Literals and templates

An integer literal uses one of these radix forms.

Radix prefixes and hexadecimal letters are lowercase. An uppercase spelling produces TPZ2001 during parsing.

A leading - is a unary expression operator applied to a literal.

In a numeric digit run, _ separates two digits of the selected radix. The run starts and ends with a digit and cannot contain repeated separators. Parsing a malformed numeric token produces TPZ2001.

An integer literal has a literal type compatible with int. Values range from -9223372036854775808 through 9223372036854775807. The signed minimum is accepted as -9223372036854775808. An out-of-range integer value produces TPZ5001 during checking.

A float literal has decimal digits on both sides of .. Exponent notation is rejected as a malformed numeric token with TPZ2001 during parsing. A float literal has a literal type compatible with float.

Decimal float literals round to the nearest binary64 value. An exact tie selects the value with an even final significand bit. Overflow produces infinity. Underflow can produce a subnormal value or zero. Unary minus retains the sign of zero. Interpolation writes the infinities as inf and -inf, and the two zeros as 0.0 and -0.0. Decimal spelling still requires a point with digits on both sides; exponent spelling is rejected.

A duration literal attaches a lowercase ms, s, m suffix to a decimal integer magnitude. It denotes an ordinary Duration value, including in bindings, function arguments and results, parenthesized expressions, and concurrent(timeout: ...). Duration has no type arguments and does not implicitly convert to or from int or float. The same literal conversion and range checks apply in every expression position. Operations follow the Duration values rule.

A Duration stores from 0 through 18446744073709551615 milliseconds. An out-of-range duration literal produces TPZ5001 during checking.

Two Duration values support addition and subtraction. Multiplication takes a Duration and an int in either order, as in d * n and n * d. Construct a duration from an integer with 1ms * n. Duration has no division and no function for converting a number to Duration.

Eq compares millisecond counts for equality. Order compares those counts numerically. Display produces a decimal millisecond literal such as 1500ms, which reads back as the same value. toMillis() -> int returns the millisecond count. Duration has no JSON capability in 6.0.

An overflowing sum or product, a subtraction below zero, any negative integer multiplier, or a millisecond count outside the int range stops with runtime TPZ4004. These faults follow the fault cleanup rule.

A length literal attaches its lowercase unit directly to a decimal integer or decimal fraction. Radix and exponent notation are not length-literal forms.

The suffix determines the compiler-owned length type. Literal formation rejects an unrepresentable result with TPZ5001 during checking.

A Length literal is converted from its exact decimal value, including its sign. It rounds to the nearest multiple of the unit step below. An exact midpoint selects the even multiple.

The fractions are exact mathematical values. Write a decimal value followed immediately by its unit in source. A literal outside this interval produces TPZ5001 during checking. Equality and the public unit-conversion methods observe the rounded result.

An unescaped { begins interpolation containing an expression followed by }. An empty interpolation produces TPZ2001 during parsing.

Single-line and multiline strings use these escapes, including strings tagged r, sh, sql.

An unknown escape produces TPZ0004 during lexing for every tag. Write an escaped backslash when a literal backslash is intended.

A single-line string cannot cross a physical LF, CRLF, or CR line break, including inside interpolation. Crossing one produces TPZ0003 during lexing. An escaped carriage return or a carriage return in an inserted string remains character data.

Ordinary strings use the following type selection.

Interpolating a value whose type contains ByteBuffer in a string or tagged template produces TPZ5001 during checking.

In an ordinary string, each interpolation expression is evaluated once in source order. A nested interpolation completes before the next outer expression. The value then uses the standard renderer, with no automatic call to a user-defined show method. Inserted spaces and newlines remain unchanged.

A value whose type contains ByteBuffer produces TPZ5001 during checking, including inside arrays, records, unions, and nominal fields.

Literal text, inserted strings, and template text preserve their Unicode scalar sequence after the specified escape, physical-line-break, indentation, and blank-line processing. They do not apply Unicode normalization or remove Default_Ignorable_Code_Point scalars. Composed and decomposed text can therefore produce different UTF-8 bytes. A name inside an interpolation follows the identifier rules because it is part of the expression.

In a multiline string, one physical LF, CRLF, or CR line break immediately after the opening delimiter is removed from the value. CRLF is one line break.

When the closing delimiter follows spaces and tabs on its own line, that exact prefix is removed from each nonblank text line. Physical LF, CRLF, and CR each start a new line for this check. A nonblank line without that prefix produces TPZ0006 during lexing. The closing indentation is also removed.

Physical LF, CRLF, and CR line breaks in multiline text become LF. CRLF becomes one LF. A newline before the closing indentation remains in the value. Normalization applies to source text before escape decoding and value insertion. An escaped carriage return or a carriage return in an inserted string remains character data. Interpolation expressions use the source expression grammar.

A multiline source text line containing just spaces (U+0020) and tabs (U+0009) contributes an empty line. Its line break remains. This applies with or without closing indentation, in ordinary and tagged strings. Removal happens before escape decoding and value insertion. Escaped and inserted whitespace remains data. Nonblank indentation compares the exact source prefix, without expanding tabs. An interpolated value cannot supply missing indentation. A prefix mismatch produces TPZ0006 during lexing. Physical LF, CRLF, and CR start new text lines and become LF in the string value.

The template tags are r, sh, sql. Write the tag immediately before a single-line or triple-quoted string. A different immediately attached identifier produces TPZ2002 during parsing.

A registered tagged template is checked as a template value.

Registered tags construct template values from cooked text parts and expressions evaluated once in source order. There is one more text part than interpolation values. tag returns the tag as a string. parts returns an array value of the literal text segments under value semantics. Interpolated values remain separate and have no public accessor.

A template value cannot be passed where string is required. Use ordinary string interpolation to obtain its display. All tags use the shared string escape grammar. For example, r"\\d" preserves a regex backslash, and \{ and \} preserve literal braces.

Path.from follows the Path construction contract. That rule defines normalization, input failures and removed construction members.