Variables

Variables in Raku

Variables are symbolic names for values or containers. Variable declarations or assignment of values may create a container on the fly. Variable names can start with or without a special character called a sigil, followed optionally by a second special character named twigil and then an identifier.

Sigils

There are four sigils. The scalar-sigil $, the positional-sigil @, the associative-sigil % and the callable-sigil &.

Sigils provide a link between syntax, the type system and containers. They provide a shortcut for the most common type constraints when declaring variables and serve as markers for string interpolation. The positional-sigil and the associative-sigil provide type constraint that enforce a base type subscripts require to know what methods to dispatch to. The callable-sigil does the same for function calls. The latter also tells the compiler where parentheses for calls can be omitted. The positional and associative-sigil also simplify assignment by flattening by default.

Sigil Type constraint Default type Assignment Examples
$ Mu (no type constraint) Any item Int, Str, Array, Hash
@ Positional Array list List, Array, Range, Buf
% Associative Hash list Hash, Map, Pair
& Callable Callable item Sub, Method, Block, Routine

Examples:

my $square = 9 ** 2;
my @array  = 123;   # Array variable with three elements 
my %hash   = London => 'UK'Berlin => 'Germany';

The type to which the variable will be bound can be set with is in the declaration of the variable. Assuming we have a FailHash class:

class FailHash is Hash {
    has Bool $!final = False;
    multi method AT-KEY ( ::?CLASS:D: Str:D \key ){
        fail X::OutOfRange.new(:what("Hash key"), :got(key),
          :range(self.keys)) if $!final && !self.EXISTS-KEY(key);
        callsame  # still not final, so do normal action from Hash 
    }
 
    method finalize() {
        $!final = True
    }
}

One can then define a %h variable of this type using is:

my %h is FailHash = oranges => "round"bananas => "bendy";

And then run the following code:

say %h<oranges>;
# OUTPUT: «round␤» 
%h.finalize;
say %h<cherry>;
CATCH { default { put .^name''.Str } }
# OUTPUT: «X::OutOfRange: Hash key out of range. Is: cherry, should be in (oranges bananas)» 

For information on variables without sigils, see sigilless variables.

Item and list assignment

There are two types of variable assignment, item assignment and list assignment.

An item assignment copies a single value from the right-hand side into a Scalar variable on the left. An assignment to anything other than a simple Scalar variable is parsed as a list assignment. A list assignment leaves the choice of what the assignment operation entails to the variable on the left. For example, Array variables (@ sigil) empty themselves on list assignment, and then iteratively copy all values from the right-hand side into themselves as elements.

The two types of assignment both use the equal sign = as their operator and are both right associative, but differ in operator precedence: item assignment has a higher precedence level (level: Item assignment) than list assignment (level: List prefix). In situations in which a comma-separated list of elements is assigned, these precedences should in particular be contrasted with that of the comma operator , which sits in between. So without any list-delimiting parentheses (or other construct to hold the list's elements together), item assignment will only assign the first element of the specified list, and not the full list.

In an assignment expression the context of the left-hand side determines whether an = means item or list assignment. As mentioned, item assignment is restricted to simple Scalar variables. Accordingly, assignment to a Scalar container (scalar-context) triggers item assignment, unless the Scalar is explicitly put in list-context by surrounding parentheses ( ):

my $a;
$a = 1,2,3;         # item assignment to Scalar 
say $a;             # OUTPUT: «1␤» ( '=' has higher precedence than ',' ) 
 
my $b = 1,2,3;      # item assignment to Scalar (same as preceding example) 
say $b;             # OUTPUT: «1␤» 
 
my $c;
($c= 4,5,6;       # list assignment to Scalar; '( )' is list-contextualizer 
say $c;             # OUTPUT:  «(4,5,6)␤» 
 
(my $d= 4,5,6;    # list assignment to Scalar (same as preceding example) 
say $d;             # OUTPUT:  «(4,5,6)␤»

Assignment to a List container (list-context) always triggers list assignment:

my @e;
@e = 7,8,9;                     # list assignment to Array 
say @e;                         # OUTPUT:  «[7,8,9]␤» 
 
my $f;
($f,) = 7,8,9;                  # list assignment to List with one element 
say $f;                         # OUTPUT:  «7␤» 
say ( ($f,) ).VAR.^name;        # OUTPUT:  «List␤» 
 
# ATTENTION: special declaration syntax! 
my ($g= 7,8,9;                # list assignment to List with one element 
say $g;                         # OUTPUT:  «7␤» 
say ( ($g) ).VAR.^name       # OUTPUT:  «List␤»

The last two examples above are simple destructuring assignments that select the first item of the right-hand side list. See for a more elaborate discussion of destructuring assignments in the context of variable declarations the section on declaring a list of variables with lexical or package scope.

Chained assignments are parsed having regard to the precedence of the assignment operators and, where applicable, their right associativity. For instance, in the example below there is one chained assignment statement comprising two assignment operators. The assignment to @array is a list assignment having a lower precedence than the item assignment to the Scalar variable $num. The assignment expression involving the item assignment to the Scalar variable $num is thus evaluated first. It returns the assigned value 42, which in turn forms part of the List (42, "str") constructed by the comma operator that also has a higher precedence than the list assignment. Finally, the List (42, "str") is list-assigned to @array:

my @array;
@array = my $num = 42"str";   # parsed as @array = ( (my $num = 42), "str ) 
say @array.perl;                # OUTPUT: «[42, "str"]␤» (an Array) 
say $num.perl;                  # OUTPUT: «42␤» (a Num)

Here's a variant:

my ( @foo$bar );
@foo = ($bar= 42"str";       # parsed as @foo = ( $bar = (42, "str") ) 
say $bar.perl;                   # OUTPUT: «$(42, "str")␤» (a List)# 
say @foo.perl;                   # OUTPUT: «[(42, "str"),]␤» (an Array)

In this case, the list contextualizer ( ) puts $bar in a list context, and thus triggers a list assignment to the Scalar variable $bar. This means that there are two chained list assignments, both having a lower precedence than the comma operator , that constructs the List (42, "str"). Due to their right associativity, the list assignment expression that is evaluated first is the assignment to $bar, which returns the assigned value $(42, "str"), i.e. a Scalar containing a two-element List. This value is in turn list-assigned to @array, such that it becomes a Array with a single element, namely a List.

