Identifiers

Normal identifiers

Most dialects support [a-zA-Z_] as first character and [a-zA-Z0-9_] as rest of the characters. The differences from this are listed below:

• BigQuery: single dashes (-) can be used, but not at the beginning or end.
• DB2: first char can be an uppercase letter (a lowercase letter gets converted to uppercase). My testing in dbfiddle shows that @, #, or $ characters can be used anywhere inside an identifier.
• DB2i: like DB2. The IBM i docs state that only the first char can be @, #, or $, but #550 suggests that these can also appear after first character.
• DuckDB: additionally $ after first char.¹
• Hive: (no differences)
• MariaDB: no first-letter restrictions. The characters [a-zA-Z0-9_$] and unicode letters are allowed everywhere. Can begin with digit, but can't only contain digits.
• MySQL: same as MariaDB.
• N1QL: (no differences)
• PL/SQL: can't start with _. Allows $, # in rest of the identifier.
• PostgreSQL: additionally $ after first char. Also unicode letters are allowed.
• Redshift: also unicode letters are allowed. # is allowed as the first char of temporary table names.
• SingleStoreDB: Same as MariaDB.⁵
• Snowflake: additionally $ after first char allowed.
• Spark: no first-letter restrictions. The characters [a-zA-Z0-9_] are allowed everywhere. Can begin with digit, but can't only contain digits. The docs are confusing...
• SQLite: (no differences)
• TiDB: Same as MySQL and MariaDB.
• Transact-SQL: @ and # are allowed as first chars plus $ in the rest. Also unicode letters are allowed. Though the beginning @ signifies a local variable or parameter and # a temporary table or procedure.
• Trino: (no differences)⁴

Notes:

1. Tested DuckDB syntax on their playground: https://shell.duckdb.org/

Delimited identifiers

SQL standard specifies double-quotes ".." for delimited identifiers. There is a considerable variation in implementations:

• `..` BigQuery
• ".." DB2 (repeated " used for escaping)
• ".." DB2i It seems that escaping is not supported. Also a single quote ' can be configured for quoting.
• ".." DuckDB (repeated " used for escaping)
• `..` Hive (repeated ` used for escaping)
• `..`, (".."¹, [..]²) MariaDB (repeated ` used for escaping)
• `..`, (".."¹) MySQL (repeated ` used for escaping)
• `..` SingleStoreDB⁵
• `..` N1QL
• ".." PL/SQL (escaping of quotes is not supported)
• "..", U&".." PostgreSQL (repeated " used for escaping)
• ".." Redshift (repeated " used for escaping)
• ".." Snowflake (repeated " used for escaping)
• `..` Spark (repeated ` used for escaping)
• "..", `..`, [..] SQLite (repeated " or ` used for escaping)
• `..`, (".."¹) TiDB (repeated ` used for escaping)
• ".."³, [..] Transact-SQL (repeated " or ] used for escaping)
• ".."⁴ Trino (repeated " used for escaping)

Notes:

1. when ANSI_QUOTES mode enabled
2. when MSSQL mode enabled
3. unless QUOTED_IDENTIFIER option has been set OFF
4. Trino grammar lists `..`-quoted identifiers and identifiers starting with number, only to print an error message saying: these aren't supported.
5. Tested SingleStoreDB syntax manually. Haven't found any documentation for these low-level details.