Alpha

These pages are built from main, turborel 0.0.0. Anything may change from one build to the next: SQL, API, protocol and files alike.

Lexical Elements

Characters

SQL text is Unicode. Its characters are those of the one character set, SQL_TEXT, which holds every Unicode character; a character string holds any of them. A line ends with a newline, \n or \r\n.

Key words, operators and punctuation are written in the characters of ASCII; identifiers and string literals may hold any character.

Tokens and separators

SQL text is a sequence of tokens: key words, identifiers, literals, operators and punctuation. A separator, one or more spaces, newlines or comments, goes between two tokens that would otherwise run together: a key word, an identifier or a number is followed by a separator or by a delimiter, as (, ,, ' or an operator.

A tab is not a separator: SQL-92's separators are the space character, newlines and comments alone.

VALUES	1;
VALUES 1abc;
VALUES 1 @;
turborel: SQLCODE -101 (SQLSTATE 2A000): "VALUES" must be followed by a <delimiter token> or a <separator>, not '\t' [X3.135-1986 §5.3 SR2] (at character 6)
turborel: SQLCODE -101 (SQLSTATE 2A000): "1" must be followed by a <delimiter token> or a <separator>, not 'a' [X3.135-1986 §5.3 SR2] (at character 8)
turborel: SQLCODE -101 (SQLSTATE 2A000): character '@' cannot start a <token> [X3.135-1986 §5.3 F] (at character 9)

The operators and delimiters are ( ) , . ; ' " + - * / = < > <= >= <> || ? and :. Not equal is written <>: != is not SQL. The distance operators of vectors, <-> <=> <#> <+> <~> <%>, are delimiters too (Vectors).

Identifiers

An identifier names a table, a column, a schema or another object.

A regular identifier is a letter, then letters, digits and underscores, at most 128 of them. It is taken in upper case, whatever case it is written in: staff, Staff and STAFF are one name, STAFF. A key word is not a regular identifier.

A delimited identifier is written in double quotes, a double quote in it written twice. It is taken as written, its case kept, and may be a key word or hold any character: "select", "Mixed Case". A delimited identifier of upper case letters, digits and underscores is the regular one: "STAFF" is STAFF.

CREATE TABLE "select" ("Mixed Case" INTEGER, abc INTEGER);
SELECT COLUMN_NAME FROM INFORMATION_SCHEMA.COLUMNS WHERE TABLE_NAME = 'select' ORDER BY 1;
CREATE TABLE value (K INTEGER);
OK
 COLUMN_NAME
--------------------------------------------------------------
 ABC
 Mixed Case
(2 rows)
turborel: SQLCODE -102 (SQLSTATE 2A000): expected an <identifier>, found VALUE (at character 13)

A name of an object in a schema may be qualified by the schema's, ANN.STAFF, and a schema's by the catalog's; a column's by its table's, or a correlation name's, S.NAME. See Names.

Literals

Literal Written Of type
Character string 'it''s', '' CHARACTER of its length
National character string N'text' CHARACTER of its length
Binary string X'1F', X'0A 0B' BINARY of its octets, two hexits each
Exact number 42, -3.75, .5, 5., 1_000_000 NUMERIC of its digits
Approximate number 1.5E3, -2E-1 FLOAT(53), of DOUBLE PRECISION's precision
Boolean TRUE, FALSE, UNKNOWN BOOLEAN, UNKNOWN its null
Date DATE '2026-10-28' DATE
Time TIME '13:45:30.25', TIME '13:45:30+02:00' TIME of its fraction's digits
Timestamp TIMESTAMP '2026-10-28 13:45:30' TIMESTAMP of its fraction's digits
Interval INTERVAL '1-6' YEAR TO MONTH, INTERVAL -'3' DAY INTERVAL of its qualifier

A quote in a character string is written twice. A string may be written in parts, each quoted, separated by a separator holding a newline: the parts are one string. A character string may begin with an introducer naming its character set, _SQL_TEXT'text', of SQL_TEXT alone. A binary string holds an even number of hexits, spaces among them passed over, and may be written in parts as a character string may. The letters X, N and E may be written in either case. SQL-92's bit string literal, B'1010', is gone, as SQL:2023 has none.