See operators for more details on precedence and associativity.

Sigilless variables

Using the \ prefix, it's possible to create variables that do not have a sigil:

my \degrees = pi / 180;
my \θ       = 15 * degrees;

Note that sigilless variable do not have associated containers. This means degrees and θ, above, actually directly represent Nums. To illustrate, try assigning to one after you've defined it:

θ = 3# Dies with the error "Cannot modify an immutable Num" 

Sigilless variables do not enforce context, so they can be used to pass something on as-is:

sub logged(&f|args{
    say('Calling ' ~ &f.name ~ ' with arguments ' ~ args.perl);
    my \result = f(|args);
    #  ^^^^^^^ not enforcing any context here 
    say(&f.name ~ ' returned ' ~ result.perl);
    return |result;
}

Sigilless variables can also be used for binding. See Binding for more information.

Twigils

Twigils influence the scoping of a variable; however, they have no influence over whether the primary sigil interpolates. That is, if $a interpolates, so do $^a, $*a, $=a, $?a, $.a, etc. It only depends on the $.

Twigil Scope
none Based only on declarator
* Dynamic
? Compile-time variable
! Attribute (class member)
. Method (not really a variable)
< Index into match object (not really a variable)
^ Self-declared formal positional parameter
: Self-declared formal named parameter
= Pod variables
~ The sublanguage seen by the parser at this lexical spot

The * twigil

This twigil is used for dynamic variables which are looked up through the caller's, not through the outer, scope. Look at the example below.

Note: So far, if you use rakudo perl6, the example below cannot run correctly in the REPL. Please test it by copy-pasting it into a file, then run the file.

my $lexical   = 1;
my $*dynamic1 = 10;
my $*dynamic2 = 100;
 
sub say-all() {
    say "$lexical$*dynamic1$*dynamic2";
}
 
say-all();    # OUTPUT: 1, 10, 100 
 
{
    my $lexical   = 2;
    my $*dynamic1 = 11;
    $*dynamic2    = 101;
 
    say-all(); # OUTPUT: 1, 11, 101 
}
 
say-all();  # OUTPUT: 1, 10, 101 

The first time &say-all is called, it prints "1, 10, 100" just as one would expect. The second time though, it prints "1, 11, 101". This is because $lexical isn't looked up in the caller's scope but in the scope &say-all was defined in. The two dynamic variables are looked up in the caller's scope and therefore have the values 11 and 101. The third time &say-all is called $*dynamic1 isn't 11 anymore, but $*dynamic2 is still 101. This stems from the fact that we declared a new dynamic variable $*dynamic1 in the block and did not assign to the old variable as we did with $*dynamic2.

The dynamic variables differ from other variable types in that referring to an undeclared dynamic variable is not a compile time error but a runtime failure, so a dynamic variable can be used undeclared as long as it's checked for definedness or used in a boolean context before using it for anything else:

sub foo() {
    $*FOO // 'foo';
}
 
say foo# OUTPUT: «foo␤» 
 
my $*FOO = 'bar';
 
say foo# OUTPUT: «bar␤» 

Dynamic variables can have lexical scope when declared with my or package scope when declared with our. Dynamic resolution and resolution through symbol tables introduced with our are two orthogonal issues.

The ? twigil

Compile-time variables may be addressed via the ? twigil. They are known to the compiler and may not be modified after being compiled in. A popular example for this is:

say "$?FILE$?LINE"# OUTPUT: "hello.p6: 23" 
                      # if this is the line 23 of a 
                      # file named "hello.p6"

For a list of these special variables, see compile-time variables.

The ! twigil

Attributes are variables that exist per instance of a class. They may be directly accessed from within the class via !:

my class Point {
    has $.x;
    has $.y;
 
    method Str() {
        "($!x$!y)"
    }
}

Note how the attributes are declared as $.x and $.y but are still accessed via $!x and $!y. This is because in Raku all attributes are private and can be directly accessed within the class by using $!attribute-name. Raku may automatically generate accessor methods for you though. For more details on objects, classes and their attributes see object orientation.

The . twigil

The . twigil isn't really for variables at all. In fact, something along the lines of

my class Point {
    has $.x;
    has $.y;
 
    method Str() {
        "($.x$.y)" # note that we use the . instead of ! this time 
    }
}

just calls the methods x and y on self, which are automatically generated for you because you used the . twigil when the attributes were declared. Note, however, that subclasses may override those methods. If you don't want this to happen, use $!x and $!y instead.

The fact that the . twigil does a method call implies that the following is also possible:

class SaySomething {
    method a() { say "a"}
    method b() { $.a}
}
 
SaySomething.b# OUTPUT: «a␤»

For more details on objects, classes and their attributes and methods see object orientation.

