Documentation Home
MySQL 8.4 Reference Manual
Related Documentation Download this Manual
PDF (US Ltr) - 40.1Mb
PDF (A4) - 40.1Mb
Man Pages (TGZ) - 259.1Kb
Man Pages (Zip) - 366.2Kb
Info (Gzip) - 4.0Mb
Info (Zip) - 4.0Mb


MySQL 8.4 Reference Manual  /  ...  /  CREATE INDEX Statement

15.1.15 CREATE INDEX Statement

CREATE [UNIQUE | FULLTEXT | SPATIAL] INDEX index_name
    [index_type]
    ON tbl_name (key_part,...)
    [index_option]
    [algorithm_option | lock_option] ...

key_part: {col_name [(length)] | (expr)} [ASC | DESC]

index_option: {
    KEY_BLOCK_SIZE [=] value
  | index_type
  | WITH PARSER parser_name
  | COMMENT 'string'
  | {VISIBLE | INVISIBLE}
  | ENGINE_ATTRIBUTE [=] 'string'
  | SECONDARY_ENGINE_ATTRIBUTE [=] 'string'
}

index_type:
    USING {BTREE | HASH}

algorithm_option:
    ALGORITHM [=] {DEFAULT | INPLACE | COPY}

lock_option:
    LOCK [=] {DEFAULT | NONE | SHARED | EXCLUSIVE}

Normally, you create all indexes on a table at the time the table itself is created with CREATE TABLE. See Section 15.1.20, “CREATE TABLE Statement”. This guideline is especially important for InnoDB tables, where the primary key determines the physical layout of rows in the data file. CREATE INDEX enables you to add indexes to existing tables.

CREATE INDEX is mapped to an ALTER TABLE statement to create indexes. See Section 15.1.9, “ALTER TABLE Statement”. CREATE INDEX cannot be used to create a PRIMARY KEY; use ALTER TABLE instead. For more information about indexes, see Section 10.3.1, “How MySQL Uses Indexes”.

InnoDB supports secondary indexes on virtual columns. For more information, see Section 15.1.20.9, “Secondary Indexes and Generated Columns”.

When the innodb_stats_persistent setting is enabled, run the ANALYZE TABLE statement for an InnoDB table after creating an index on that table.

The expr for a key_part specification can also take the form (CAST json_expression AS type ARRAY) to create a multi-valued index on a JSON column. See Multi-Valued Indexes.

An index specification of the form (key_part1, key_part2, ...) creates an index with multiple key parts. Index key values are formed by concatenating the values of the given key parts. For example (col1, col2, col3) specifies a multiple-column index with index keys consisting of values from col1, col2, and col3.

A key_part specification can end with ASC or DESC to specify whether index values are stored in ascending or descending order. The default is ascending if no order specifier is given.

ASC and DESC are not supported for HASH indexes, multi-valued indexes or SPATIAL indexes.

The following sections describe different aspects of the CREATE INDEX statement:

Column Prefix Key Parts

For string columns, indexes can be created that use only the leading part of column values, using col_name(length) syntax to specify an index prefix length:

If a specified index prefix exceeds the maximum column data type size, CREATE INDEX handles the index as follows:

  • For a nonunique index, either an error occurs (if strict SQL mode is enabled), or the index length is reduced to lie within the maximum column data type size and a warning is produced (if strict SQL mode is not enabled).

  • For a unique index, an error occurs regardless of SQL mode because reducing the index length might enable insertion of nonunique entries that do not meet the specified uniqueness requirement.

The statement shown here creates an index using the first 10 characters of the name column (assuming that name has a nonbinary string type):

CREATE INDEX part_of_name ON customer (name(10));

If names in the column usually differ in the first 10 characters, lookups performed using this index should not be much slower than using an index created from the entire name column. Also, using column prefixes for indexes can make the index file much smaller, which could save a lot of disk space and might also speed up INSERT operations.

Functional Key Parts

A normal index indexes column values or prefixes of column values. For example, in the following table, the index entry for a given t1 row includes the full col1 value and a prefix of the col2 value consisting of its first 10 characters:

CREATE TABLE t1 (
  col1 VARCHAR(10),
  col2 VARCHAR(20),
  INDEX (col1, col2(10))
);

Functional key parts that index expression values canalso be used in place of column or column prefix values. Use of functional key parts enables indexing of values not stored directly in the table. Examples:

CREATE TABLE t1 (col1 INT, col2 INT, INDEX func_index ((ABS(col1))));
CREATE INDEX idx1 ON t1 ((col1 + col2));
CREATE INDEX idx2 ON t1 ((col1 + col2), (col1 - col2), col1);
ALTER TABLE t1 ADD INDEX ((col1 * 40) DESC);

An index with multiple key parts can mix nonfunctional and functional key parts.

