Rules / Operators / structural

<=> Skeleton

infix <=>

<=>
infixstructural 14/27non-assocRaku++ runs this — matches Rakudo on all 7docs.raku.org ↗

The rules for <=> have not been written yet. Everything below is extracted rather than authored: the official documentation supplies the classification, and the interpreters were asked directly about the rest.

What is established #

Classification #

<=> is parsed as an infix at the structural precedence level, non-associative.

That is level 14 of 27 on the ladder, counting from the tightest. Operators sharing a level bind equally; grouping among them is decided by associativity alone.

Shares its precedence level with #

does, but, cmp, coll, unicmp, leg, .., ..^, ^.., ^..^.

These all bind equally tightly, so an expression mixing them groups by associativity alone.

What the interpreter said #

Raku++ parses this spelling. Asked to compile:

my $a = 6; my $b = 3; sink($a <=> $b)

…it got through the parser. On its own that establishes very little: it says nothing about whether the construct executes, let alone whether the result is right. Where the behaviour table below exists, that is the stronger evidence — those expressions were actually run.

Covered upstream #

The official documentation has a section for this construct: docs.raku.org. Until the rules here are written, that is the better reference.

Observed behaviour #

Each expression below was executed by both interpreters when this page was built; the results are recorded, not written. The operands are chosen to cross the boundaries that catch people out: string against number, a list where a scalar was meant, a boolean, an undefined value, and an exact rational.

Raku++ and Rakudo agree on every row below.

ExpressionRaku++Rakudo
1 <=> 2Order | Less
"a" <=> "b"rejected by both — Cannot convert string to number: base-10 number must begin with valid digits or '.' in 'a'
1 <=> "2"Order | Less
(1, 2) <=> (3, 4, 5)Order | Less
True <=> FalseOrder | More
Nil <=> 1Order | Less
1/2 <=> 1/3Order | More