The ^ twigil

The ^ twigil declares a formal positional parameter to blocks or subroutines; that is, variables of the form $^variable are a type of placeholder variable. They may be used in bare blocks to declare formal parameters to that block. So the block in the code

my @powers-of-three = 1,3,9100;
say reduce { $^b - $^a }0|@powers-of-three;
# OUTPUT: «61␤»

has two formal parameters, namely $a and $b. Note that even though $^b appears before $^a in the code, $^a is still the first formal parameter to that block. This is because the placeholder variables are sorted in Unicode order. If you have self-declared a parameter using $^a once, you may refer to it using only $a thereafter.

Although it is possible to use nearly any valid identifier as a placeholder variable, it is recommended to use short names or ones that can be trivially understood in the correct order, to avoid surprise on behalf of the reader.

Normal blocks and subroutines may also make use of placeholder variables but only if they do not have an explicit parameter list.

sub say-it    { say $^a} # valid 
sub say-it()  { say $^a} # invalid 
              { say $^a} # valid 
-> $x$y$x { say $^a} # invalid 

Placeholder variables cannot have type constraints or a variable name with a single upper-case letter (this is disallowed to enable catching some Perl5-isms).

The : twigil

The : twigil declares a formal named parameter to a block or subroutine. Variables declared using this form are a type of placeholder variable too. Therefore the same things that apply to variables declared using the ^ twigil also apply here (with the exception that they are not positional and therefore not ordered using Unicode order, of course). For instance:

say { $:add ?? $^a + $^b !! $^a - $^b }45 ) :!add
# OUTPUT: «-1␤»

See ^ for more details about placeholder variables.

The = twigil

The = twigil is used to access Pod variables. Every Pod block in the current file can be accessed via a Pod object, such as $=data, $=SYNOPSIS or =UserBlock. That is: a variable with the same name of the desired block and a = twigil.

  =begin code
  =begin Foo
  ...
  =end Foo
 
  #after that, $=Foo gives you all Foo-Pod-blocks
  =end code

You may access the Pod tree which contains all Pod structures as a hierarchical data structure through $=pod.

Note that all those $=someBlockName support the Positional and the Associative roles.

The ~ twigil

The ~ twigil is for referring to sublanguages (called slangs). The following are useful:

$~MAIN the current main language (e.g. Perl statements)
$~Quote the current root of quoting language
$~Quasi the current root of quasiquoting language
$~Regex the current root of regex language
$~Trans the current root of transliteration language
$~P5Regex the current root of the Perl 5 regex language

You augment these languages in your current lexical scope.

use MONKEY-TYPING;
augment slang Regex {  # derive from $~Regex and then modify $~Regex 
    token backslash:std<\Y> { YY };
}

Variable declarators and scope

Most of the time it's enough to create a new variable using the my keyword:

my $amazing-variable = "World";
say "Hello $amazing-variable!"# OUTPUT: «Hello World!␤»

However, there are many declarators that change the details of scoping beyond what Twigils can do.

Declarator Effect
my Introduces lexically scoped names
our Introduces package-scoped names
has Introduces attribute names
anon Introduces names that are private to the construct
state Introduces lexically scoped but persistent names
augment Adds definitions to an existing name
supersede Replaces definitions of an existing name

There are also two prefixes that resemble declarators but act on predefined variables:

Prefix Effect
temp Restores a variable's value at the end of scope
let Restores a variable's value at the end of scope if the block exits unsuccessfully
constant Declares that a container value is not going to change during its lifetime

The my declarator

Declaring a variable with my gives it lexical scope. This means it only exists within the current block. For example:

{
    my $foo = "bar";
    say $foo# OUTPUT: «"bar"␤» 
}
say $foo# Exception! "Variable '$foo' is not declared" 

This dies because $foo is only defined as long as we are in the same scope.

In order to create more than one variable with a lexical scope in the same sentence surround the variables with parentheses:

my ( $foo$bar );

see also Declaring a list of variables with lexical or package scope.

Additionally, lexical scoping means that variables can be temporarily redefined in a new scope:

my $location = "outside";
 
sub outer-location {
    # Not redefined: 
    say $location;
}
 
outer-location# OUTPUT: «outside␤» 
 
sub in-building {
    my $location = "inside";
    say $location;
}
 
in-building;    # OUTPUT: «inside␤» 
 
outer-location# OUTPUT: «outside␤»

If a variable has been redefined, any code that referenced the outer variable will continue to reference the outer variable. So here, &outer-location still prints the outer $location:

sub new-location {
    my $location = "nowhere";
    outer-location;
}
 
new-location# OUTPUT: «outside␤» 

To make new-location() print nowhere, make $location a dynamic variable using the * twigil. This twigil makes the compiler look up the symbol in the calling scope instead of the outer scope after trying the local scope.

my is the default scope for subroutines, so my sub x() {} and sub x() {} do exactly the same thing.

The our declarator

our variables work just like my variables, except that they also introduce an alias into the symbol table.

module M {
    our $Var;
    # $Var available here 
}
 
# Available as $M::Var here.

In order to create more than one variable with package scope, at the same time, surround the variables with parentheses:

our ( $foo$bar );

see also the section on declaring a list of variables with lexical or package scope.