An exact number holds at most 18 digits; an approximate number is within the range of DOUBLE PRECISION. A datetime literal is a valid date or time:

VALUES ('ab'
  'cd', 'it''s', '', N'n', X'1F', x'0a 0b'
  'FF', 1.5E3, .5, 5.);
VALUES 'ab' 'cd';
VALUES 123456789012345678901;
VALUES 1E400;
VALUES TIME '25:00:00';
      |      |  |   |    |        |                          |     |
------+------+--+---+----+--------+--------------------------+-----+-----
 abcd | it's |  | n | 1F | 0A0BFF |                    1.5E3 | 0.5 |   5
(1 row)
turborel: SQLCODE -102 (SQLSTATE 2A000): expected end of statement, found 'cd' (at character 12)
turborel: SQLCODE -900 (SQLSTATE 0A000): <exact numeric literal> 123456789012345678901 has 21 digits; this implementation's maximum exact precision is 18 [X3.135-1986 §5.2 SR4] (at character 7)
turborel: SQLCODE -900 (SQLSTATE 0A000): 1E400 is outside the range of this implementation's approximate numbers [X3.135-1986 §5.2 SR5] (at character 7)
turborel: SQLCODE -103 (SQLSTATE 2A000): TIME '25:00:00' is no valid <datetime literal> [ISO/IEC 9075:1992 §5.3 SR21, SR22; ISO/IEC 9075-2:2023 §5.3 SR43] (at character 7)

The parts of 'ab' 'cd' on one line are two strings, not one.

A number may hold an underscore between two of its digits, in its whole part, its fraction or its exponent, to group them. It is the same number without them, and they count toward no limit. One anywhere else, first, last, beside a period or another underscore, is refused. CAST reads a string's number the same way, but a datetime's or an interval's string holds no underscore.

VALUES (1_000_000, 3.141_592, 6.022_140_76E23, CAST('1_000' AS INTEGER));
VALUES 1__000;
           |           |                          |
-----------+-----------+--------------------------+--------------
   1000000 |  3.141592 |            6.02214076E23 |         1000
(1 row)
turborel: SQLCODE -101 (SQLSTATE 2A000): "1" must be followed by a <delimiter token> or a <separator>, not '_' [X3.135-1986 §5.3 SR2] (at character 8)

Comments

A comment begins with two minus signs and runs to the end of the line:

SELECT NAME   -- the name alone
  FROM STAFF;

A comment is a separator. It needs the newline that ends it; the client ends one that the end of its input cuts short.

A bracketed comment, of SQL:2023's Feature T351, begins with /* and ends with */, across lines if need be. One begun within it is nested and ends first, so a comment may hold one. A statement whose comment is not ended is refused (SQLSTATE 2A000, SQLCODE -101):

VALUES (1 /* one /* nested */ still the comment */ + 2);

A bracketed comment that begins a statement and whose text begins with + is a hint. /*+ VECTOR_WAIT_FRESH */ asks for the vectors waiting for the queue after commit first (see CREATE VECTOR PIPELINE). A hint not known is a comment alone.

Key words

The key words are SQL-92's reserved words, in any case: select is SELECT. None is an identifier, so a column named as one, VALUE or POSITION, is written delimited, "VALUE". A few words are key words only where a statement writes them, and identifiers everywhere else: SQL-92's non-reserved words that Turborel reads, as SERIALIZABLE and COMMITTED, and INDEX, PASSWORD, DATABASE and OWNER, of Turborel's extensions. SET DATABASE followed by a setting's name is SET DATABASE; otherwise, as in SET DATABASE = 1, DATABASE is the name of the setting SET sets. Key Words lists them all.

The standard: ISO/IEC 9075:1992 §5.1–§5.3; X3.135-1986 §5.1–§5.3; ISO/IEC 9075-2:2023 §5.3 for underscores, binary string literals and boolean literals.