistent Statistics.


STATS_SAMPLE_PAGES indicates how many pages are used to sample index statistics. If 0 or DEFAULT, the default value, the innodb_stats_sample_pages value is used. See InnoDB Persistent Statistics.


TRANSACTIONAL is only applicable for Aria tables. In future Aria tables created with this option will be fully transactional, but currently this provides a form of crash protection. See Aria Storage Engine for more details.


UNION must be specified when you create a MERGE table. This option contains a comma-separated list of MyISAM tables which are accessed by the new table. The list is enclosed between parenthesis. Example: UNION = (t1,t2)


WITH SYSTEM VERSIONING is used for creating System-versioned tables.


partition_options: PARTITION BY { [LINEAR] HASH(expr) | [LINEAR] KEY(column_list) | RANGE(expr) | LIST(expr) | SYSTEM_TIME [INTERVAL time_quantity time_unit] [LIMIT num] } [PARTITIONS num] [SUBPARTITION BY { [LINEAR] HASH(expr) | [LINEAR] KEY(column_list) } [SUBPARTITIONS num] ] [(partition_definition [, partition_definition] ...)]

partition_definition: [PARTITION] partition_name [VALUES {LESS THAN {(expr) | MAXVALUE} | IN (value_list)}] [[STORAGE] ENGINE [=] engine_name] [COMMENT [=] 'comment_text' ] [DATA DIRECTORY [=] 'data_dir'] [INDEX DIRECTORY [=] 'index_dir'] [MAX_ROWS [=] max_number_of_rows] [MIN_ROWS [=] min_number_of_rows] [TABLESPACE [=] tablespace_name] [NODEGROUP [=] node_group_id] [(subpartition_definition [, subpartition_definition] ...)]

subpartition_definition: SUBPARTITION logical_name [[STORAGE] ENGINE [=] engine_name] [COMMENT [=] 'comment_text' ] [DATA DIRECTORY [=] 'data_dir'] [INDEX DIRECTORY [=] 'index_dir'] [MAX_ROWS [=] max_number_of_rows] [MIN_ROWS [=] min_number_of_rows] [TABLESPACE [=] tablespace_name] [NODEGROUP [=] node_group_id]

If the PARTITION BY clause is used, the table will be partitioned. A partition method must be explicitly indicated for partitions and subpartitions. Partition methods are:

  • [LINEAR] HASH creates a hash key which will be used to read and write rows. The partition function can be any valid SQL expression which returns an INTEGER number. Thus, it is possible to use the HASH method on an integer column, or on functions which accept integer columns as an argument. However, VALUES LESS THAN and VALUES IN clauses can not be used with HASH. An example:

[LINEAR] HASH can be used for subpartitions, too.

  • [LINEAR] KEY is similar to HASH, but the index has an even distribution of data. Also, the expression can only be a column or a list of columns. VALUES LESS THAN and VALUES IN clauses can not be used with KEY.
  • RANGE partitions the rows using on a range of values, using the VALUES LESS THAN operator. VALUES IN is not allowed with RANGE. The partition function can be any valid SQL expression which returns a single value.
  • LIST assigns partitions based on a table's column with a restricted set of possible values. It is similar to RANGE, but VALUES IN must be used for at least 1 columns, and VALUES LESS THAN is disallowed.
  • SYSTEM_TIME partitioning is used for System-versioned tables to store historical data separately from current data.

Only HASH and KEY can be used for subpartitions, and they can be [LINEAR].

It is possible to define up to 8092 partitions and subpartitions.

The number of defined partitions can be optionally specified as PARTITION count. This can be done to avoid specifying all partitions individually. But you can also declare each individual partition and, additionally, specify a PARTITIONS count clause; in the case, the number of PARTITIONs must equal count.

Also see Partitioning Types Overview.

MariaDB starting with 10.7.1

From MariaDB 10.7, the PARTITION keyword is now optional as part of the partition definition, for example, instead of:

create or replace table t1 (x int)
  partition by range(x) (
    partition p1 values less than (10),
    partition p2 values less than (20),
    partition p3 values less than (30),
    partition p4 values less than (40),
    partition p5 values less than (50),
    partition pn values less than maxvalue);

the following can also be used:

create or replace table t1 (x int)
  partition by range(x) (
    p1 values less than (10),
    p2 values less than (20),
    p3 values less than (30),
    p4 values less than (40),
    p5 values less than (50),
    pn values less than maxvalue);


CREATE TABLE can also be used to create a SEQUENCE. See CREATE SEQUENCE and Sequence Overview.

Atomic DDL

MariaDB starting with 10.6.1

MariaDB 10.6.1 supports Atomic DDL. CREATE TABLE is atomic, except for CREATE OR REPLACE, which is only crash safe.


create table if not exists test (
a bigint auto_increment primary key,
name varchar(128) charset utf8,
key name (name(32))
) engine=InnoDB default charset latin1;

This example shows a couple of things:

  • Usage of IF NOT EXISTS; If the table already existed, it will not be created. There will not be any error for the client, just a warning.
  • How to create a PRIMARY KEY that is automatically generated.
  • How to specify a table-specific character set and another for a column.
  • How to create an index (name) that is only partly indexed (to save space).

The following clauses will work from MariaDB 10.2.1 only.

  a int DEFAULT (1+1),
  b int DEFAULT (a+1),

See Also


Comments loading...
Content reproduced on this site is the property of its respective owners, and this content is not reviewed in advance by MariaDB. The views, information and opinions expressed by this content do not necessarily represent those of MariaDB or any other party.