Declaring a list of variables with lexical (my) or package (our) scope

It is possible to scope more than one variable at a time, but both my and our require variables to be placed into parentheses:

my  (@a,  $s,  %h);   # same as my @a; my $s; my %h; 
our (@aa$ss%hh);  # same as our @aa; our $ss; our %hh;

This can be used in conjunction with destructuring assignment. Any assignment to such a list will take the number of elements provided in the left list and assign corresponding values from the right list to them. Any missing elements are left will result in undefined values according to the type of the variables.

my (Str $aStr $bInt $c= <a b>;
say [$a$b$c].perl;
# OUTPUT: «["a", "b", Int]␤»

To destructure a list into a single value, create a list literal with one element by using ($var,). When used with a variable declarator, providing parentheses around a single variable is sufficient.

sub f { 1,2,3 };
my ($a= f;
say $a.perl;
# OUTPUT: «1␤»

To skip elements in the list use the anonymous state variable $.

my ($,$a,$,%h= ('a''b', [1,2,3], {:1th});
say [$a%h].perl;
# OUTPUT: «["b", {:th(1)}]␤»

The has declarator

has scopes attributes to instances of a class or role, and methods to classes or roles. has is implied for methods, so has method x() {} and method x() {} do the same thing.

See object orientation for more documentation and some examples.

The anon declarator

The anon declarator prevents a symbol from getting installed in the lexical scope, the method table and everywhere else.

For example, you can use it to declare subroutines which know their own name, but still aren't installed in a scope:

my %operations =
    half   => anon sub half($x{ $x / 2 },
    square => anon sub square($x{ $x * $x },
    ;
say %operations<square>.name;       # square 
say %operations<square>(8);         # 64

Since it is a declarator, it can be applied anywhere anything is declared, for instance for classes or even sigilless variables.

say anon class þ {};     # OUTPUT: «(þ)␤» 
say anon sub þ  { 42 };  # OUTPUT: «&þ␤»

Since these symbols are not installed in the scope, they can't be used by name. They are useful, however, if they need to be assigned to an external variable and they need to know their own name, but this can be retrieved using introspection.

my $anon-class = anon class {
    has $.bar;
    method equal( ::?CLASS $foo ) {
      return $foo.bar == $.bar;
    }
};
say $anon-class.new:3bar).equal$anon-class.new:3bar ) );

The state declarator

state declares lexically scoped variables, just like my. However, initialization happens exactly once the first time the initialization is encountered in the normal flow of execution. Thus, state variables will retain their value across multiple executions of the enclosing block or routine.

Therefore, the subroutine

sub a {
    state @x;
    state $l = 'A';
    @x.push($l++);
};
 
say a for 1..6;

will continue to increment $l and append it to @x each time it is called. So it will output:

[A]
[A B]
[A B C]
[A B C D]
[A B C D E]
[A B C D E F]

Since they have a lexical scope, they are tied to the block in which they are declared.

sub foo () {
  for 0..1 {
    state $foo = 1;
    say $foo++;
  }
};
foo;  # OUTPUT: «1␤2␤» 
foo;  # OUTPUT: «1␤2␤» 

In this case, a new state variable is created every time the block that runs the for loop is entered, which is why the state variable is reset in every call to foo.

This works per "clone" of the containing code object, as in this example:

({ state $i = 1$i++.say} xx 3).map: {$_(), $_()}# says 1 then 2 thrice 

Note that this is not a thread-safe construct when the same clone of the same block is run by multiple threads. Also remember that methods only have one clone per class, not per object.

As with my, a declaration of multiple state variables must be placed in parentheses which can be omitted for a single variable.

Many operators come with implicit binding which can lead to actions at a distance.

Use .clone or coercion to create a new container that can be bound to.

my @a;
my @a-cloned;
sub f() {
    state $i;
    $i++;
    @a\      .push: "k$i" => $i;
    @a-cloned.push: "k$i" => $i.clone;
};
 
f for 1..3;
say @a;        # OUTPUT: «[k1 => 3 k2 => 3 k3 => 3]␤» 
say @a-cloned# OUTPUT: «[k1 => 1 k2 => 2 k3 => 3]␤»

State variables are shared between all threads. The result can be unexpected.

sub code(){ state $i = 0say ++$i$i };
await
    start { loop { last if code() >= 5 } },
    start { loop { last if code() >= 5 } };
 
# OUTPUT: «1␤2␤3␤4␤4␤3␤5␤» 
# OUTPUT: «2␤1␤3␤4␤5␤» 
# many other more or less odd variations can be produced

The $ variable

In addition to explicitly declared named state variables, $ can be used as an anonymous state variable without an explicit state declaration.

say "1-a 2-b 3-c".subst(:g, /\d/{<one two three>[$++]});
# OUTPUT: «one-a two-b three-c␤»

Furthermore, state variables can be used outside of subroutines. You could, for example, use $ in a one-liner to number the lines in a file.

perl6 -ne 'say ++$ ~ " $_"' example.txt

Each reference to $ within a lexical scope is in effect a separate variable.

perl6 -e '{ say ++$; say $++  } for ^5'
# OUTPUT: «1␤0␤2␤1␤3␤2␤4␤3␤5␤4␤» 

That is why, if you need to reference the same $ variable (or, for that matter, any of the other anon state variables @ and %) more than once, a possible solution is to bind another variable to it, although in this example it would be more straightforward to just declare state $x and not use the magical/anonymous $ variable:

sub foo () {
    my $x := $;
    $x++;
    say $x;
    $x = $x + 1;
}
 
foo() for ^3# OUTPUT: «1␤3␤5␤» 

In general, it is better style to declare a named state variable in case you have to refer to it several times.