ASC and DESC are supported for functional key parts.

Functional key parts must adhere to the following rules. An error occurs if a key part definition contains disallowed constructs.

  • In index definitions, enclose expressions within parentheses to distinguish them from columns or column prefixes. For example, this is permitted; the expressions are enclosed within parentheses:

    INDEX ((col1 + col2), (col3 - col4))

    This produces an error; the expressions are not enclosed within parentheses:

    INDEX (col1 + col2, col3 - col4)
  • A functional key part cannot consist solely of a column name. For example, this is not permitted:

    INDEX ((col1), (col2))

    Instead, write the key parts as nonfunctional key parts, without parentheses:

    INDEX (col1, col2)
  • A functional key part expression cannot refer to column prefixes. For a workaround, see the discussion of SUBSTRING() and CAST() later in this section.

  • Functional key parts are not permitted in foreign key specifications.

For CREATE TABLE ... LIKE, the destination table preserves functional key parts from the original table.

Functional indexes are implemented as hidden virtual generated columns, which has these implications:

UNIQUE is supported for indexes that include functional key parts. However, primary keys cannot include functional key parts. A primary key requires the generated column to be stored, but functional key parts are implemented as virtual generated columns, not stored generated columns.

SPATIAL and FULLTEXT indexes cannot have functional key parts.

If a table contains no primary key, InnoDB automatically promotes the first UNIQUE NOT NULL index to the primary key. This is not supported for UNIQUE NOT NULL indexes that have functional key parts.

Nonfunctional indexes raise a warning if there are duplicate indexes. Indexes that contain functional key parts do not have this feature.

To remove a column that is referenced by a functional key part, the index must be removed first. Otherwise, an error occurs.

Although nonfunctional key parts support a prefix length specification, this is not possible for functional key parts. The solution is to use SUBSTRING() (or CAST(), as described later in this section). For a functional key part containing the SUBSTRING() function to be used in a query, the WHERE clause must contain SUBSTRING() with the same arguments. In the following example, only the second SELECT is able to use the index because that is the only query in which the arguments to SUBSTRING() match the index specification:

CREATE TABLE tbl (
  col1 LONGTEXT,
  INDEX idx1 ((SUBSTRING(col1, 1, 10)))
);
SELECT * FROM tbl WHERE SUBSTRING(col1, 1, 9) = '123456789';
SELECT * FROM tbl WHERE SUBSTRING(col1, 1, 10) = '1234567890';

Functional key parts enable indexing of values that cannot be indexed otherwise, such as JSON values. However, this must be done correctly to achieve the desired effect. For example, this syntax does not work:

CREATE TABLE employees (
  data JSON,
  INDEX ((data->>'$.name'))
);

The syntax fails because:

  • The ->> operator translates into JSON_UNQUOTE(JSON_EXTRACT(...)).

  • JSON_UNQUOTE() returns a value with a data type of LONGTEXT, and the hidden generated column thus is assigned the same data type.

  • MySQL cannot index LONGTEXT columns specified without a prefix length on the key part, and prefix lengths are not permitted in functional key parts.

To index the JSON column, you could try using the CAST() function as follows:

CREATE TABLE employees (
  data JSON,
  INDEX ((CAST(data->>'$.name' AS CHAR(30))))
);

The hidden generated column is assigned the VARCHAR(30) data type, which can be indexed. But this approach produces a new issue when trying to use the index:

  • CAST() returns a string with the collation utf8mb4_0900_ai_ci (the server default collation).

  • JSON_UNQUOTE() returns a string with the collation utf8mb4_bin (hard coded).

As a result, there is a collation mismatch between the indexed expression in the preceding table definition and the WHERE clause expression in the following query:

SELECT * FROM employees WHERE data->>'$.name' = 'James';

The index is not used because the expressions in the query and the index differ. To support this kind of scenario for functional key parts, the optimizer automatically strips CAST() when looking for an index to use, but only if the collation of the indexed expression matches that of the query expression. For an index with a functional key part to be used, either of the following two solutions work (although they differ somewhat in effect):

  • Solution 1. Assign the indexed expression the same collation as JSON_UNQUOTE():

    CREATE TABLE employees (
      data JSON,
      INDEX idx ((CAST(data->>"$.name" AS CHAR(30)) COLLATE utf8mb4_bin))
    );
    INSERT INTO employees VALUES
      ('{ "name": "james", "salary": 9000 }'),
      ('{ "name": "James", "salary": 10000 }'),
      ('{ "name": "Mary", "salary": 12000 }'),
      ('{ "name": "Peter", "salary": 8000 }');
    SELECT * FROM employees WHERE data->>'$.name' = 'James';

