Switch to SpeakEasy.net DSL

The Modular Manual Browser

Home Page
Manual: (Debian-5.0)
Apropos / Subsearch:
optional field

Class::Accessor(3pm)  User Contributed Perl Documentation Class::Accessor(3pm)

         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(400000);  # 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 past-
       ing.  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 it's also simply not lazy, which is the real

       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

       What Makes This Different?

       What makes this module special compared to all the other method gener-
       ating 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 mem-
       ory than most other solutions which generate a new method for each


           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

       It takes an optional %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);

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

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



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

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

       See "Overriding autogenerated accessors" in CAVEATS AND TRICKS for



       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(42);    # BOOM!  Naughty you.



       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 it's here
       for orthoginality and because it's 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(42);      # 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 pro-
       vides default get() and set() methods which your class can override.
       They're detailed later.


       In Damian's Perl Best Practices book he recommends separate get and set
       methods with the prefix set_ and get_ to make it explicit what you
       intend to do.  If you want to create those accessor methods instead of
       the default ones, call:


       accessor_name_for / mutator_name_for

       You may have your own crazy ideas for the names of the accessors, so
       you can make those happen by overriding "accessor_name_for" and "muta-
       tor_name_for" in your subclass.  (I copied that idea from Class::DBI.)

       Modifying the behavior of the accessor

       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 $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 $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 $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.

                      Rate   Basic Average    Fast  Faster  Direct
         Basic    189150/s      --    -42%    -51%    -55%    -89%
         Average  327679/s     73%      --    -16%    -22%    -82%
         Fast     389212/s    106%     19%      --     -8%    -78%
         Faster   421646/s    123%     29%      8%      --    -76%
         Direct  1771243/s    836%    441%    355%    320%      --

                      Rate   Basic Average    Fast  Faster  Direct
         Basic    173769/s      --    -34%    -53%    -59%    -90%
         Average  263046/s     51%      --    -29%    -38%    -85%
         Fast     371158/s    114%     41%      --    -13%    -78%
         Faster   425821/s    145%     62%     15%      --    -75%
         Direct  1699081/s    878%    546%    358%    299%      --

       Class::Accessor::Fast is faster than methods written by an average pro-
       grammer (where "average" is based on Schwern's example code).

       Class::Accessor is slower than average, but more flexible.

       Class::Accessor::Faster is even faster than Class::Accessor::Fast.  It
       uses an array internally, not a hash.  This could be a good or bad fea-
       ture depending on your point of view.

       Direct hash access is, of course, much faster than all of these, but it
       provides no encapsulation.

       Of course, it's not as simple as saying "Class::Accessor is slower than
       average".  These are benchmarks for a simple accessor.  If your acces-
       sors 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::Acces-
       sor and the ones you write will be roughly the same speed.

       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";


           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.

       Don't make a field called DESTROY

       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".

       Overriding autogenerated accessors

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

       There's a subtle problem in the last example, and it's 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 2007 Marty Pauley <marty+perlATkasei.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 <schwernATpobox.com>


       Liz and RUZ for performance tweaks.

       Tels, for his big feature request/bug report.


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

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

perl v5.8.8                       2007-07-11              Class::Accessor(3pm)