Lexical::Accessor(3) true private attributes for Moose/Moo/Mouse


my $accessor = lexical_has identifier => (
is => 'rw',
isa => Int,
default => sub { 0 },

# or...
lexical_has identifier => (
is => 'rw',
isa => Int,
default => sub { 0 },
accessor => \$accessor,

# later...
say $self->$accessor; # says 0
$self->$accessor( 1 ); # setter
say $self->$accessor; # says 1


Lexical::Accessor generates coderefs which can be used as methods to access private attributes for objects.

The private attributes are stored inside-out, and do not add any accessors to the class' namespace, so are completely invisible to any outside code, including any subclasses. This gives your attribute complete privacy: subclasses can define a private (or even public) attribute with the same name as your private one and they will not interfere with each other.

Private attributes can not be initialized by Moose/Moo/Mouse constructors, but you can safely initialize them inside a "BUILD" sub.


"lexical_has $name?, %options"
This module exports a function lexical_has which acts much like Moose's "has" function, but sets up a private (lexical) attribute instead of a public one.

Because lexical attributes are stored inside-out, the $name is completely optional; however a name is recommended because it allows better error messages to be generated.

The lexical_has function supports the following options:

Moose/Mouse/Moo-style "ro", "rw", "rwp" and "lazy" values are supported. These control what sort of coderef is returned by the "lexical_has" function itself.

   my $reader            = lexical_has "foo" => (is => "ro");
   my $accessor          = lexical_has "foo" => (is => "rw");
   my ($reader, $writer) = lexical_has "foo" => (is => "rwp");

If generating more than one method it is probably clearer to pass in scalar references to the "reader", "writer", etc methods, rather than relying on the return value of the "lexical_has" function.

"reader", "writer", "accessor", "predicate", "clearer"
These accept scalar references. The relevant coderefs will be plonked into them:

   my ($get_foo, $set_foo);
   lexical_has foo => (
      reader      => \$get_foo,
      writer      => \$set_foo,
"default", "builder", "lazy"
Lazy defaults and builders are allowed. Eager (non-lazy) defaults and builders are currently disallowed. (Use a "BUILD" sub to set private attribute values at object construction time.)

The default may be either a non-reference value, or a coderef which will be called as a method to return the value.

Builders probably make less sense than defaults because they require a method in the class' namespace. The builder may be a method name, or the special value '1' which will be interpreted as meaning the attribute name prefixed by ``_build_''. If a coderef is provided, this is automatically installed into the class' namespace with the ``_build_'' prefix. (This last feature requires Sub::Name.)

A type constraint for the attribute. Moo-style coderefs are accepted (including those generated by MooX::Types::MooseLike), as are Moose::Meta::TypeConstraint/MooseX::Types objects, and Mouse::Meta::TypeConstraint/MouseX::Types objects, and of course Type::Tiny type constraints.

String type constraints may also be accepted, but only if Type::Utils is installed. (String type constraints are reified using "dwim_type".)

As an alternative to "isa", you can provide a role name in the "does" option.
A coderef or Type::Coercion object is accepted.

If the special value '1' is provided, the type constraint object is consulted to find the coercion. (This doesn't work for coderef type constraints.)

A method name or coderef to trigger when a new value is set.
Boolean indicating whether to automatically dereference array and hash values if called in list context.
Must be "undef" if provided at all.
Must be false if provided at all.
Boolean. Makes the setter weaken any references it is called with.
Delegates methods. Has slightly different syntax to Moose's option of the same name - is required to be an arrayref of pairs such that each pair is a scalar ref followed by a method name, a coderef, or an arrayref (where the first element is a method name or coderef and subsequent elements are curried arguments).

   my ($get, $post);
   lexical_has ua => (
      isa      => 'HTTP::Tiny',
      default  => sub { 'HTTP::Tiny'->new },
      handles  => [
         \$get   => 'get',
         \$post  => 'post',
   # later...
   my $response = $self->$get('http://example.net/');

"initializer", "traits", "lazy_build"
Not currently implemented. Providing any of these options throws an error.
"documentation", "definition_context"
Don't do anything, but are allowed; effectively inline comments.

Class Methods

This function may also be called as a class method.

Comparison (benchmarking, etc)

Lexical::Accessor is almost three times faster than MooX::PrivateAttributes, and almost twenty time faster than MooseX::Privacy. I'd also argue that it's a more ``correct'' implementation of private accessors as (short of performing impressive PadWalker manipulations), the accessors generated by this module are completely invisible to subclasses, method dispatch, etc.

Compared to the usual Moose convention of using a leading underscore to indicate a private method (which is a very loose convention; it is quite common for subclasses to override such methods!), Lexical::Accessor clearly offers much better method privacy. There should be little performance hit from using lexical accessors compared to normal Moose accessors. (However they are nowhere near the speed of the XS-powered accessors that Moo sometimes uses and Mouse usually uses.)

See also: "examples/benchmark.pl" bundled with this release.


IRC: support is available through in the #moops channel on irc.perl.org <http://www.irc.perl.org/channels.html>.


Toby Inkster <[email protected]>.


This software is copyright (c) 2013-2014 by Toby Inkster.

This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.