    The ->> operator is the same as JSON_UNQUOTE(JSON_EXTRACT(...)), and JSON_UNQUOTE() returns a string with collation utf8mb4_bin. The comparison is thus case-sensitive, and only one row matches:

    +------------------------------------+
    | data                               |
    +------------------------------------+
    | {"name": "James", "salary": 10000} |
    +------------------------------------+
  • Solution 2. Specify the full expression in the query:

    CREATE TABLE employees (
      data JSON,
      INDEX idx ((CAST(data->>"$.name" AS CHAR(30))))
    );
    INSERT INTO employees VALUES
      ('{ "name": "james", "salary": 9000 }'),
      ('{ "name": "James", "salary": 10000 }'),
      ('{ "name": "Mary", "salary": 12000 }'),
      ('{ "name": "Peter", "salary": 8000 }');
    SELECT * FROM employees WHERE CAST(data->>'$.name' AS CHAR(30)) = 'James';

    CAST() returns a string with collation utf8mb4_0900_ai_ci, so the comparison case-insensitive and two rows match:

    +------------------------------------+
    | data                               |
    +------------------------------------+
    | {"name": "james", "salary": 9000}  |
    | {"name": "James", "salary": 10000} |
    +------------------------------------+

Be aware that although the optimizer supports automatically stripping CAST() with indexed generated columns, the following approach does not work because it produces a different result with and without an index (Bug#27337092):

mysql> CREATE TABLE employees (
         data JSON,
         generated_col VARCHAR(30) AS (CAST(data->>'$.name' AS CHAR(30)))
       );
Query OK, 0 rows affected, 1 warning (0.03 sec)

mysql> INSERT INTO employees (data)
       VALUES ('{"name": "james"}'), ('{"name": "James"}');
Query OK, 2 rows affected, 1 warning (0.01 sec)
Records: 2  Duplicates: 0  Warnings: 1

mysql> SELECT * FROM employees WHERE data->>'$.name' = 'James';
+-------------------+---------------+
| data              | generated_col |
+-------------------+---------------+
| {"name": "James"} | James         |
+-------------------+---------------+
1 row in set (0.00 sec)

mysql> ALTER TABLE employees ADD INDEX idx (generated_col);
Query OK, 0 rows affected, 1 warning (0.03 sec)
Records: 0  Duplicates: 0  Warnings: 1

mysql> SELECT * FROM employees WHERE data->>'$.name' = 'James';
+-------------------+---------------+
| data              | generated_col |
+-------------------+---------------+
| {"name": "james"} | james         |
| {"name": "James"} | James         |
+-------------------+---------------+
2 rows in set (0.01 sec)

Unique Indexes

A UNIQUE index creates a constraint such that all values in the index must be distinct. An error occurs if you try to add a new row with a key value that matches an existing row. If you specify a prefix value for a column in a UNIQUE index, the column values must be unique within the prefix length. A UNIQUE index permits multiple NULL values for columns that can contain NULL.

If a table has a PRIMARY KEY or UNIQUE NOT NULL index that consists of a single column that has an integer type, you can use _rowid to refer to the indexed column in SELECT statements, as follows:

  • _rowid refers to the PRIMARY KEY column if there is a PRIMARY KEY consisting of a single integer column. If there is a PRIMARY KEY but it does not consist of a single integer column, _rowid cannot be used.

  • Otherwise, _rowid refers to the column in the first UNIQUE NOT NULL index if that index consists of a single integer column. If the first UNIQUE NOT NULL index does not consist of a single integer column, _rowid cannot be used.

Full-Text Indexes

FULLTEXT indexes are supported only for InnoDB and MyISAM tables and can include only CHAR, VARCHAR, and TEXT columns. Indexing always happens over the entire column; column prefix indexing is not supported and any prefix length is ignored if specified. See Section 14.9, “Full-Text Search Functions”, for details of operation.

Multi-Valued Indexes

InnoDB supports multi-valued indexes. A multi-valued index is a secondary index defined on a column that stores an array of values. A normal index has one index record for each data record (1:1). A multi-valued index can have multiple index records for a single data record (N:1). Multi-valued indexes are intended for indexing JSON arrays. For example, a multi-valued index defined on the array of zip codes in the following JSON document creates an index record for each zip code, with each index record referencing the same data record.

{
    "user":"Bob",
    "user_id":31,
    "zipcode":[94477,94536]
}
Creating multi-valued Indexes

You can create a multi-valued index in a CREATE TABLE, ALTER TABLE, or CREATE INDEX statement. This requires using CAST(... AS ... ARRAY) in the index definition, which casts same-typed scalar values in a JSON array to an SQL data type array. A virtual column is then generated transparently with the values in the SQL data type array; finally, a functional index (also referred to as a virtual index) is created on the virtual column. It is the functional index defined on the virtual column of values from the SQL data type array that forms the multi-valued index.