Note that the implicit state declarator is only applied to the variable itself, not the expression that may contain an initializer. If the initializer has to be called exactly once, the state declarator has to be provided.

for ^3 {       $ = .say } # OUTPUT: «0␤1␤2␤» 
for ^3 { state $ = .say } # OUTPUT: «0␤» 

The @ variable

Similar to the $ variable, there is also a Positional anonymous state variable @.

sub foo($x{
    say (@).push($x);
}
 
foo($_for ^3;
 
# OUTPUT: «[0] 
#          [0 1] 
#          [0 1 2]␤»

The @ here is parenthesized in order to disambiguate the expression from a class member variable named @.push. Indexed access doesn't require this disambiguation but you will need to copy the value in order to do anything useful with it.

sub foo($x{
    my $v = @;
    $v[$x= $x;
    say $v;
}
 
foo($_for ^3;
 
# OUTPUT: «[0] 
#          [0 1] 
#          [0 1 2]␤»

As with $, each mention of @ in a scope introduces a new anonymous array.

The % variable

In addition, there's an Associative anonymous state variable %.

sub foo($x{
    say (%).push($x => $x);
}
 
foo($_for ^3;
 
# OUTPUT: «{0 => 0} 
#          {0 => 0, 1 => 1} 
#          {0 => 0, 1 => 1, 2 => 2}␤»

The same caveat about disambiguation applies. As you may expect, indexed access is also possible (with copying to make it useful).

sub foo($x{
    my $v = %;
    $v{$x} = $x;
    say $v;
}
 
foo($_for ^3;
 
# OUTPUT: «{0 => 0} 
#          {0 => 0, 1 => 1} 
#          {0 => 0, 1 => 1, 2 => 2}␤»

As with the other anonymous state variables, each mention of % within a given scope will effectively introduce a separate variable.

The augment declarator

With augment, you can add methods, but not attributes, to existing classes and grammars, provided you activated the MONKEY-TYPING pragma first.

Since classes are usually our scoped, and thus global, this means modifying global state, which is strongly discouraged. For almost all situations, there are better solutions.

# don't do this 
use MONKEY-TYPING;
augment class Int {
    method is-answer { self == 42 }
}
say 42.is-answer;       # OUTPUT: «True␤»

(In this case, the better solution would be to use a function).

The temp prefix

Like my, temp restores the old value of a variable at the end of its scope. However, temp does not create a new variable.

my $in = 0# temp will "entangle" the global variable with the call stack 
            # that keeps the calls at the bottom in order. 
sub f(*@c{
    (temp $in)++;
     "<f>\n"
     ~ @c».indent($in).join("\n")
     ~ (+@c ?? "\n" !! "")
     ~ '</f>'
};
sub g(*@c{
    (temp $in)++;
    "<g>\n"
    ~ @c».indent($in).join("\n")
    ~ (+@c ?? "\n" !! "")
    ~ "</g>"
};
print g(g(f(g()), g(), f()));
 
# OUTPUT: «<g> 
#           <g> 
#            <f> 
#             <g> 
#             </g> 
#            </f> 
#            <g> 
#            </g> 
#            <f> 
#            </f> 
#           </g> 
#          </g>␤»

The let prefix

Restores the previous value if the block exits unsuccessfully. A successful exit means the block returned a defined value or a list.

my $answer = 42;
 
{
    let $answer = 84;
    die if not Bool.pick;
    CATCH {
        default { say "it's been reset :(" }
    }
    say "we made it 84 sticks!";
}
 
say $answer;

In the above case, if the Bool.pick returns true, the answer will stay as 84 because the block returns a defined value (say returns True). Otherwise the die statement will cause the block to exit unsuccessfully, resetting the answer to 42.

The constant prefix

The constant prefix declares that a container value is not going to change during its lifetime.

constant $pi2 = pi * 2;
$pi2 = 6# OUTPUT: «(exit code 1) Cannot assign to an immutable value␤ 

The value is assigned at compile time. Please check the section on constants in the Terms page for additional information.

Type constraints and initialization

Variables have a type constraint via the container they are bound to, which goes between the declarator and the variable name. The default type constraint is Mu. You can also use the trait of to set a type constraint.

    my Int $x = 42;
    $x = 'a string';
    CATCH { default { put .^name''.Str } }
    # OUTPUT: «X::TypeCheck::Assignment: Type check failed in assignment to $x; 
    expected Int but got Str ("a string")␤»

If a scalar variable has a type constraint but no initial value, it's initialized with the type object of the default value of the container it's bound to.

my Int $x;
say $x.^name;       # OUTPUT: «Int␤» 
say $x.defined;     # OUTPUT: «False␤»

Scalar variables without an explicit type constraint are typed as Mu but default to the Any type object.

Variables with the @ sigil are initialized with an empty Array; variables with the % sigil are initialized with an empty Hash.

The default value of a variable can be set with the is default trait, and re-applied by assigning Nil to it:

my Real $product is default(1);
say $product;                       # OUTPUT: «1␤» 
$product *= 5;
say $product;                       # OUTPUT: «5␤» 
$product = Nil;
say $product;                       # OUTPUT: «1␤»

Default defined variables pragma

To force all variables to have a definiteness constraint, use the pragma use variables :D. The pragma is lexically scoped and can be switched off with use variables :_.

use variables :D;
my Int $i;
# OUTPUT: «===SORRY!=== Error while compiling <tmp>␤Variable definition of type Int:D (implicit :D by pragma) requires an initializer ... 
my Int $i = 1# that works 
{ use variables :_my Int $i} # switch it off in this block 

Note that assigning Nil will revert the variable to its default value, which is often not a definite value and as such would fail the constraint:

use variables :D;
my Int $x = 42;
$x = Nil;
# OUTPUT: «Type check failed in assignment to $x; expected type Int:D cannot be itself…» 

As the name suggests, this pragma applies only to variables. To effect the same behavior on parameters, use the use parameters :D pragma (currently NYI in Rakudo).

