Skip to content
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
660 lines (467 sloc) 20.6 KB
=begin pod :kind("Type") :subkind("class") :category("basic")
=TITLE class Mu
=SUBTITLE The root of the Perl 6 type hierarchy.
class Mu { }
The root of the Perl 6 type hierarchy. For the origin of the name, see
L<>. One can also say that
there are many undefined values in Perl 6, and C<Mu> is the I<most undefined>
Note that most classes do not derive from C<Mu> directly, but rather from
=head1 Methods
=head2 method iterator
Defined as:
method iterator(--> Iterator)
Coerces the invocant to a C<list> by applying its L«C<.list>|/routine/list»
method and uses L«C<iterator>|/type/Iterable#method_iterator» on it.
my $it = Mu.iterator;
say $it.pull-one; # OUTPUT: «(Mu)␤»
say $it.pull-one; # OUTPUT: «IterationEnd␤»
=head2 method defined X<|method,defined>
Declared as
multi method defined( --> Bool:D)
Returns C<False> on a type object, and C<True> otherwise.
say Int.defined; # OUTPUT: «False␤»
say 42.defined; # OUTPUT: «True␤»
A few types (like L<Failure|/type/Failure>) override C<defined> to return
C<False> even for instances:
sub fails() { fail 'oh noe' };
say fails().defined; # OUTPUT: «False␤»
=head2 routine defined
Declared as
multi sub defined(Mu --> Bool:D)
invokes the C<.defined> method on the object and returns its result.
=head2 routine isa
multi method isa(Mu $type --> Bool:D)
multi method isa(Str:D $type --> Bool:D)
Returns C<True> if the invocant is an instance of class C<$type>, a subset type
or a derived class (through inheritance) of C<$type>.
L<C<does>|/routine/does#(Mu)_routine_does> is similar, but includes roles.
my $i = 17;
say $i.isa("Int"); # OUTPUT: «True␤»
say $i.isa(Any); # OUTPUT: «True␤»
role Truish {};
my $but-true = 0 but Truish;
say $but-true.^name; # OUTPUT: «Int+{Truish}␤»
say $but-true.does(Truish); # OUTPUT: «True␤»
say $but-true.isa(Truish); # OUTPUT: «False␤»
=head2 routine does
method does(Mu $type --> Bool:D)
Returns C<True> if and only if the invocant conforms to type C<$type>.
=for code
my $d ='2016-06-03');
say $d.does(Dateish); # True (Date does role Dateish)
say $d.does(Any); # True (Date is a subclass of Any)
say $d.does(DateTime); # False (Date is not a subclass of DateTime)
Unlike L<C<isa>|/routine/isa#(Mu)_routine_isa>, which
returns C<True> only for superclasses, C<does> includes both superclasses and
=for code :preamble<my $d ='2016-06-03');>
say $d.isa(Dateish); # OUTPUT: «False␤»
Using the smartmatch operator L<~~|/routine/~~> is a more idiomatic alternative.
my $d ='2016-06-03');
say $d ~~ Dateish; # OUTPUT: «True␤»
say $d ~~ Any; # OUTPUT: «True␤»
say $d ~~ DateTime; # OUTPUT: «False␤»
=head2 routine Bool
multi sub Bool(Mu --> Bool:D)
multi method Bool( --> Bool:D)
Returns C<False> on the type object, and C<True> otherwise.
Many built-in types override this to be C<False> for empty collections, the
empty L<string|/type/Str> or numerical zeros
say Mu.Bool; # OUTPUT: «False␤»
say; # OUTPUT: «True␤»
say [1, 2, 3].Bool; # OUTPUT: «True␤»
say [].Bool; # OUTPUT: «False␤»
say %( hash => 'full' ).Bool; # OUTPUT: «True␤»
say {}.Bool; # OUTPUT: «False␤»
say "".Bool; # OUTPUT: «False␤»
say 0.Bool; # OUTPUT: «False␤»
say 1.Bool; # OUTPUT: «True␤»
say "0".Bool; # OUTPUT: «True␤»
=head2 method Capture
Declared as:
method Capture(Mu:D: --> Capture:D)
Returns a L<Capture|/type/Capture> with named arguments corresponding to
invocant's public attributes:
class Foo {
has $.foo = 42;
has $.bar = 70;
method bar { 'something else' }
}.new.Capture.say; # OUTPUT: «\(:bar("something else"), :foo(42))␤»
=head2 method Str
multi method Str(--> Str)
Returns a string representation of the invocant, intended to be machine
readable. Method C<Str> warns on type objects, and produces the empty string.