The examples in the following list show the three different ways in which a multi-valued index zips can be created on an array $.zipcode on a JSON column custinfo in a table named customers. In each case, the JSON array is cast to an SQL data type array of UNSIGNED integer values.

  • CREATE TABLE only:

    CREATE TABLE customers (
        id BIGINT NOT NULL AUTO_INCREMENT PRIMARY KEY,
        modified DATETIME DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
        custinfo JSON,
        INDEX zips( (CAST(custinfo->'$.zipcode' AS UNSIGNED ARRAY)) )
        );
  • CREATE TABLE plus ALTER TABLE:

    CREATE TABLE customers (
        id BIGINT NOT NULL AUTO_INCREMENT PRIMARY KEY,
        modified DATETIME DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
        custinfo JSON
        );
    
    ALTER TABLE customers ADD INDEX zips( (CAST(custinfo->'$.zipcode' AS UNSIGNED ARRAY)) );

  • CREATE TABLE plus CREATE INDEX:

    CREATE TABLE customers (
        id BIGINT NOT NULL AUTO_INCREMENT PRIMARY KEY,
        modified DATETIME DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
        custinfo JSON
        );
    
    CREATE INDEX zips ON customers ( (CAST(custinfo->'$.zipcode' AS UNSIGNED ARRAY)) );

A multi-valued index can also be defined as part of a composite index. This example shows a composite index that includes two single-valued parts (for the id and modified columns), and one multi-valued part (for the custinfo column):

CREATE TABLE customers (
    id BIGINT NOT NULL AUTO_INCREMENT PRIMARY KEY,
    modified DATETIME DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
    custinfo JSON
    );

ALTER TABLE customers ADD INDEX comp(id, modified,
    (CAST(custinfo->'$.zipcode' AS UNSIGNED ARRAY)) );

Only one multi-valued key part can be used in a composite index. The multi-valued key part may be used in any order relative to the other parts of the key. In other words, the ALTER TABLE statement just shown could have used comp(id, (CAST(custinfo->'$.zipcode' AS UNSIGNED ARRAY), modified)) (or any other ordering) and still have been valid.

Using multi-valued Indexes

The optimizer uses a multi-valued index to fetch records when the following functions are specified in a WHERE clause:

We can demonstrate this by creating and populating the customers table using the following CREATE TABLE and INSERT statements:

mysql> CREATE TABLE customers (
    ->     id BIGINT NOT NULL AUTO_INCREMENT PRIMARY KEY,
    ->     modified DATETIME DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
    ->     custinfo JSON
    ->     );
Query OK, 0 rows affected (0.51 sec)

mysql> INSERT INTO customers VALUES
    ->     (NULL, NOW(), '{"user":"Jack","user_id":37,"zipcode":[94582,94536]}'),
    ->     (NULL, NOW(), '{"user":"Jill","user_id":22,"zipcode":[94568,94507,94582]}'),
    ->     (NULL, NOW(), '{"user":"Bob","user_id":31,"zipcode":[94477,94507]}'),
    ->     (NULL, NOW(), '{"user":"Mary","user_id":72,"zipcode":[94536]}'),
    ->     (NULL, NOW(), '{"user":"Ted","user_id":56,"zipcode":[94507,94582]}');
Query OK, 5 rows affected (0.07 sec)
Records: 5  Duplicates: 0  Warnings: 0

First we execute three queries on the customers table, one each using MEMBER OF(), JSON_CONTAINS(), and JSON_OVERLAPS(), with the result from each query shown here:

mysql> SELECT * FROM customers
    ->     WHERE 94507 MEMBER OF(custinfo->'$.zipcode');
+----+---------------------+-------------------------------------------------------------------+
| id | modified            | custinfo                                                          |
+----+---------------------+-------------------------------------------------------------------+
|  2 | 2019-06-29 22:23:12 | {"user": "Jill", "user_id": 22, "zipcode": [94568, 94507, 94582]} |
|  3 | 2019-06-29 22:23:12 | {"user": "Bob", "user_id": 31, "zipcode": [94477, 94507]}         |
|  5 | 2019-06-29 22:23:12 | {"user": "Ted", "user_id": 56, "zipcode": [94507, 94582]}         |
+----+---------------------+-------------------------------------------------------------------+
3 rows in set (0.00 sec)

