Class::Meta - Class automation, introspection, and data validation

INTERFACE

Class Methods

default_error_handler
handle_error
for_key
keys
clear

Constructors

new
add_constructor
add_attribute
add_method

Instance Methods

class
build

TO DO
BUGS
AUTHOR
SEE ALSO
COPYRIGHT AND LICENSE

NAME

Class::Meta - Class automation, introspection, and data validation

SYNOPSIS

Generate a class:

  package MyApp::Thingy;
  use strict;
  use Class::Meta;
  use Class::Meta::Types::String;
  use Class::Meta::Types::Numeric;

  BEGIN {

      # Create a Class::Meta object for this class.
      my $cm = Class::Meta->new( key => 'thingy' );

      # Add a constructor.
      $cm->add_constructor(
          name   => 'new',
          create => 1,
      );

      # Add a couple of attributes with generated methods.
      $cm->add_attribute(
          name     => 'uuid',
          authz    => Class::Meta::READ,
          type     => 'string',
          required => 1,
          default  => sub { Data::UUID->new->create_str },
      );
      $cm->add_attribute(
          name     => 'name',
          is       => 'string',
          required => 1,
          default  => undef,
      );
      $cm->add_attribute(
          name    => 'age',
          is      => 'integer',
          default => undef,
      );

     # Add a custom method.
      $cm->add_method(
          name => 'chk_pass',
          view => Class::Meta::PUBLIC,
      );
      $cm->build;
       }

Then use the class:

  use MyApp::Thingy;

  my $thingy = MyApp::Thingy->new;
  print "ID: ", $thingy->id, $/;
  $thingy->name('Larry');
  print "Name: ", $thingy->name, $/;
  $thingy->age(42);
  print "Age: ", $thingy->age, $/;

Or make use of the introspection API:

  use MyApp::Thingy;

  my $class = MyApp::Thingy->my_class;
  my $thingy;

  print "Examining object of class ", $class->package, $/;

  print "\nConstructors:\n";
  for my $ctor ($class->constructors) {
      print "  o ", $ctor->name, $/;
      $thingy = $ctor->call($class->package);
  }

  print "\nAttributes:\n";
  for my $attr ($class->attributes) {
      print "  o ", $attr->name, " => ", $attr->get($thingy), $/;
      if ($attr->authz >= Class::Meta::SET && $attr->type eq 'string') {
          $attr->get($thingy, 'hey there!');
          print "    Changed to: ", $attr->get($thingy), $/;
      }
  }

  print "\nMethods:\n";
  for my $meth ($class->methods) {
      print "  o ", $meth->name, $/;
      $meth->call($thingy);
  }

DESCRIPTION

Class::Meta provides an interface for automating the creation of Perl classes with attribute data type validation. It differs from other such modules in that it includes an introspection API that can be used as a unified interface for all Class::Meta-generated classes. In this sense, it is an implementation of the "Facade" design pattern.

JUSTIFICATION

One might argue that there are already too many class automation and parameter validation modules on CPAN. And one would be right. They range from simple accessor generators, such as Class::Accessor, to simple parameter validators, such as Params::Validate, to more comprehensive systems, such as Class::Contract and Class::Tangram. But, naturally, none of them could do exactly what I needed.

What I needed was an implementation of the "Facade" design pattern. Okay, this isn't a facade like the GOF meant it, but it is in the respect that it creates classes with a common API so that objects of these classes can all be used identically, calling the same methods on each. This is done via the implementation of an introspection API. So the process of creating classes with Class::Meta not only creates attributes and accessors, but also creates objects that describe those classes. Using these descriptive objects, client applications can determine what to do with objects of Class::Meta-generated classes. This is particularly useful for user interface code.

USAGE

Before we get to the introspection API, let's take a look at how to create classes with Class::Meta. Unlike many class automation modules for Perl, the classes that Class::Meta builds do not inherit from Class::Meta. This frees you from any dependencies on the interfaces that such a base class might compel. For example, you can create whatever constructors you like, and name them whatever you like.

I recommend that you create your Class::Meta classes in a BEGIN block. Although this is not strictly necessary, it helps to ensure that the classes you're building are completely constructed and ready to go by the time compilation has completed. Creating classes with Class::Meta is easy, using the Class::Meta object oriented interface. Here is an example of a very simple class:

  package MyApp::Dog;
  use strict;
  use Class::Meta;
  use Class::Meta::Types::Perl;

  BEGIN {

      # Create a Class::Meta object for this class.
      my $cm = Class::Meta->new( key => 'dog' );

      # Add a constructor.
      $cm->add_constructor(
          name   => 'new',
          create => 1,
      );

      # Add an attribute.
      $cm->add_attribute(
          name => 'tail',
          type => 'scalar',
      );

      # Add a custom method.
      $cm->add_method( name => 'wag' );
      $cm->build;
        }

  sub wag {
      my $self = shift;
      print "Wagging ", $self->tail;
  }

This simple example shows of the construction of all three types of objects supported by Class::Meta: constructors, attributes, and methods. Here's how it does it:

