ENUM

Why ENUM?

ENUM provides a way to provide an attribute for a row. Attributes with a small number of non-numeric options work best. Examples:

reply ENUM('yes', 'no')
gender ENUM('male', 'female', 'other', 'decline-to-state')

The values are strings:

INSERT ... VALUES ('yes', 'female')
SELECT ... --> yes female

VARCHAR as an alternative

Let’s say we have

type ENUM('fish','mammal','bird')

An alternative is

type VARCHAR(20)  COMENT "fish, bird, etc"

This is quite open-ended in that new types are trivially added.

Comparison, and whether better or worse than ENUM:

• (same) INSERT: simply provide the string
• (worse?) On INSERT a typo will go unnoticed
• (same) SELECT: the actual string is returned
• (worse) A lot more space is consumed

Adding a new option

ALTER TABLE tbl MODIFY COLUMN type ENUM('fish','mammal','bird','insect');

Notes

• As with all cases of MODIFY COLUMN, you must include NOT NULL, and any other qualifiers that originally existed, else they will be lost.
• If you add to the end of the list and the list is under 256 items, the ALTER is done by merely changing the schema. That is there will not be a lengthy table copy. (Old versions of MySQL did not have this optimization.)

NULL vs NOT NULL

Examples of what happens when NULL and ‘bad-value’ are stored into nullable and not nullable columns. Also shows usage of casting to numeric via +0.

CREATE TABLE enum (
    e     ENUM('yes', 'no')   NOT NULL,
    enull ENUM('x', 'y', 'z')     NULL
        );
INSERT INTO enum (e, enull)
    VALUES
        ('yes', 'x'),
        ('no',  'y'),
        (NULL,  NULL),
        ('bad-value', 'bad-value');
Query OK, 4 rows affected, 3 warnings (0.00 sec)
Records: 4  Duplicates: 0  Warnings: 3

mysql>SHOW WARNINGS;
+---------+------+--------------------------------------------+
| Level   | Code | Message                                    |
+---------+------+--------------------------------------------+
| Warning | 1048 | Column 'e' cannot be null                  |
| Warning | 1265 | Data truncated for column 'e' at row 4     |
| Warning | 1265 | Data truncated for column 'enull' at row 4 |
+---------+------+--------------------------------------------+
3 rows in set (0.00 sec)

What is in the table after those inserts. This uses “+0” to cast to numeric see what is stored.

mysql>SELECT e, e+0 FROM enum;
+-----+-----+
| e   | e+0 |
+-----+-----+
| yes |   1 |
| no  |   2 |
|     |   0 |  -- NULL
|     |   0 |  -- 'bad-value'
+-----+-----+
4 rows in set (0.00 sec)

mysql>SELECT enull, enull+0 FROM enum;
+-------+---------+
| enull | enull+0 |
+-------+---------+
| x     |       1 |
| y     |       2 |
| NULL  |    NULL |
|       |       0 |  -- 'bad-value'
+-------+---------+
4 rows in set (0.00 sec)

TINYINT as an alternative

Let’s say we have

type ENUM('fish','mammal','bird')

An alternative is

type TINYINT UNSIGNED

plus

CREATE TABLE AnimalTypes (
    type TINYINT UNSIGNED NOT NULL AUTO_INCREMENT,
    name VARCHAR(20) NOT NULL  COMMENT "('fish','mammal','bird')",
    PRIMARY KEY(type),
    INDEX(name)
) ENGINE=InnoDB

which is very much like a many-to-many table.

Comparison, and whether better or worse than ENUM:

• (worse) INSERT: need to lookup the type
• (worse) SELECT: need to JOIN to get the string (ENUM gives you the string with no effort)
• (better) Adding new types: Simply insert into this table. With ENUM, you need to do an ALTER TABLE.
• (same) Either technique (for up to 255 values) takes only 1 byte.
• (mixed) There’s also an issue of data integrity: TINYINT will admit invalid values; whereas ENUM sets them to a special empty-string value (unless strict SQL mode is enabled, in which case they are rejected). Better data integrity can be achieved with TINYINT by making it a foreign key into a lookup table: which, with appropriate queries/joins, but there is still the small cost of reaching into the other table. (FOREIGN KEYs are not free.)