mysql> SELECT * FROM customers
    ->     WHERE JSON_CONTAINS(custinfo->'$.zipcode', CAST('[94507,94582]' AS JSON));
+----+---------------------+-------------------------------------------------------------------+
| id | modified            | custinfo                                                          |
+----+---------------------+-------------------------------------------------------------------+
|  2 | 2019-06-29 22:23:12 | {"user": "Jill", "user_id": 22, "zipcode": [94568, 94507, 94582]} |
|  5 | 2019-06-29 22:23:12 | {"user": "Ted", "user_id": 56, "zipcode": [94507, 94582]}         |
+----+---------------------+-------------------------------------------------------------------+
2 rows in set (0.00 sec)

mysql> SELECT * FROM customers
    ->     WHERE JSON_OVERLAPS(custinfo->'$.zipcode', CAST('[94507,94582]' AS JSON));
+----+---------------------+-------------------------------------------------------------------+
| id | modified            | custinfo                                                          |
+----+---------------------+-------------------------------------------------------------------+
|  1 | 2019-06-29 22:23:12 | {"user": "Jack", "user_id": 37, "zipcode": [94582, 94536]}        |
|  2 | 2019-06-29 22:23:12 | {"user": "Jill", "user_id": 22, "zipcode": [94568, 94507, 94582]} |
|  3 | 2019-06-29 22:23:12 | {"user": "Bob", "user_id": 31, "zipcode": [94477, 94507]}         |
|  5 | 2019-06-29 22:23:12 | {"user": "Ted", "user_id": 56, "zipcode": [94507, 94582]}         |
+----+---------------------+-------------------------------------------------------------------+
4 rows in set (0.00 sec)

Next, we run EXPLAIN on each of the previous three queries:

mysql> EXPLAIN SELECT * FROM customers
    ->     WHERE 94507 MEMBER OF(custinfo->'$.zipcode');
+----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+
| id | select_type | table     | partitions | type | possible_keys | key  | key_len | ref  | rows | filtered | Extra       |
+----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | customers | NULL       | ALL  | NULL          | NULL | NULL    | NULL |    5 |   100.00 | Using where |
+----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+
1 row in set, 1 warning (0.00 sec)

mysql> EXPLAIN SELECT * FROM customers
    ->     WHERE JSON_CONTAINS(custinfo->'$.zipcode', CAST('[94507,94582]' AS JSON));
+----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+
| id | select_type | table     | partitions | type | possible_keys | key  | key_len | ref  | rows | filtered | Extra       |
+----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | customers | NULL       | ALL  | NULL          | NULL | NULL    | NULL |    5 |   100.00 | Using where |
+----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+
1 row in set, 1 warning (0.00 sec)

mysql> EXPLAIN SELECT * FROM customers
    ->     WHERE JSON_OVERLAPS(custinfo->'$.zipcode', CAST('[94507,94582]' AS JSON));
+----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+
| id | select_type | table     | partitions | type | possible_keys | key  | key_len | ref  | rows | filtered | Extra       |
+----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | customers | NULL       | ALL  | NULL          | NULL | NULL    | NULL |    5 |   100.00 | Using where |
+----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+
1 row in set, 1 warning (0.01 sec)

None of the three queries just shown are able to use any keys. To solve this problem, we can add a multi-valued index on the zipcode array in the JSON column (custinfo), like this:

mysql> ALTER TABLE customers
    ->     ADD INDEX zips( (CAST(custinfo->'$.zipcode' AS UNSIGNED ARRAY)) );
Query OK, 0 rows affected (0.47 sec)
Records: 0  Duplicates: 0  Warnings: 0

When we run the previous EXPLAIN statements again, we can now observe that the queries can (and do) use the index zips that was just created:

mysql> EXPLAIN SELECT * FROM customers
    ->     WHERE 94507 MEMBER OF(custinfo->'$.zipcode');
+----+-------------+-----------+------------+------+---------------+------+---------+-------+------+----------+-------------+
| id | select_type | table     | partitions | type | possible_keys | key  | key_len | ref   | rows | filtered | Extra       |
+----+-------------+-----------+------------+------+---------------+------+---------+-------+------+----------+-------------+
|  1 | SIMPLE      | customers | NULL       | ref  | zips          | zips | 9       | const |    1 |   100.00 | Using where |
+----+-------------+-----------+------------+------+---------------+------+---------+-------+------+----------+-------------+
1 row in set, 1 warning (0.00 sec)

mysql> EXPLAIN SELECT * FROM customers
    ->     WHERE JSON_CONTAINS(custinfo->'$.zipcode', CAST('[94507,94582]' AS JSON));
+----+-------------+-----------+------------+-------+---------------+------+---------+------+------+----------+-------------+
| id | select_type | table     | partitions | type  | possible_keys | key  | key_len | ref  | rows | filtered | Extra       |
+----+-------------+-----------+------------+-------+---------------+------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | customers | NULL       | range | zips          | zips | 9       | NULL |    6 |   100.00 | Using where |
+----+-------------+-----------+------------+-------+---------------+------+---------+------+------+----------+-------------+
1 row in set, 1 warning (0.00 sec)

