Class::Accessor - Automated accessor generation

  package Employee;
  use base qw(Class::Accessor);
  Employee->mk_accessors(qw(name role salary));

  # Meanwhile, in a nearby piece of code!
  # Class::Accessor provides new().
  my $mp = Foo->new({ name => "Marty", role => "JAPH" });

  my $job = $mp->role;  # gets $mp->{role}
  $mp->salary(4);  # sets $mp->{salary} = 400000 (I wish)

  # like my @info = @{$mp}{qw(name role)}
  my @info = $mp->get(qw(name role));

  # $mp->{salary} = 400000
  $mp->set('salary', 400000);

This module automagically generates accessors/mutators for your class.

Most of the time, writing accessors is an exercise in cutting and pasting. You usually wind up with a series of methods like this:

    sub name {
        my $self = shift;
        if(@_) {
            $self->{name} = $_[0];
        }
        return $self->{name};
    }

    sub salary {
        my $self = shift;
        if(@_) {
            $self->{salary} = $_[0];
        }
        return $self->{salary};
    }

  # etc...

One for each piece of data in your object. While some will be unique, doing value checks and special storage tricks, most will simply be exercises in repetition. Not only is it Bad Style to have a bunch of repetitious code, but its also simply not lazy, which is the real tragedy.

If you make your module a subclass of Class::Accessor and declare your accessor fields with mk_accessors() then you'll find yourself with a set of automatically generated accessors which can even be customized!

The basic set up is very simple:

    package My::Class;
    use base qw(Class::Accessor);
    My::Class->mk_accessors( qw(foo bar car) );

Done. My::Class now has simple foo(), bar() and car() accessors defined.

What makes this module special compared to all the other method generating modules (SEE ALSO)? By overriding the get() and set() methods you can alter the behavior of the accessors class-wide. Also, the accessors are implemented as closures which should cost a bit less memory than most other solutions which generate a new method for each accessor.

    my $obj = Class->new;
    my $obj = $other_obj->new;

    my $obj = Class->new(\%fields);
    my $obj = $other_obj->new(\%fields);

Class::Accessor provides a basic constructor. It generates a hash-based object and can be called as either a class method or an object method.

It takes an optional CW%fields hash which is used to initialize the object (handy if you use read-only accessors). The fields of the hash correspond to the names of your accessors, so...

    package Foo;
    use base qw(Class::Accessor);
    Foo->mk_accessors('foo');

    my $obj = Class->new({ foo => 42 });
    print $obj->foo;    # 42

however CW%fields can contain anything, new() will shove them all into your object. Don't like it? Override it.

    Class->mk_accessors(@fields);

This creates accessor/mutator methods for each named field given in CW@fields. Foreach field in CW@fields it will generate two accessors. One called field() and the other called _field_accessor(). For example:

    # Generates foo(), _foo_accessor(), bar() and _bar_accessor().
    Class->mk_accessors(qw(foo bar));

See Overriding autogenerated accessors in CAVEATS AND TRICKS for details.

  Class->mk_ro_accessors(@read_only_fields);

Same as mk_accessors() except it will generate read-only accessors (ie. true accessors). If you attempt to set a value with these accessors it will throw an exception. It only uses get() and not set().

    package Foo;
    use base qw(Class::Accessor);
    Class->mk_ro_accessors(qw(foo bar));

    # Let's assume we have an object $foo of class Foo...
    print $foo->foo;  # ok, prints whatever the value of $foo->{foo} is
    $foo->foo(4);    # BOOM!  Naughty you.

  Class->mk_wo_accessors(@write_only_fields);

Same as mk_accessors() except it will generate write-only accessors (ie. mutators). If you attempt to read a value with these accessors it will throw an exception. It only uses set() and not get().