say Mu.Str; # Use of uninitialized value of type Mu in string context.
my @foo = [2,3,1];
say @foo.Str # OUTPUT: «2 3 1␤»
=head2 routine gist
multi sub gist(+args --> Str)
multi method gist( --> Str)
Returns a string representation of the invocant, optimized for fast recognition
by humans. As such lists will be truncated at 100 elements. Use C<.perl> to get
all elements.
The default C<gist> method in C<Mu> re-dispatches to the L<perl|/routine/perl>
method for defined invocants, and returns the type name in parenthesis for type
object invocants. Many built-in classes override the case of instances to
something more specific that may truncate output.
C<gist> is the method that L<say|/routine/say> calls implicitly, so C<say
$something> and C<say $something.gist> generally produce the same output.
say Mu.gist; # OUTPUT: «(Mu)␤»
say; # OUTPUT: «␤»
=head2 routine perl
multi method perl(--> Str)
Returns a Perlish representation of the object (i.e., can usually be
re-evaluated with L<EVAL|/routine/EVAL> to regenerate the object). The exact
output of C<perl> is implementation specific, since there are generally many
ways to write a Perl expression that produces a particular value.
=head2 method item
method item(Mu \item:) is raw
Forces the invocant to be evaluated in item context and returns the value of it.
say [1,2,3].item.perl; # OUTPUT: «$[1, 2, 3]␤»
say %( apple => 10 ).item.perl; # OUTPUT: «${:apple(10)}␤»
say "abc".item.perl; # OUTPUT: «"abc"␤»
=head2 method self
method self(--> Mu)
Returns the object it is called on.
=head2 method clone
multi method clone(Mu:U: *%twiddles)
multi method clone(Mu:D: *%twiddles)
This method will clone type objects, or die if it's invoked with any argument.
=for code :skip-test<illustrates failure>
say Num.clone( :yes )
# OUTPUT: «(exit code 1) Cannot set attribute values when cloning a type object␤ in block <unit>␤␤»
If invoked with value objects, it creates a shallow clone of the invocant,
including shallow cloning of private attributes. Alternative values for
I<public> attributes can be provided via named arguments with names matching the
attributes' names.
=begin code
class Point2D {
has ($.x, $.y);
multi method gist(Point2D:D:) {
"Point($.x, $.y)";
my $p = => 2, y => 3);
say $p; # OUTPUT: «Point(2, 3)␤»
say $p.clone(y => -5); # OUTPUT: «Point(2, -5)␤»
=end code
Note that C<.clone> does not go the extra mile to shallow-copy C<@.> and C<%.>
sigiled attributes and, if modified, the modifications will still be available
in the original object:
=begin code
class Foo {
has $.foo is rw = 42;
has &.boo is rw = { say "Hi" };
has = <a b>;
has %.baz = <a b c d>;
my $o1 =;
with my $o2 = $o1.clone {
.foo = 70;
.bar = <Z Y>;
.baz = <Z Y X W>;
.boo = { say "Bye" };
# Hash and Array attribute modifications in clone appear in original as well:
say $o1;
# OUTPUT: « => 42, bar => ["Z", "Y"], baz => {:X("W"), :Z("Y")}, …␤»
say $o2;
# OUTPUT: « => 70, bar => ["Z", "Y"], baz => {:X("W"), :Z("Y")}, …␤»
$; # OUTPUT: «Hi␤»
$; # OUTPUT: «Bye␤»
=end code
To clone those, you could implement your own C<.clone> that clones the
appropriate attributes and passes the new values to C<Mu.clone>, for example,
via L«C<nextwith>|/language/functions#sub_nextwith».