mysql> EXPLAIN SELECT * FROM customers
    ->     WHERE JSON_OVERLAPS(custinfo->'$.zipcode', CAST('[94507,94582]' AS JSON));
+----+-------------+-----------+------------+-------+---------------+------+---------+------+------+----------+-------------+
| id | select_type | table     | partitions | type  | possible_keys | key  | key_len | ref  | rows | filtered | Extra       |
+----+-------------+-----------+------------+-------+---------------+------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | customers | NULL       | range | zips          | zips | 9       | NULL |    6 |   100.00 | Using where |
+----+-------------+-----------+------------+-------+---------------+------+---------+------+------+----------+-------------+
1 row in set, 1 warning (0.01 sec)

A multi-valued index can be defined as a unique key. If defined as a unique key, attempting to insert a value already present in the multi-valued index returns a duplicate key error. If duplicate values are already present, attempting to add a unique multi-valued index fails, as shown here:

mysql> ALTER TABLE customers DROP INDEX zips;
Query OK, 0 rows affected (0.55 sec)
Records: 0  Duplicates: 0  Warnings: 0

mysql> ALTER TABLE customers
    ->     ADD UNIQUE INDEX zips((CAST(custinfo->'$.zipcode' AS UNSIGNED ARRAY)));
ERROR 1062 (23000): Duplicate entry '[94507, ' for key 'customers.zips'
mysql> ALTER TABLE customers
    ->     ADD INDEX zips((CAST(custinfo->'$.zipcode' AS UNSIGNED ARRAY)));
Query OK, 0 rows affected (0.36 sec)
Records: 0  Duplicates: 0  Warnings: 0
Characteristics of Multi-Valued Indexes

Multi-valued indexes have the additional characteristics listed here:

  • DML operations that affect multi-valued indexes are handled in the same way as DML operations that affect a normal index, with the only difference being that there may be more than one insert or update for a single clustered index record.

  • Nullability and multi-valued indexes:

    • If a multi-valued key part has an empty array, no entries are added to the index, and the data record is not accessible by an index scan.

    • If multi-valued key part generation returns a NULL value, a single entry containing NULL is added to the multi-valued index. If the key part is defined as NOT NULL, an error is reported.

    • If the typed array column is set to NULL, the storage engine stores a single record containing NULL that points to the data record.

    • JSON null values are not permitted in indexed arrays. If any returned value is NULL, it is treated as a JSON null and an Invalid JSON value error is reported.

  • Because multi-valued indexes are virtual indexes on virtual columns, they must adhere to the same rules as secondary indexes on virtual generated columns.

  • Index records are not added for empty arrays.

Limitations and Restrictions on Multi-valued Indexes

Multi-valued indexes are subject to the limitations and restrictions listed here:

  • Only one multi-valued key part is permitted per multi-valued index. However, the CAST(... AS ... ARRAY) expression can refer to multiple arrays within a JSON document, as shown here:

    CAST(data->'$.arr[*][*]' AS UNSIGNED ARRAY)

    In this case, all values matching the JSON expression are stored in the index as a single flat array.

  • An index with a multi-valued key part does not support ordering and therefore cannot be used as a primary key. For the same reason, a multi-valued index cannot be defined using the ASC or DESC keyword.

  • A multi-valued index cannot be a covering index.

  • The maximum number of values per record for a multi-valued index is determined by the amount of data than can be stored on a single undo log page, which is 65221 bytes (64K minus 315 bytes for overhead), which means that the maximum total length of key values is also 65221 bytes. The maximum number of keys depends on various factors, which prevents defining a specific limit. Tests have shown a multi-valued index to permit as many as 1604 integer keys per record, for example. When the limit is reached, an error similar to the following is reported: ERROR 3905 (HY000): Exceeded max number of values per record for multi-valued index 'idx' by 1 value(s).

  • The only type of expression that is permitted in a multi-valued key part is a JSON expression. The expression need not reference an existing element in a JSON document inserted into the indexed column, but must itself be syntactically valid.

  • Because index records for the same clustered index record are dispersed throughout a multi-valued index, a multi-valued index does not support range scans or index-only scans.

  • Multi-valued indexes are not permitted in foreign key specifications.

  • Index prefixes cannot be defined for multi-valued indexes.

  • Multi-valued indexes cannot be defined on data cast as BINARY (see the description of the CAST() function).

  • Online creation of a multi-value index is not supported, which means the operation uses ALGORITHM=COPY. See Performance and Space Requirements.

