Relix

Language reference

String, number and boolean literals

Syntax

"<text>"          -- a string, in double quotes
'<text>'          -- the same string, in single quotes
<digits>[.<digits>][e[±]<digits>]   -- a number: 42, 3.25, 0.5, 1e6, 2.5E-3
TRUE | FALSE      -- a boolean, in any case

π name, "in stock" → status, price * 1.2 → gross, TRUE → listed (Products)

Description

A literal is a value written straight into a query: a piece of text to compare a column with, a number to multiply by, a flag to add. Text goes in quotes, either double or single, numbers are written as plain digits, and the two booleans are TRUE and FALSE. Dates, times, timestamps and durations have literals of their own, written as a keyword followed by quoted text, such as DATE '2026-01-31'.

Technical Description

A string literal is text between two double quotes or two single quotes. The two forms are the same string, so "Asha" and 'Asha' compare equal. Inside either, a backslash escapes the next character: \" and \' write a quote, \\ a backslash, \n a newline and \t a tab. Any other character after a backslash is a syntax error. A quote of the other kind needs no escape, so "it's" and 'say "hi"' are both fine. As in SQL, the string's own quote doubled is one quote: 'it''s' is the same string as 'it\'s' and "it's", and '''' is a single quote.

A number literal is a run of digits with an optional fraction, such as 42, 3.25 or 0.5. A fraction needs a digit on each side of the point, so .5 and 5. are syntax errors. A number may end in an exponent, e or E and a power of ten with an optional sign: 1e6 is 1000000, 2.5E-3 is 0.0025 and 1.5e+2 is 150. It is read as the number written out in full, so it is as exact as any other. A negative number is the minus operator applied to a literal: -7 is -(7). Numbers are exact decimals of any size, so 0.1 + 0.2 = 0.3 holds and a 64-bit id keeps every digit.

A boolean literal is TRUE or FALSE, matched in any case.

NULL, or ⊥, is the missing value itself, written where a value goes: IIf(late, NULL, fee), COUNT(IIf(paid, id, NULL)). It takes its type from where it stands, the other branch of a conditional, and is ANY alone. In a null test it is not a value but part of the test: x = NULL and x IS NULL both ask whether x is missing. Comparing with it in any other way, x < NULL or NULL = x, is refused, since such a comparison is never true. In an inline table a missing value is an empty cell, ⊥ or NULL.

In an inline table, cells are bare values and are not written as literals: a cell holding "Asha" is the text with its quotes.

π .5 → half (Products)
π "a\qb" → s (Products)

Examples

Compare a column with a string:

σ name = "Asha" (Users)

The same comparison in single quotes:

σ name = 'Asha' (Users)

A quote of the other kind inside a string, and an escaped one:

π name, "it's" → note, 'say "hi"' → greeting, "say \"hi\"" → same (Users)

SQL's doubled quote, the same string as 'it\'s':

π name, 'it''s' → note (Users)

Numbers in exponent notation:

π name, price * 1e3 → milli, tax * 2.5E-3 → scaled (Products)

Numbers in arithmetic and a comparison:

π name, price * 1.2 + 0.5 → gross (σ price > 10 (Products))

A negative number and a boolean column:

π name, -1 → sentinel, FALSE → archived (Products)

Limitations

No hexadecimal. An exponent may be at most 1000 either way. A column named null is written with backticks, `null`, since the bare word is the NULL literal. A string cannot span lines except by writing \n.

Alternatives

The temporal literals for dates and times; an inline table for a set of rows; a $name parameter for a value supplied when the query runs.

See Also

date-literal, timestamp-literal, duration-literal, inline-table, is-null

Notes

TRUE and FALSE are case-insensitive, like every keyword.