=begin code
class Bar {
has $.quux;
has = <a b>;
has = <a b c d>;
method clone { nextwith :foo(@!foo.clone), :bar(%!bar.clone), |%_ }
my $o1 = :42quux );
with my $o2 = $o1.clone {
.foo = <Z Y>;
.bar = <Z Y X W>;
# Hash and Array attribute modifications in clone do not affect original:
say $o1;
# OUTPUT: « => 42, foo => ["a", "b"], bar => {:a("b"), :c("d")})␤»
say $o2;
# OUTPUT: « => 42, foo => ["Z", "Y"], bar => {:X("W"), :Z("Y")})␤»
=end code
The C<|%_> is needed to slurp the rest of the attributes that would have been
copied via shallow copy.
=head2 method new
multi method new(*%attrinit)
multi method new($, *@)
Default method for constructing (create + initialize) new objects
of a class. This method expects only named arguments which are then
used to initialize attributes with accessors of the same name.
Classes may provide their own C<new> method to override this default.
C<new> triggers an object construction mechanism that calls submethods named
C<BUILD> in each class of an inheritance hierarchy, if they exist. See
L<the documentation on object construction|/language/objects#Object_Construction>
for more information.
=head2 method bless
method bless(*%attrinit --> Mu:D)
Low-level object construction method, usually called from within C<new>,
implicitly from the default constructor, or explicitly if you create your own
constructor. C<bless> creates a new object of the same type as the invocant,
using the named arguments to initialize attributes and returns the created
It is usually invoked within custom C<new> method implementations:
=begin code
class Point {
has $.x;
has $.y;
multi method new($x, $y) {
self.bless(:$x, :$y);
my $p =, 1);
=end code
In this case we are declaring C<new> as a C<multi method> so that we can still
use the default constructor like this: C« x => 3, y => 8 )». In this
case we are declaring this C<new> method simply to avoid the extra syntax of
using pairs when creating the object. C<self.bless> returns the object, which is
in turn returned by C<new>.
However, in general, implementing a customized C<new> method might not be the best
way of initializing a class, even more so if the default constructor is
disabled, since it can make it harder to correctly initialize the class from a
subclass. For instance, in the above example, the C<new> implementation takes
two positional arguments that must be passed from the subclass to the superclass
in the exact order. That is not a real problem if it's documented, but take
into account C<bless> will eventually be calling C<BUILD>
in the class that is being instantiated. This might result in some unwanted
problems, like having to create a C<BUILD> submethod to serve it correctly:
=begin code
class Point {
has Int $.x;
has Int $.y;
multi method new($x, $y) {
self.bless(:$x, :$y);
class Point-with-ID is Point {
has Int $.ID is rw = 0;
submethod BUILD( *%args ) {
say %args; # OUTPUT: «{x => 1, y => 2}␤»
for self.^attributes -> $attr {
if $attr.Str ~~ /ID/ {
$attr.set_value( self, "*" ~ %args<x> ~ "-" ~ %args<y> ) ;
my $p =,2);
say $p.perl;
# OUTPUT: « => "*1-2", x => 1, y => 2)␤»
=end code
In this code, C<bless>, called within C<>, is eventually calling
C<BUILD> with the same parameters. We have to create a convoluted way of using
the C<$.ID> attribute using the meta-object protocol so that we can instantiate
it and thus serve that C<new> constructor, which can be called on
C<Point-with-ID> since it is a subclass.
We might have to use something similar if we want to instantiate superclasses.