  • Character sets and collations other than the following two combinations of character set and collation are not supported for multi-valued indexes:

    1. The binary character set with the default binary collation

    2. The utf8mb4 character set with the default utf8mb4_0900_as_cs collation.

  • As with other indexes on columns of InnoDB tables, a multi-valued index cannot be created with USING HASH; attempting to do so results in a warning: This storage engine does not support the HASH index algorithm, storage engine default was used instead. (USING BTREE is supported as usual.)

Spatial Indexes

The MyISAM, InnoDB, NDB, and ARCHIVE storage engines support spatial columns such as POINT and GEOMETRY. (Section 13.4, “Spatial Data Types”, describes the spatial data types.) However, support for spatial column indexing varies among engines. Spatial and nonspatial indexes on spatial columns are available according to the following rules.

Spatial indexes on spatial columns have these characteristics:

  • Available only for InnoDB and MyISAM tables. Specifying SPATIAL INDEX for other storage engines results in an error.

  • An index on a spatial column must be a SPATIAL index. The SPATIAL keyword is thus optional but implicit for creating an index on a spatial column.

  • Available for single spatial columns only. A spatial index cannot be created over multiple spatial columns.

  • Indexed columns must be NOT NULL.

  • Column prefix lengths are prohibited. The full width of each column is indexed.

  • Not permitted for a primary key or unique index.

Nonspatial indexes on spatial columns (created with INDEX, UNIQUE, or PRIMARY KEY) have these characteristics:

  • Permitted for any storage engine that supports spatial columns except ARCHIVE.

  • Columns can be NULL unless the index is a primary key.

  • The index type for a non-SPATIAL index depends on the storage engine. Currently, B-tree is used.

  • Permitted for a column that can have NULL values only for InnoDB, MyISAM, and MEMORY tables.

Index Options

Following the key part list, index options can be given. An index_option value can be any of the following:

  • KEY_BLOCK_SIZE [=] value

    For MyISAM tables, KEY_BLOCK_SIZE optionally specifies the size in bytes to use for index key blocks. The value is treated as a hint; a different size could be used if necessary. A KEY_BLOCK_SIZE value specified for an individual index definition overrides a table-level KEY_BLOCK_SIZE value.

    KEY_BLOCK_SIZE is not supported at the index level for InnoDB tables. See Section 15.1.20, “CREATE TABLE Statement”.

  • index_type

    Some storage engines permit you to specify an index type when creating an index. For example:

    CREATE TABLE lookup (id INT) ENGINE = MEMORY;
    CREATE INDEX id_index ON lookup (id) USING BTREE;

    Table 15.1, “Index Types Per Storage Engine” shows the permissible index type values supported by different storage engines. Where multiple index types are listed, the first one is the default when no index type specifier is given. Storage engines not listed in the table do not support an index_type clause in index definitions.

    Table 15.1 Index Types Per Storage Engine

    Storage Engine Permissible Index Types
    InnoDB BTREE
    MyISAM BTREE
    MEMORY/HEAP HASH, BTREE
    NDB HASH, BTREE (see note in text)

    The index_type clause cannot be used for FULLTEXT INDEX specifications. Full-text index implementation is storage-engine dependent. Spatial indexes are implemented as R-tree indexes.

    If you specify an index type that is not valid for a given storage engine, but another index type is available that the engine can use without affecting query results, the engine uses the available type. The parser recognizes RTREE as a type name. This is permitted only for SPATIAL indexes.

    BTREE indexes are implemented by the NDB storage engine as T-tree indexes.

    Note

    For indexes on NDB table columns, the USING option can be specified only for a unique index or primary key. USING HASH prevents the creation of an ordered index; otherwise, creating a unique index or primary key on an NDB table automatically results in the creation of both an ordered index and a hash index, each of which indexes the same set of columns.

    For unique indexes that include one or more NULL columns of an NDB table, the hash index can be used only to look up literal values, which means that IS [NOT] NULL conditions require a full scan of the table. One workaround is to make sure that a unique index using one or more NULL columns on such a table is always created in such a way that it includes the ordered index; that is, avoid employing USING HASH when creating the index.

    If you specify an index type that is not valid for a given storage engine, but another index type is available that the engine can use without affecting query results, the engine uses the available type. The parser recognizes RTREE as a type name, but currently this cannot be specified for any storage engine.

    Note

    Use of the index_type option before the ON tbl_name clause is deprecated; expect support for use of the option in this position to be removed in a future MySQL release. If an index_type option is given in both the earlier and later positions, the final option applies.

    TYPE type_name is recognized as a synonym for USING type_name. However, USING is the preferred form.

    The following tables show index characteristics for the storage engines that support the index_type option.