Special variables

Raku attempts to use long, descriptive names for special variables. There are only three special variables that are extra short.

Pre-defined lexical variables

There are three special variables that are available in every block:

Variable Meaning
$_ topic variable
$/ regex match
$! exceptions

The $_ variable

$_ is the topic variable. It's the default parameter for blocks that do not have an explicit signature, so constructs like for @array { ... } and given $var { ... } bind the value or values of the variable to $_ by invoking the block.

for <a b c> { say $_ }  # sets $_ to 'a', 'b' and 'c' in turn 
say $_ for <a b c>;     # same, even though it's not a block 
given 'a'   { say $_ }  # sets $_ to 'a' 
say $_ given 'a';       # same, even though it's not a block

CATCH blocks set $_ to the exception that was caught. The ~~ smartmatch operator sets $_ on the right-hand side expression to the value of the left-hand side.

Calling a method on $_ can be shortened by leaving off the variable name:

.say;                   # same as $_.say

m/regex/ and /regex/ regex matches and s/regex/subst/ substitutions work on $_:

say "Looking for strings with non-alphabetic characters...";
for <ab:c d$e fgh ij*> {
    .say if m/<-alpha>/;
}
 
# OUTPUT: «Looking for strings with non-alphabetic characters... 
#          ab:c 
#          d$e 
#          ij*␤»

The $/ variable

$/ is the match variable. It stores the result of the last Regex match and so usually contains objects of type Match.

'abc 12' ~~ /\w+/;  # sets $/ to a Match object 
say $/.Str;         # OUTPUT: «abc␤»

The Grammar.parse method also sets the caller's $/ to the resulting Match object. For the following code:

use XML::Grammar# zef install XML 
XML::Grammar.parse("<p>some text</p>");
say $/;
 
# OUTPUT: «「<p>some text</p>」 
#           root => 「<p>some text</p>」 
#            name => 「p」 
#            child => 「some text」 
#             text => 「some text」 
#             textnode => 「some text」 
#           element => 「<p>some text</p>」 
#            name => 「p」 
#            child => 「some text」 
#             text => 「some text」 
#             textnode => 「some text」␤» 

Prior to the 6.d version, you could use $() shortcut to get the ast value from $/ Match if that value is truthy, or the stringification of the Match object otherwise.

'test' ~~ /.../;
# 6.c language only: 
say $(); # OUTPUT: «tes␤»; 
$/.make: 'McTesty';
say $(); # OUTPUT: «McTesty␤»;

This (non-)feature has been deprecated as of version 6.d.

Positional attributes

$/ can have positional attributes if the Regex had capture-groups in it, which are just formed with parentheses.

'abbbbbcdddddeffg' ~~ / a (b+) c (d+ef+) g /;
say $/[0]; # OUTPUT: «「bbbbb」␤» 
say $/[1]; # OUTPUT: «「dddddeff」␤»

These can also be accessed by the shortcuts $0, $1, $2, etc.

say $0# OUTPUT: «「bbbbb」␤» 
say $1# OUTPUT: «「dddddeff」␤»

To get all of the positional attributes, you can use $/.list or @$/. Before 6.d, you can also use the @() shortcut (no spaces inside the parentheses).

say @$/.join# OUTPUT: «bbbbbdddddeff␤» 
 
# 6.c language only: 
say @().join# OUTPUT: «bbbbbdddddeff␤»

This magic behavior of @() has been deprecated as of 6.d

Named attributes

$/ can have named attributes if the Regex had named capture-groups in it, or if the Regex called out to another Regex.

'I... see?' ~~ / \w+ $<punctuation>=[ <-[\w\s]>+ ] \s* $<final-word> = [ \w+ . ] /;
say $/<punctuation># OUTPUT: «「....」␤» 
say $/<final-word>;  # OUTPUT: «「see?」␤» 

These can also be accessed by the shortcut $<named>.

say $<punctuation># OUTPUT: «「....」␤» 
say $<final-word>;  # OUTPUT: «「see?」␤»

To get all of the named attributes, you can use $/.hash or %$/. Before 6.d language, you could also use the %() shortcut (no spaces inside the parentheses).

say %$/.join;       # OUTPUT: «"punctuation     ....final-word  see?"␤» 
 
# 6.c language only 
say %().join;       # OUTPUT: «"punctuation     ....final-word  see?"␤»

This behavior has been deprecated as of the 6.d version.

The $! variable

$! is the error variable. If a try block or statement prefix catches an exception, that exception is stored in $!. If no exception was caught, $! is set to the Any type object.

Note that CATCH blocks do not set $!. Rather, they set $_ inside the block to the caught exception.

Compile-time variables

All compile time variables have a question mark as part of the twigil. Being compile time they cannot be changed at runtime, however they are valuable in order to introspect the program. The most common compile time variables are the following:

$?FILE Which file am I in?
$?LINE Which line am I at? [indexed from 1]
::?CLASS Which class am I in?
%?RESOURCES The files associated with the "Distribution" of the current compilation unit.

$?FILE and $?LINE are also available from CallFrame as the file and line methods, respectively.

%?RESOURCES

A compile-time variable available within code of a Distribution.

It contains a hash that provides compile and runtime access to files associated with the Distribution of the current compilation unit. It is used to access a special storage for Distribution-wide static files (for example, examples of configuration files, templates etc).