NOTE I'm not entirely sure why this is useful, but I'm sure someone will need it. If you've found a use, let me know. Right now its here for orthoginality and because its easy to implement.

    package Foo;
    use base qw(Class::Accessor);
    Class->mk_wo_accessors(qw(foo bar));

    # Let's assume we have an object $foo of class Foo...
    $foo->foo(4);      # OK.  Sets $self->{foo} = 42
    print $foo->foo;    # BOOM!  Can't read from this accessor.

An accessor generated by Class::Accessor looks something like this:

    # Your foo may vary.
    sub foo {
        my($self) = shift;
        if(@_) {    # set
            return $self->set('foo', @_);
        }
        else {
            return $self->get('foo');
        }
    }

Very simple. All it does is determine if you're wanting to set a value or get a value and calls the appropriate method. Class::Accessor provides default get() and set() methods which your class can override. They're detailed later.

Rather than actually modifying the accessor itself, it is much more sensible to simply override the two key methods which the accessor calls. Namely set() and get().

If you -really- want to, you can override make_accessor().

    $obj->set($key, $value);
    $obj->set($key, @values);

set() defines how generally one stores data in the object.

override this method to change how data is stored by your accessors.

    $value  = $obj->get($key);
    @values = $obj->get(@keys);

get() defines how data is retreived from your objects.

override this method to change how it is retreived.

    $accessor = Class->make_accessor($field);

Generates a subroutine reference which acts as an accessor for the given CW$field. It calls get() and set().

If you wish to change the behavior of your accessors, try overriding get() and set() before you start mucking with make_accessor().

    $read_only_accessor = Class->make_ro_accessor($field);

Generates a subroutine refrence which acts as a read-only accessor for the given CW$field. It only calls get().

Override get() to change the behavior of your accessors.

    $read_only_accessor = Class->make_wo_accessor($field);

Generates a subroutine refrence which acts as a write-only accessor (mutator) for the given CW$field. It only calls set().

Override set() to change the behavior of your accessors.

If something goes wrong Class::Accessor will warn or die by calling Carp::carp or Carp::croak. If you don't like this you can override _carp() and _croak() in your subclass and do whatever else you want.

Class::Accessor does not employ an autoloader, thus it is much faster than you'd think. Its generated methods incur no special penalty over ones you'd write yourself.

Here are Schwern's results of benchmarking Class::Accessor, Class::Accessor::Fast, a hand-written accessor, and direct hash access.

  Benchmark: timing 500000 iterations of By Hand - get, By Hand - set, 
    C::A - get, C::A - set, C::A::Fast - get, C::A::Fast - set, 
    Direct - get, Direct - set...

  By Hand - get:  4 wallclock secs ( 5.09 usr +  0.00 sys =  5.09 CPU) 
                  @ 98231.83/s (n=500000)
  By Hand - set:  5 wallclock secs ( 6.06 usr +  0.00 sys =  6.06 CPU) 
                  @ 82508.25/s (n=500000)
  C::A - get:  9 wallclock secs ( 9.83 usr +  0.01 sys =  9.84 CPU) 
               @ 50813.01/s (n=500000)
  C::A - set: 11 wallclock secs ( 9.95 usr +  0.00 sys =  9.95 CPU) 
               @ 50251.26/s (n=500000)
  C::A::Fast - get:  6 wallclock secs ( 4.88 usr +  0.00 sys =  4.88 CPU) 
                     @ 102459.02/s (n=500000)
  C::A::Fast - set:  6 wallclock secs ( 5.83 usr +  0.00 sys =  5.83 CPU) 
                     @ 85763.29/s (n=500000)
  Direct - get:  0 wallclock secs ( 0.89 usr +  0.00 sys =  0.89 CPU) 
                 @ 561797.75/s (n=500000)
  Direct - set:  2 wallclock secs ( 0.87 usr +  0.00 sys =  0.87 CPU) 
                 @ 574712.64/s (n=500000)

So Class::Accessor::Fast is just as fast as one you'd write yourself while Class::Accessor is twice as slow, a price paid for flexibility. Direct hash access is about six times faster, but provides no encapsulation and no flexibility.

