Arrow operator#

The single most heavily used operator in working Perl, after =. It does three things at three syntactic positions, and they all start by dereferencing something on the left.

Form	Means	Example
`$ref->[i]`	element of an arrayref	`$aref->[0]`
`$ref->{k}`	element of a hashref	`$href->{name}`
`$ref->(...)`	call a coderef	`$cref->(@args)`
`$obj->meth`	method call on an object	`$obj->print`
`$obj->meth(@a)`	method call with arguments	`$obj->set_x(42)`
`$class->meth`	class method call	`Foo::Bar->new`

The first three forms are dereferencing. The last three are method dispatch. Both are spelled with ->.

Dereferencing#

A reference is a scalar that points at a container or a coderef. You cannot index into a reference directly with [] or {}; you must dereference first, and -> is the canonical way:

my @arr  = (10, 20, 30);
my $aref = \@arr;

$arr[0]            # 10  - direct array access
$$aref[0]          # 10  - dereference, then subscript (old syntax)
$aref->[0]         # 10  - dereference and subscript in one move
$aref->[-1]        # 30  - negative indices work the same way

my %h    = (name => 'John');
my $href = \%h;

$h{name}           # 'John'
${$href}{name}     # 'John' - full deref + subscript
$href->{name}      # 'John' - same, idiomatic form

Coderef invocation#

my $f = sub { $_[0] * 2 };

$f->(5)            # 10  - invoke the coderef with argument 5
&$f(5)             # same - older syntax, still works

The ->() form is preferred in modern Perl. Empty argument list is $f->().

Arrow elision between subscripts#

When two subscripts directly follow each other on a deref chain, the second arrow can be omitted. The grammar implies it:

$grid[1]->[2]      # explicit
$grid[1][2]        # arrow elided - the only way most people write this

$tree->{children}[0]{name}    # parses with arrows between every pair
                              # - the chain is read as a sequence
                              # of subscripts on the same deref path

The first arrow (between a scalar variable and the first subscript) is required. Only arrows between consecutive subscripts may be elided.

$aref->[0]->[1]    # explicit
$aref->[0][1]      # elided second arrow - fine
$aref[0][1]        # WRONG - $aref[0] reads from @aref, not the
                   # arrayref in $aref. Different variable!

The first-arrow rule is the source of subtle bugs when refactoring $aref->[0] into a deeper expression - keep the leading arrow.

Method calls#

When the right side of -> is an identifier (or a parenthesised name expression), it is interpreted as a method call rather than a subscript:

$obj->print;                     # call $obj->print()
$obj->name;                      # accessor: returns $obj->name
$obj->set_name("Alice");         # mutator: sets and may return $obj
Foo::Bar->new(arg => 1);         # class method call

$obj->$method_name(@args);       # method name in a variable
$class->can('foo')               # can() - does method exist?

Method calls dispatch through the package the object is blessed into. The exact resolution path is governed by @ISA and the chosen MRO; see the OOP guide for the full story.

Indirect-object syntax - don’t use it#

Perl 5 historically allowed print $fh "foo" and new Foo @args without an arrow. These forms still parse but are context-sensitive in ways that surprise - the indirect-object syntax can resolve the new (or whatever) at the wrong time. The modern guidance is to always use -> for method calls:

my $obj = Foo->new(@args);                    # correct
my $obj = new Foo @args;                       # avoid (still parses)

print $fh "hello\n";                           # OK - print is special
$fh->print("hello\n");                         # also fine, prefer this
                                               # for IO::File-style handles

Precedence#

-> is row 2 of the precedence table - second only to terms and list operators. It binds tighter than every other operator, which is what you want: $h->{a} + $h->{b} parses as the sum of two hash lookups, not as a lookup of {a + $h->{b}}.