    Table 15.2 InnoDB Storage Engine Index Characteristics

    Index Class Index Type Stores NULL VALUES Permits Multiple NULL Values IS NULL Scan Type IS NOT NULL Scan Type
    Primary key BTREE No No N/A N/A
    Unique BTREE Yes Yes Index Index
    Key BTREE Yes Yes Index Index
    FULLTEXT N/A Yes Yes Table Table
    SPATIAL N/A No No N/A N/A

    Table 15.3 MyISAM Storage Engine Index Characteristics

    Index Class Index Type Stores NULL VALUES Permits Multiple NULL Values IS NULL Scan Type IS NOT NULL Scan Type
    Primary key BTREE No No N/A N/A
    Unique BTREE Yes Yes Index Index
    Key BTREE Yes Yes Index Index
    FULLTEXT N/A Yes Yes Table Table
    SPATIAL N/A No No N/A N/A

    Table 15.4 MEMORY Storage Engine Index Characteristics

    Index Class Index Type Stores NULL VALUES Permits Multiple NULL Values IS NULL Scan Type IS NOT NULL Scan Type
    Primary key BTREE No No N/A N/A
    Unique BTREE Yes Yes Index Index
    Key BTREE Yes Yes Index Index
    Primary key HASH No No N/A N/A
    Unique HASH Yes Yes Index Index
    Key HASH Yes Yes Index Index

    Table 15.5 NDB Storage Engine Index Characteristics

    Index Class Index Type Stores NULL VALUES Permits Multiple NULL Values IS NULL Scan Type IS NOT NULL Scan Type
    Primary key BTREE No No Index Index
    Unique BTREE Yes Yes Index Index
    Key BTREE Yes Yes Index Index
    Primary key HASH No No Table (see note 1) Table (see note 1)
    Unique HASH Yes Yes Table (see note 1) Table (see note 1)
    Key HASH Yes Yes Table (see note 1) Table (see note 1)

    Table note:

    1. USING HASH prevents creation of an implicit ordered index.

  • WITH PARSER parser_name

    This option can be used only with FULLTEXT indexes. It associates a parser plugin with the index if full-text indexing and searching operations need special handling. InnoDB and MyISAM support full-text parser plugins. If you have a MyISAM table with an associated full-text parser plugin, you can convert the table to InnoDB using ALTER TABLE. See Full-Text Parser Plugins and Writing Full-Text Parser Plugins for more information.

  • COMMENT 'string'

    Index definitions can include an optional comment of up to 1024 characters.

    The MERGE_THRESHOLD for index pages can be configured for individual indexes using the index_option COMMENT clause of the CREATE INDEX statement. For example:

    CREATE TABLE t1 (id INT);
    CREATE INDEX id_index ON t1 (id) COMMENT 'MERGE_THRESHOLD=40';

    If the page-full percentage for an index page falls below the MERGE_THRESHOLD value when a row is deleted or when a row is shortened by an update operation, InnoDB attempts to merge the index page with a neighboring index page. The default MERGE_THRESHOLD value is 50, which is the previously hardcoded value.

    MERGE_THRESHOLD can also be defined at the index level and table level using CREATE TABLE and ALTER TABLE statements. For more information, see Section 17.8.11, “Configuring the Merge Threshold for Index Pages”.

  • VISIBLE, INVISIBLE

    Specify index visibility. Indexes are visible by default. An invisible index is not used by the optimizer. Specification of index visibility applies to indexes other than primary keys (either explicit or implicit). For more information, see Section 10.3.12, “Invisible Indexes”.

  • The ENGINE_ATTRIBUTE and SECONDARY_ENGINE_ATTRIBUTE are used to specify index attributes for primary and secondary storage engines. The options are reserved for future use.

    The value assigned to this option is a string literal containing a valid JSON document or an empty string (''). Invalid JSON is rejected.

    CREATE INDEX i1 ON t1 (c1) ENGINE_ATTRIBUTE='{"key":"value"}';

    ENGINE_ATTRIBUTE and SECONDARY_ENGINE_ATTRIBUTE values can be repeated without error. In this case, the last specified value is used.

    ENGINE_ATTRIBUTE and SECONDARY_ENGINE_ATTRIBUTE values are not checked by the server, nor are they cleared when the table's storage engine is changed.

Table Copying and Locking Options

ALGORITHM and LOCK clauses may be given to influence the table copying method and level of concurrency for reading and writing the table while its indexes are being modified. They have the same meaning as for the ALTER TABLE statement. For more information, see Section 15.1.9, “ALTER TABLE Statement”

NDB Cluster supports online operations using the same ALGORITHM=INPLACE syntax used with the standard MySQL Server. See Section 25.6.12, “Online Operations with ALTER TABLE in NDB Cluster”, for more information.