10.9.8 Converting Between 3-Byte and 4-Byte Unicode Character Sets

This section describes issues that you may face when converting character data between the utf8mb3 and utf8mb4 character sets.

Note

This discussion focuses primarily on converting between utf8mb3 and utf8mb4, but similar principles apply to converting between the ucs2 character set and character sets such as utf16 or utf32.

The utf8mb3 and utf8mb4 character sets differ as follows:

utf8mb3 supports only characters in the Basic Multilingual Plane (BMP). utf8mb4 additionally supports supplementary characters that lie outside the BMP.
utf8mb3 uses a maximum of three bytes per character. utf8mb4 uses a maximum of four bytes per character.

Note

This discussion refers to the utf8mb3 and utf8mb4 character set names to be explicit about referring to 3-byte and 4-byte UTF-8 character set data. The exception is that in table definitions, utf8 is used because MySQL converts instances of utf8mb3 specified in such definitions to utf8, which is an alias for utf8mb3.

One advantage of converting from utf8mb3 to utf8mb4 is that this enables applications to use supplementary characters. One tradeoff is that this may increase data storage space requirements.

In terms of table content, conversion from utf8mb3 to utf8mb4 presents no problems:

For a BMP character, utf8mb4 and utf8mb3 have identical storage characteristics: same code values, same encoding, same length.
For a supplementary character, utf8mb4 requires four bytes to store it, whereas utf8mb3 cannot store the character at all. When converting utf8mb3 columns to utf8mb4, you need not worry about converting supplementary characters because there are none.

In terms of table structure, these are the primary potential incompatibilities:

For the variable-length character data types (VARCHAR and the TEXT types), the maximum permitted length in characters is less for utf8mb4 columns than for utf8mb3 columns.
For all character data types (CHAR, VARCHAR, and the TEXT types), the maximum number of characters that can be indexed is less for utf8mb4 columns than for utf8mb3 columns.

Consequently, to convert tables from utf8mb3 to utf8mb4, it may be necessary to change some column or index definitions.

Tables can be converted from utf8mb3 to utf8mb4 by using ALTER TABLE. Suppose that a table has this definition:

CREATE TABLE t1 (
  col1 CHAR(10) CHARACTER SET utf8 COLLATE utf8_unicode_ci NOT NULL,
  col2 CHAR(10) CHARACTER SET utf8 COLLATE utf8_bin NOT NULL
) CHARACTER SET utf8;

The following statement converts t1 to use utf8mb4:

ALTER TABLE t1
  DEFAULT CHARACTER SET utf8mb4,
  MODIFY col1 CHAR(10)
    CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci NOT NULL,
  MODIFY col2 CHAR(10)
    CHARACTER SET utf8mb4 COLLATE utf8mb4_bin NOT NULL;

The catch when converting from utf8mb3 to utf8mb4 is that the maximum length of a column or index key is unchanged in terms of bytes. Therefore, it is smaller in terms of characters because the maximum length of a character is four bytes instead of three. For the CHAR, VARCHAR, and TEXT data types, watch for these issues when converting your MySQL tables:

Check all definitions of utf8mb3 columns and make sure they do not exceed the maximum length for the storage engine.
Check all indexes on utf8mb3 columns and make sure they do not exceed the maximum length for the storage engine. Sometimes the maximum can change due to storage engine enhancements.

If the preceding conditions apply, you must either reduce the defined length of columns or indexes, or continue to use utf8mb3 rather than utf8mb4.

Here are some examples where structural changes may be needed:

A TINYTEXT column can hold up to 255 bytes, so it can hold up to 85 3-byte or 63 4-byte characters. Suppose that you have a TINYTEXT column that uses utf8mb3 but must be able to contain more than 63 characters. You cannot convert it to utf8mb4 unless you also change the data type to a longer type such as TEXT.
Similarly, a very long VARCHAR column may need to be changed to one of the longer TEXT types if you want to convert it from utf8mb3 to utf8mb4.
InnoDB has a maximum index length of 767 bytes for tables that use COMPACT or REDUNDANT row format, so for utf8mb3 or utf8mb4 columns, you can index a maximum of 255 or 191 characters, respectively. If you currently have utf8mb3 columns with indexes longer than 191 characters, you must index a smaller number of characters.
In an InnoDB table that uses COMPACT or REDUNDANT row format, these column and index definitions are legal:
```
col1 VARCHAR(500) CHARACTER SET utf8, INDEX (col1(255))
```
To use utf8mb4 instead, the index must be smaller:
```
col1 VARCHAR(500) CHARACTER SET utf8mb4, INDEX (col1(191))
```
Note
For InnoDB tables that use COMPRESSED or DYNAMIC row format, you can enable the innodb_large_prefix option to permit index key prefixes longer than 767 bytes (up to 3072 bytes). Creating such tables also requires the option values innodb_file_format=barracuda and innodb_file_per_table=true.) In this case, enabling the innodb_large_prefix option enables you to index a maximum of 1024 or 768 characters for utf8mb3 or utf8mb4 columns, respectively. For related information, see Section 14.23, “InnoDB Limits”.

The preceding types of changes are most likely to be required only if you have very long columns or indexes. Otherwise, you should be able to convert your tables from utf8mb3 to utf8mb4 without problems, using ALTER TABLE as described previously.

The following items summarize other potential incompatibilities:

SET NAMES 'utf8mb4' causes use of the 4-byte character set for connection character sets. As long as no 4-byte characters are sent from the server, there should be no problems. Otherwise, applications that expect to receive a maximum of three bytes per character may have problems. Conversely, applications that expect to send 4-byte characters must ensure that the server understands them.
For replication, if character sets that support supplementary characters are to be used on the source, all replicas must understand them as well.
Also, keep in mind the general principle that if a table has different definitions on the source and replica, this can lead to unexpected results. For example, the differences in maximum index key length make it risky to use utf8mb3 on the source and utf8mb4 on the replica.

If you have converted to utf8mb4, utf16, utf16le, or utf32, and then decide to convert back to utf8mb3 or ucs2 (for example, to downgrade to an older version of MySQL), these considerations apply:

utf8mb3 and ucs2 data should present no problems.
The server must be recent enough to recognize definitions referring to the character set from which you are converting.
For object definitions that refer to the utf8mb4 character set, you can dump them with mysqldump prior to downgrading, edit the dump file to change instances of utf8mb4 to utf8, and reload the file in the older server, as long as there are no 4-byte characters in the data. The older server sees utf8 in the dump file object definitions and creates new objects that use the (3-byte) utf8 character set.