C<bless> will help us with that, since it is calling across all the hierarchy:
=begin code
class Str-with-ID is Str {
my $.counter = 0;
has Int $.ID is rw = 0;
multi method new( $str ) {
self.bless( value => $str, ID => $.counter++ );
submethod BUILD( *%args ) {
for self.^attributes -> $attr {
if $attr.Str ~~ /ID/ {
$attr.set_value( self, %args<ID> ) ;
say"1.1,2e2").ID; # OUTPUT: «0␤»
my $enriched-str ="3,4");
say "$enriched-str, {$enriched-str.^name}, {$enriched-str.ID}";
# OUTPUT: «3,4, Str-with-ID, 1␤»
=end code
We are I<enriching> C<Str> with an auto-incrementing ID. We create a C<new>
since we want to initialize it with a string and, besides, we need to
instantiate the superclass. We do so using C<bless> from within C<new>. C<bless>
is going to call C<Str.BUILD>. It will *capture* the value it's looking for, the
pair C<value => $str> and initialize itself. But we have to initialize also the
properties of the subclass, which is why within C<BUILD> we use the previously
explained method to initialize C<$.ID> with the value that is in the C<%args>
variable. As shown in the output, the objects will be correctly initialized with
its ID, and will correctly behave as C<Str>, converting themselves in just the
string in the C<say> statement, and including the C<ID> property as required.
For more details see
L<the documentation on object construction|/language/objects#Object_Construction>.
=head2 method CREATE
method CREATE(--> Mu:D)
Allocates a new object of the same type as the invocant, without
initializing any attributes.
say Mu.CREATE.defined; # OUTPUT: «True␤»
=head2 method print
multi method print(--> Bool:D)
Prints value to C<$*OUT> after stringification using C<.Str> method without
adding a newline at end.
"abc\n".print; # RESULT: «abc␤»
=head2 method put
multi method put(--> Bool:D)
Prints value to C<$*OUT>, adding a newline at end, and if necessary,
stringifying non-C<Str> object using the C<.Str> method.
"abc".put; # RESULT: «abc␤»
=head2 method say
multi method say()
Will L<C<say>|/type/IO::Handle#method_say> to
L<standard output|/language/variables#index-entry-$*OUT>.
say 42; # OUTPUT: «42␤»
What C<say> actually does is, thus, deferred to the actual subclass. In L<most
cases|/routine/say> it calls L<C<.gist>|/routine/gist> on the object, returning
a compact string representation.
In non-sink context, C<say> will always return C<True>.
say (1,[1,2],"foo",Mu).map: so *.say ;
# OUTPUT: «1␤[1 2]␤foo␤(Mu)␤(True True True True)␤»
However, this behavior is just conventional and you shouldn't trust it for
your code. It's useful, however, to explain certain behaviors.
C<say> is first printing out in C<*.say>, but the outermost C<say> is printing
the C<True> values returned by the C<so> operation.
=head2 method ACCEPTS
multi method ACCEPTS(Mu:U: $other)
C<ACCEPTS> is the method that smartmatching with the L<infix ~~|/routine/~~>
operator and given/when invokes on the right-hand side (the matcher).
The C<Mu:U> multi performs a type check. Returns C<True> if C<$other> conforms
to the invocant (which is always a type object or failure).
say 42 ~~ Mu; # OUTPUT: «True␤»
say 42 ~~ Int; # OUTPUT: «True␤»
say 42 ~~ Str; # OUTPUT: «False␤»
Note that there is no multi for defined invocants; this is to allow
autothreading of L<junctions|/type/Junction>, which happens as a fallback
mechanism when no direct candidate is available to dispatch to.
=head2 method WHICH
multi method WHICH(--> ObjAt:D)
Returns an object of type L<ObjAt|/type/ObjAt> which uniquely identifies the
object. Value types override this method which makes sure that two equivalent
objects return the same return value from C<WHICH>.
say 42.WHICH eq 42.WHICH; # OUTPUT: «True␤»
=head2 method WHERE
method WHERE(--> Int)
Returns an C<Int> representing the memory address of the object.
=head2 method WHY
multi method WHY(--> Pod::Block::Declarator)
Returns the attached Pod::Block::Declarator.
For instance:
=for code :preamble<class Spell {};sub do-raw-magic(Spell) {};> :method<False>
#| Initiate a specified spell normally
sub cast(Spell $s) {
#= (do not use for class 7 spells)
say &cast.WHY;
# OUTPUT: «Initiate a specified spell normally␤(do not use for class 7 spells)␤»
See L<Pod declarator blocks|/language/pod#Declarator_blocks> for
details about attaching Pod to variables, classes, functions, methods, etc.
=head2 trait is export
multi sub trait_mod:<is>(Mu:U \type, :$export!)