To add a file to a distribution, it is placed under resources directory:

Module-Foo/
├── lib
│   └── Module
│       └── Foo.pm6
├── META6.json
├── README.md
└── resources
    └── images
        └── foo.jpg

Then a relative path (starting from the root directory of a distribution) to a file is specified under "resources" field in META6.json file:

"resources": [
    "images/foo.jpg"
]

Every resource file is added to an installed Distribution and is accessible using a Hash-like access to %?RESOURCES:

my $foo-IO = %?RESOURCES<images/foo.jpg>;          # gets an object you can slurp 
my $foo-IO = %?RESOURCES<images/foo.jpg>.absolute# gets an absolute path to a file 
my $foo-IO = %?RESOURCES<images/foo.jpg>.open;     # gets an opened IO::Handle to work with 

Note that paths and names of resource files can be mangled in an installed distribution, so do not rely on their values in any other case besides using them as keys for the %?RESOURCES variable.

The %?RESOURCES variable is not implemented as a plain Hash, but as an instance of the Distribution::Resources type, so do not expect to see all available resource files in a distribution by printing or other ways to inspect its value. Instead, use the API described above to access particular files.

Introspection compile-time variables

The following compile time variables allow for a deeper introspection:

$?PACKAGE Which package am I in?
$?MODULE Which module am I in?
$?CLASS Which class am I in? (as variable)
$?ROLE Which role am I in? (as variable)
$?TABSTOP How many spaces is a tab in a heredoc or virtual margin?
$?NL What a vertical newline "\n" means: LF, CR or CRLF
$?DISTRIBUTION The Distribution of the current compilation unit.

With particular regard to the $?NL, see the newline pragma.

Rakudo-specific compile-time variables

These variables are Rakudo specific, with all the corresponding caveats:

$?BITS Number of data-path bits of the platform the program is being compiled upon.

&?ROUTINE

The compile time variable &?ROUTINE provides introspection about which routine the program is actually within. It returns an instance of Sub attached to the current routine. It does support the method .name to obtain the name of the called routine, as well as .signature and others method related to Sub:

sub awesome-sub { say &?ROUTINE.name }
awesome-sub# OUTPUT: «awesome-sub␤»

It also allows also for recursion:

my $counter = 10;
sub do-work {
    say 'Calling myself other ' ~ $counter-- ~ ' times';
    &?ROUTINE() if ( $counter > 0 );
}
do-work;

&?BLOCK

The special compile variable ?&BLOCK behaves similarly to ?&ROUTINE but it allows to introspect a single block of code. It holds a Sub and allows for recursion within the same block:

for '.' {
    .Str.say when !.IO.d;
    .IO.dir()».&?BLOCK when .IO.d # lets recurse a little! 
}

$?DISTRIBUTION

$?DISTRIBUTION provides access to the Distribution of the current compilation unit. This gives module authors a way to reference other files in the distribution by their original relative path names, or to view the metadata (via the .meta method), without needing to know the underlying file structure (such as how CompUnit::Repository::Installation changes the file layout on installation).

unit module MyFoo;
 
sub module-version {
    say "MyFoo is version:";
    say $?DISTRIBUTION.meta<ver>;
}
 
sub module-source {
    say "MyFoo source code:";
    say $?DISTRIBUTION.content('lib/MyFoo.pm6');
}

Dynamic variables

All dynamically scoped variables have the * twigil, and their name is (conventionally) written in uppercase.

These variables are related to the arguments passed to a script.

$*ARGFILES

An IO::ArgFiles (an empty subclass of IO::CatHandle) that uses @*ARGS as source files, if it contains any files, or $*IN otherwise. When $*IN is used, its :nl-in, :chomp, :encoding, and :bin will be set on the IO::ArgFiles object.

As of the 6.d version, $*ARGFILES inside sub MAIN is always set to $*IN, even when @*ARGS is not empty. See the class documentation for examples and more context.

@*ARGS

@*ARGS is an array of Str containing the arguments from the command line.

&*ARGS-TO-CAPTURE

A dynamic variable available inside any custom ARGS-TO-CAPTURE subroutine that can be used to perform the default argument parsing. Takes the same parameters as are expected of the custom ARGS-TO-CAPTURE subroutine.

&*GENERATE-USAGE

A dynamic variable available inside any custom GENERATE-USAGE subroutine that can be used to perform the default usage message creation. Takes the same parameters as are expected of the custom GENERATE-USAGE subroutine.

Special filehandles: STDIN, STDOUT and STDERR

For more information about special filehandles please see also the Input and Output page and the IO::Special class. IO::Handle contains several examples of using $*IN for reading standard input.

Runtime environment

These dynamic variables contain information related to the environment the script or program is running in.

%*ENV

Operating system environment variables. Numeric values are provided as allomorphs.

$*REPO

This variable holds information about modules installed/loaded.

$*INIT-INSTANT

$*INIT-INSTANT is an Instant object representing program startup time. In particular, this is when the core code starts up, so the value of $*INIT-INSTANT may be a few milliseconds earlier than INIT now or even BEGIN now executed in your program.

$*TZ

$*TZ contains the system's local time zone offset, as the number of seconds from GMT.

$*CWD

It contains the Current Working Directory.

$*KERNEL

$*KERNEL contains a Kernel instance, the .gist of it being the current running kernel.

say $*KERNEL# OUTPUT: «linux (4.4.92.31.default)␤»

$*DISTRO

This object (of type Distro) contains information about the current operating system distribution. For instance:

say "Some sort of Windows" if $*DISTRO.is-win;

$*DISTRO.name takes a set of values that depend on the operating system. These names will vary with version and implementation, so you should double-check and test before using them in your programs; since these names are implementation defined and not in the specification, they could vary and change at any moment.

