Character Sets and Collations

Declaration

CREATE TABLE foo ( ...
    name CHARACTER SET utf8mb4
    ... );

Connection

Vital to using character sets is to tell the MySQL-server what encoding the client’s bytes are. Here is one way:

SET NAMES utf8mb4;

Each language (PHP, Python, Java, …) has its own way the it usually preferable to SET NAMES.

For example: SET NAMES utf8mb4, together with a column declared CHARACTER SET latin1 — this will convert from latin1 to utf8mb4 when INSERTing and convert back when SELECTing.

Which CHARACTER SET and COLLATION?

There are dozens of character sets with hundreds of collations. (A given collation belongs to only one character set.) See the output of SHOW COLLATION;.

There are usually only 4 CHARACTER SETs that matter:

ascii -- basic 7-bit codes.
latin1 -- ascii, plus most characters needed for Western European languages.
utf8 -- the 1-, 2-, and 3-byte subset of utf8.  This excludes Emoji and some of Chinese.
utf8mb4 -- the full set of UTF8 characters, covering all current languages.

All include English characters, encoded identically. utf8 is a subset of utf8mb4.

Best practice…

Use utf8mb4 for any TEXT or VARCHAR column that can have a variety of languages in it.
Use ascii (latin1 is ok) for hex strings (UUID, MD5, etc) and simple codes (country_code, postal_code, etc).

utf8mb4 did not exist until version 5.5.3, so utf8 was the best available before that.

Outside of MySQL, “UTF8” means the same things as MySQL’s utf8mb4, not MySQL’s utf8.

Collations start with the charset name and usually end with _ci for “case and accent insensitive” or _bin for “simply compare the bits.

The ‘latest’ utf8mb4 collation is utf8mb4_unicode_520_ci, based on Unicode 5.20. If you are working with a single language, you might want, say, utf8mb4_polish_ci, which will rearrange the letters slightly, based on Polish conventions.

Setting character sets on tables and fields

You can set a character set both per table, as well as per individual field using the CHARACTER SET and CHARSET statements:

CREATE TABLE Address (
    `AddressID`   INTEGER NOT NULL PRIMARY KEY,
    `Street`      VARCHAR(80) CHARACTER SET ASCII,
    `City`        VARCHAR(80),
    `Country`     VARCHAR(80) DEFAULT "United States",
    `Active`      BOOLEAN DEFAULT 1,
) Engine=InnoDB default charset=UTF8;

City and Country will use UTF8, as we set that as the default character set for the table. Street on the other hand will use ASCII, as we’ve specifically told it to do so.

Setting the right character set is highly dependent on your dataset, but can also highly improve portability between systems working with your data.