Marks a type as being exported, that is, available to external users.
my class SomeClass is export { }
A user of a module or class automatically gets all the symbols imported that
are marked as C<is export>.
See L<Exporting and Selective Importing Modules|/language/modules#Exporting_and_selective_importing>
for more details.
=head2 method return
method return()
The method C<return> will stop execution of a subroutine or method, run all
relevant L<phasers|/language/phasers#Block_phasers> and provide invocant as a
return value to the caller. If a return
L<type constraint|/type/Signature#Constraining_return_types> is provided it will be
checked unless the return value is C<Nil>. A control exception is raised and
can be caught with L<CONTROL|/language/phasers#CONTROL>.
sub f { (1|2|3).return };
say f(); # OUTPUT: «any(1, 2, 3)␤»
=head2 method return-rw
Same as method L<C<return>|/type/Mu#method_return> except that C<return-rw> returns a writable
container to the invocant (see more details here: L<C<return-rw>|/language/control#return-rw>).
=head2 method emit
method emit()
Emits the invocant into the enclosing
L<supply|/language/concurrency#index-entry-supply_(on-demand)> or
L<react|/language/concurrency#index-entry-react> block.
react { whenever supply { .emit for "foo", 42, .5 } {
say "received {.^name} ($_)";
# received Str (foo)
# received Int (42)
# received Rat (0.5)
=head2 method take
method take()
Returns the invocant in the enclosing L<gather|/language/control#gather/take> block.
sub insert($sep, +@list) {
gather for @list {
FIRST .take, next;
take slip $sep, .item
say insert ':', <a b c>;
# OUTPUT: «(a : b : c)␤»
=head2 routine take
sub take(\item)
Takes the given item and passes it to the enclosing C<gather> block.
#| randomly select numbers for lotto
my $num-selected-numbers = 6;
my $max-lotto-numbers = 49;
gather for ^$num-selected-numbers {
take (1 .. $max-lotto-numbers).pick(1);
}.say; # six random values
=head2 routine take-rw
sub take-rw(\item)
Returns the given item to the enclosing C<gather> block, without introducing a new container.
my @a = 1...3;
sub f(@list){ gather for @list { take-rw $_ } };
for f(@a) { $_++ };
say @a;
# OUTPUT: «[2 3 4]␤»
=head2 method so
method so()
Evaluates the item in boolean context (and thus, for instance, collapses Junctions),
and returns the result.
It is the opposite of C<not>, and equivalent to the
L«C<?> operator|/language/operators#prefix_?».
One can use this method similarly to the English sentence: "If that is B<so>,
then do this thing". For instance,
my @args = <-a -e -b -v>;
my $verbose-selected = any(@args) eq '-v' | '-V';
if $ {
say "Verbose option detected in arguments";
} # OUTPUT: «Verbose option detected in arguments␤»
The C<$verbose-selected> variable in this case contains a
L<C<Junction>|/type/Junction>, whose value is C<any(any(False, False),
any(False, False), any(False, False), any(True, False))>. That is actually a
I<truish> value; thus, negating it will yield C<False>. The negation of that
result will be C<True>. C<so> is performing all those operations under the hood.
=head2 method not
method not()
Evaluates the item in boolean context (leading to final evaluation of Junctions, for instance),
and negates the result.
It is the opposite of C<so> and its behavior is equivalent to the
L«C<!> operator|/language/operators#prefix_!».
my @args = <-a -e -b>;
my $verbose-selected = any(@args) eq '-v' | '-V';
if $verbose-selected.not {
say "Verbose option not present in arguments";
} # OUTPUT: «Verbose option not present in arguments␤»
Since there is also a
L«prefix version of C<not>|/language/operators#prefix_not»,
this example reads better as:
my @args = <-a -e -b>;
my $verbose-selected = any(@args) eq '-v' | '-V';
if not $verbose-selected {
say "Verbose option not present in arguments";
} # OUTPUT: «Verbose option not present in arguments␤»
=end pod
# vim: expandtab softtabstop=4 shiftwidth=4 ft=perl6
You can’t perform that action at this time.