The $*DISTRO gist is displayed by using say:

say $*DISTRO# OUTPUT: «debian (9.stretch)␤»

This shows additional information on the operating system and version it's using, but as a matter of fact, this variable contains information which is useful to create portable programs, such as the path separator:

say $*DISTRO.perl;
# OUTPUT: «Distro.new(release => "42.3", is-win => Bool::False, 
#          path-sep => ":", name => "opensuse", 
#          auth => "https://www.opensuse.org/", version => v42.3, 
#          signature => Blob, desc => "2018-12-13T08:50:59.213619+01:00")␤»

$*VM

This variable contains the current virtual machine running the code, as well as additional information on the inner workings of aforementioned VM.

say $*VM.precomp-ext" "$*VM.precomp-target# OUTPUT: «moarvm mbc␤»

These two methods, for instance, will show the extension used in the precompiled bytecode scripts and the target used. This is what is found in the Moar Virtual Machine, but it could also vary with version and implementation. Other VM, such as Java, will show different values for them. $*VM.config includes all configuration values used to create the virtual machine, e.g.

say $*VM.config<versionmajor>"."$*VM.config<versionminor>;
# OUTPUT: «2018.11␤»

which are the version of the virtual machine, generally the same one as the one used in the interpreter and the overall Raku environment.

$*RAKU

This object of the Raku class contains information on the current implementation of the Raku language:

say $*RAKU.compiler.version# OUTPUT: «v2020.01␤»

but its gist includes the name of the language, followed by the major version of the compiler:

say $*RAKU# OUTPUT: «Raku (6.d)␤»

It stringifies to Raku:

$*RAKU.put# OUTPUT: «Raku␤»

$*PERL

For the foreseeable future, the same as $*RAKU. Will be deprecated at some point.

$*PID

Object containing an integer describing the current Process IDentifier (operating system dependent).

$*PROGRAM-NAME

This contains the path to the current executable as it was entered on the command line, or -e if perl was invoked with the -e flag.

$*PROGRAM

Contains the location (in the form of an IO::Path object) of the Raku program being executed.

&*EXIT

This is a Callable that contains the code that will be executed when doing an exit() call. Intended to be used in situations where Raku is embedded in another language runtime (such as Inline::Perl6 in Perl 5).

$*EXECUTABLE

Contains an IO::Path absolute path of the perl executable that is currently running.

$*EXECUTABLE-NAME

Contains the name of the Perl executable that is currently running. (e.g. perl6-p, perl6-m). Favor $*EXECUTABLE over this one, since it's not guaranteed that the perl executable is in PATH.

$*USAGE

This is the default usage message generated from the signatures of MAIN subs available from inside sub MAIN and sub USAGE. The variable is read-only.

$*USER

An Allomorph with information about the user that is running the program. It will evaluate to the username if treated as a string and the numeric user id if treated as a number.

$*GROUP

An Allomorph with the primary group of the user who is running the program. It will evaluate to the groupname only if treated as a string and the numeric group id if treated as a number.

$*HOMEDRIVE

Contains information about the "home drive" of the user that is running the program on Windows. It's not defined in other operating systems.

$*HOMEPATH

Contains information about the path to the user directory that is running the program on Windows. It's not defined in other operating systems.

$*HOME

Contains an IO::Path object representing the "home directory" of the user that is running the program. Uses %*ENV<HOME> if set.

On Windows, uses %*ENV<HOMEDRIVE> ~ %*ENV<HOMEPATH>. If the home directory cannot be determined, it will be Any.

$*SPEC

Contains the appropriate IO::Spec sub-class for the platform that the program is running on. This is a higher-level class for the operating system; it will return Unix, for instance, in the case of Linux (in the form of the IO::Spec class used for the current implementation).

$*TMPDIR

This is an IO::Path object representing the "system temporary directory" as determined by .tmpdir IO::Spec::* method.

$*THREAD

Contains a Thread object representing the currently executing thread.

$*SCHEDULER

This is a ThreadPoolScheduler object representing the current default scheduler.

By default this imposes a maximum of 64 threads on the methods .hyper, .race and other thread-pool classes that use that scheduler such as Promises or Supplys. This is, however, implementation, dependent and might be subject to change. To change the maximum number of threads, you can either set the environment variable RAKUDO_MAX_THREADS before running perl6 or create a scoped copy with the default changed before using them:

my $*SCHEDULER = ThreadPoolScheduler.newmax_threads => 128 );

This behavior is not tested in the spec tests and is subject to change.

$*SAMPLER

The current Telemetry::Sampler used for making snapshots of system state. Only available if Telemetry has been loaded.

$*USAGE

This is a Str object that holds value of the auto-generated USAGE message.

sub MAIN($a:$bUInt :$ehehe{
    say $*USAGE# OUTPUT: «Usage:␤  perl6.pl6 [-a=<Int>] [-b=<Str>] [--<opts>=...]» 
}

It is accessible only inside of MAIN sub.

Runtime variables

These variables affect the behavior of certain functions, and in some cases its value can be changed during runtime.

$*COLLATION

This is a Collation object that can be used to configure Unicode collation levels.

$*TOLERANCE

Variable used by the =~= operator, and any operations that depend on it, to decide if two values are approximately equal. Defaults to 1e-15.

$*DEFAULT-READ-ELEMS

Affects the number of bytes read by default by IO::Handle.read. Its default value is 65536.

Naming conventions

It is helpful to know our naming conventions in order to understand what codes do directly. However, there is not yet (and might never be) an official list of; still, we list several conventions that are widely held.