Almost every SQL statement refers in some way to a database or its constituent elements. This section describes the syntax and case sensitivity rules for identifiers that refer to databases, tables, columns, indexes, and aliases.
Referring to Elements of Databases
When you use identifiers to refer to elements of databases, you are constrained by the characters you can use and the length that identifiers can be. The format of identifiers also depends on the context in which you use them. Another factor that affects naming rules is that the server can be configured to use different SQL modes.
- Legal characters in identifiers. Unquoted identifiers can consist of any alphanumeric characters in the system default character set (utf8), plus the characters ‘_‘ and ‘$‘. Identifiers can start with any character that is legal in an identifier, including a digit. However, an identifier cannot consist entirely of digits because that would make it indistinguishable from a number. MySQL’s support for identifiers that begin with a number is somewhat unusual among database systems. If you use such an identifier, be particularly careful if it contains an ‘E‘ or ‘e‘ because those characters can lead to ambiguous expressions. For example, the expression 23e + 14 (with spaces surrounding the ‘+‘ sign) means column 23e plus the number 14, but what about 23e+14? Does it mean the same thing, or is it a number in scientific notation? You should also be careful about using identifiers such as 0x1020 that begin with 0x because they might be interpreted as hexadecimal constants.
Identifiers can be quoted (delimited) within backtick characters (‘´‘), which allows use of any character except backtick or a byte with value 0 or 255:
CREATE TABLE ´my table´ (´my column´ INT);
Quoting is useful when an identifier is an SQL keyword or contains spaces or other special characters. Quoting an identifier also allows it to be entirely numeric, something that is not true of unquoted identifiers. To include an identifier quote within a quoted identifier, double it.
For database and table identifiers, there are two additional constraints, even for identifiers that are quoted. First, you cannot use the ‘.‘ character, because it is used as the separator character in qualified name notation of the forms db_name.tbl_name and db_name.tbl_name.col_name. Second, you cannot use the Unix or Windows pathname separator characters (‘/‘ or ‘\‘). The pathname separator is disallowed in database and table identifiers because databases are represented on disk by directories, and tables are represented on disk by at least one file. Consequently, these types of identifiers must contain only characters that are legal in directory names and filenames. The Unix pathname separator is disallowed on Windows (and vice versa) to make it easier to transfer databases and tables between servers running on different platforms. (Suppose that you were allowed to use a slash in a table name on Windows. That would make it impossible to move the table to Unix, because filenames on that platform cannot contain slashes.)
Your operating system might impose additional constraints on database and table identifiers. See “Operating System Constraints on Database and Table Naming,” in Chapter 10, “The MySQL Data Directory.”
Column and table aliases can be fairly arbitrary. You should quote an alias within identifier quoting characters if it is an SQL keyword, is entirely numeric, or contains spaces or other special characters. Column aliases also can be quoted with single quotes or double quotes.
Server SQL mode. If the ANSI_QUOTES SQL mode is enabled, you can quote identifiers with double quotes (although backticks still are allowable).
CREATE TABLE "my table" ("my column" INT);
Note: Enabling ANSI_QUOTES has the additional effect that string literals must be written using single quotes. If you use double quotes, the server will interpret the value as an identifier, not as a string.
Function names normally are not reserved and can be used as identifiers without quotes. However, if the IGNORE_SPACES SQL mode is enabled, function names become reserved and must be quoted if used as identifiers.
For instructions on setting the SQL mode, see “The Server SQL Mode” later in this chapter.
Identifier length. Identifiers for databases, tables, columns, and indexes can be up to 64 characters long. Identifiers are stored using utf8 characters. (Before MySQL 4.1.5, the maximum identifier length is 64 bytes, not characters. Because utf8 characters take from one to three bytes each, the effective maximum identifier length is less than 64 characters if you use multi-byte characters.) Aliases can be up to 256 characters long.
Identifier qualifiers. Depending on context, an identifier might need to be qualified to make clear what it refers to. To refer to a database, just specify its name:
- A fully qualified table name consists of a database identifier and a table identifier:
SHOW COLUMNS FROM db_name.tbl_name;
SELECT * FROM db_name.tbl_name;
A table identifier by itself refers to a table in the default (current) database. If sampdb is the default database, the following statements are equivalent:
SELECT * FROM member;
SELECT * FROM sampdb.member;
- A name written as db_name.tbl_name.col_name is fully qualified.
- A partially qualified name written as tbl_name.col_name refers to a column in the named table in the default database.
- An unqualified name written simply as col_name refers to whatever table is indicated by the surrounding context. The following two queries use the same column names, but the context supplied by the FROM clause of each statement indicates which table to select the columns from:
SELECT last_name, first_name FROM president;
SELECT last_name, first_name FROM members;
SHOW TABLES FROM db_name;
To refer to a table, you have two choices:
If no database has been selected, you cannot refer to a table without specifying a database qualifier because the server cannot tell which database the table belongs to.
To refer to a column, there are three choices: fully qualified, partially qualified, and unqualified.
It’s usually unnecessary to supply fully qualified names, although it’s always legal to do so if you like. If you select a database with a USE statement, that database becomes the default database and is implicit in every unqualified table reference. If you’re using a SELECT statement that refers to only one table, that table is implicit for every column reference in the statement. It’s necessary to qualify identifiers only when a table or database cannot be determined from context. For example, if a statement refers to tables from multiple databases, any table not in the default database must be referenced using the db_name.tbl_name form to let MySQL know which database contains the table. Similarly, if a query uses multiple tables and refers to a column name that is used in more than one table, it’s necessary to qualify the column identifier with a table identifier to make it clear which column you mean.
If you use quotes when referring to a qualified name, quote individual parts of the name separately. For example:
SELECT * FROM ´sampdb´.´member´ WHERE ´sampdb´.´member´.´member_id´ > 100;
Do not quote the name as a whole. This statement is illegal:
SELECT * FROM ´sampdb.member´ WHERE ´sampdb.member.member_id´ >