Of course, its not as simple as saying Class::Accessor is twice as slow as one you write yourself. These are benchmarks for the simplest possible accessor, if your accessors do any sort of complicated work (such as talking to a database or writing to a file) the time spent doing that work will quickly swamp the time spend just calling the accessor. In that case, Class::Accessor and the ones you write will tend to be just as fast.

Here's an example of generating an accessor for every public field of your class.

    package Altoids;

    use base qw(Class::Accessor Class::Fields);
    use fields qw(curiously strong mints);
    Altoids->mk_accessors( Altoids->show_fields('Public') );

    sub new {
        my $proto = shift;
        my $class = ref $proto || $proto;
        return fields::new($class);
    }

    my Altoids $tin = Altoids->new;

    $tin->curiously('Curiouser and curiouser');
    print $tin->{curiously};    # prints 'Curiouser and curiouser'

    # Subclassing works, too.
    package Mint::Snuff;
    use base qw(Altoids);

    my Mint::Snuff $pouch = Mint::Snuff->new;
    $pouch->strong('Blow your head off!');
    print $pouch->{strong};     # prints 'Blow your head off!'

Here's a simple example of altering the behavior of your accessors.

    package Foo;
    use base qw(Class::Accessor);
    Foo->mk_accessor(qw(this that up down));

    sub get {
        my $self = shift;

        # Note every time someone gets some data.
        print STDERR "Getting @_\n";

        $self->SUPER::get(@_);
    }

    sub set {
        my ($self, $key) = splice(@_, 0, 2);

        # Note every time someone sets some data.
        print STDERR "Setting $key to @_\n";

        $self->SUPER::set($key, @_);
    }

Class::Accessor has to do some internal wackiness to get its job done quickly and efficiently. Because of this, there's a few tricks and traps one must know about.

Hey, nothing's perfect.

This is bad. Since DESTROY is a magical method it would be bad for us to define an accessor using that name. Class::Accessor will carp if you try to use it with a field named DESTROY.

You may want to override the autogenerated accessor with your own, yet have your custom accessor call the default one. For instance, maybe you want to have an accessor which checks its input. Normally, one would expect this to work:

    package Foo;
    use base qw(Class::Accessor);
    Foo->mk_accessors(qw(email this that whatever));

    # Only accept addresses which look valid.
    sub email {
        my($self) = shift;
        my($email) = @_;

        if( @_ ) {  # Setting
            require Email::Valid;
            unless( Email::Valid->address($email) ) {
                carp("$email doesn't look like a valid address.");
                return;
            }
        }

        return $self->SUPER::email(@_);
    }

There's a subtle problem in the last example, and its in this line:

    return $self->SUPER::email(@_);

If we look at how Foo was defined, it called mk_accessors() which stuck email() right into Foo's namespace. There *is* no SUPER::email() to delegate to! Two ways around this... first is to make a pure base class for Foo. This pure class will generate the accessors and provide the necessary super class for Foo to use:

    package Pure::Organic::Foo;
    use base qw(Class::Accessor);
    Pure::Organic::Foo->mk_accessors(qw(email this that whatever));

    package Foo;
    use base qw(Pure::Organic::Foo);

And now Foo::email() can override the generated Pure::Organic::Foo::email() and use it as SUPER::email().

This is probably the most obvious solution to everyone but me. Instead, what first made sense to me was for mk_accessors() to define an alias of email(), _email_accessor(). Using this solution, Foo::email() would be written with:

    return $self->_email_accessor(@_);

instead of the expected SUPER::email().

Copyright 2005 Marty Pauley <marty+perl@kasei.com>

This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself. That means either (a) the GNU General Public License or (b) the Artistic License.

Michael G Schwern <schwern@pobox.com>

Liz, for performance tweaks.

Tels, for his big feature request/bug report.

Class::Accessor::Fast

These are some modules which do similar things in different ways Class::Struct, Class::Methodmaker, Class::Generate, Class::Class, Class::Contract

Class::DBI for an example of this module in use.