First we load Class::Meta and Class::Meta::Types::Perl. The latter module creates data types that can be used for attributes, including a "scalar" data type.
Second, we create a Class::Meta object. It's okay to create it within the BEGIN block, as it won't be needed beyond that. All Class::Meta classes have a key that uniquely identifies them across an application. If none is provided, the class name will be used, instead.
Next, we create a Class::Meta::Constructor object to describe a constructor method for the class. The create parameter to the add_constructor() method tells Class::Meta to create the constructor named "new()".
Then we call add_attribute() to create a single attribute, "tail". This is a simple scalar attribute, meaning that any scalar value can be stored in it. Class::Meta will create a Class::Meta::Attribute object that describes this attribute, and will also shortly create accessor methods for the attribute.
The add_method() method constructs a Class::Meta::Method object to describe any methods written for the class. In this case, we've told Class::Meta that there will be a wag() method.
And finally, we tell Class::Meta to build the class. This is the point at which all constructors and accessor methods will be created in the class. In this case, these include the new() constructor and a tail() accessor for the "tail" attribute. And finally, Class::Meta will install another method, my_class(). This method will return a Class::Meta::Class object that describes the class, and provides the complete introspection API.

Thus, the class the above code creates has this interface:

  sub my_class;
  sub new;
  sub tail;
  sub wag;

Data Types

By default, Class::Meta loads no data types. If you attempt to create an attribute without creating or loading the appropriate data type, you will get an error.

But I didn't want to leave you out in the cold, so I created a whole bunch of data types to get you started. Any of these will automatically be loaded by Class::Meta if it is used to create an attribute. They can also be loaded simply by useing the appropriate module. The modules are:

Class::Meta::Types::Perl: Typical Perl data types.
Class::Meta::Types::String

string

Attributes of this type must contain a string value. Essentially, this means anything other than a reference.
Class::Meta::Types::Boolean

boolean
bool

Attributes of this type store a boolean value. Implementation-wise, this means either a 1 or a 0.
Class::Meta::Types::Numeric: These data types are validated by the functions provided by Data::Types.

Other data types may be added in the future. See the individual data type modules for more information.

Accessors

Class::Meta supports the creation of three different types of attribute accessors: typical Perl single-method accessors, "affordance" accessors, and "semi-affordance" accessors. The single accessors are named for their attributes, and typically tend to look like this:

  sub tail {
      my $self = shift;
      return $self->{tail} unless @_;
      return $self->{tail} = shift;
  }

Although this can be an oversimplification if the data type has associated validation checks.

Affordance accessors provide at up to two accessors for every attribute: One to set the value and one to retrieve the value. They tend to look like this:

  sub get_tail { shift->{tail} }

  sub set_tail { shift->{tail} = shift }

These accessors offer a bit less overhead than the traditional Perl accessors, in that they don't have to check whether they're called to get or set a value. They also have the benefit of creating a psychological barrier to misuse. Since traditional Perl accessors can be created as read-only or write-only accessors, one can't tell just by looking at them which is the case. The affordance accessors make this point moot, as they make clear what their purpose is.

Semi-affordance accessors are similar to affordance accessors in that they provide at least two accessors for every attribute. However, the accessor that fetches the value is named for the attribute. Thus, they tend to look like this:

  sub tail { shift->{tail} }

  sub set_tail { shift->{tail} = shift }

To get Class::Meta's data types to create affordance accessors, simply pass the string "affordance" to them when you load them:

  use Class::Meta::Types::Perl 'affordance';

Likewise, to get them to create semi-affordance accessors, pass the string "semi-affordance":

  use Class::Meta::Types::Perl 'semi-affordance';

The boolean data type is the only one that uses a slightly different approach to the creation of affordance accessors: It creates three of them. Assuming you're creating a boolean attribute named "alive", it will create these accessors:

  sub is_alive      { shift->{alive} }
  sub set_alive_on  { shift->{alive} = 1 }
  sub set_alive_off { shift->{alive} = 0 }

Incidentally, I stole the term "affordance" from Damian Conway's "Object Oriented Perl," pp 83-84, where he borrows it from Donald Norman.

See Class::Meta::Type for details on creating new data types.

Introspection API

Class::Meta provides four classes the make up the introspection API for Class::Meta-generated classes. Those classes are:

Class::Meta::Class

Describes the class. Each Class::Meta-generated class has a single constructor object that can be retrieved by calling a class' my_class() class method. Using the Class::Meta::Class object, you can get access to all of the other objects that describe the class. The relevant methods are:

constructors: Provides access to all of the Class::Meta::Constructor objects that describe the class' constructors, and provide indirect access to those constructors.
attributes: Provides access to all of the Class::Meta::Attribute objects that describe the class' attributes, and provide methods for indirectly getting and setting their values.
methods: Provides access to all of the Class::Meta::Method objects that describe the class' methods, and provide indirect execution of those constructors.

Class::Meta::Constructor

Describes a class constructor. Typically a class will have only a single constructor, but there could be more, and client code doesn't necessarily know its name. Class::Meta::Constructor objects resolve these issues by describing all of the constructors in a class. The most useful methods are:

name: Returns the name of the constructor, such as "new".
call: Calls the constructor on an object, passing in the arguments passed to call() itself.

Class::Meta::Attribute

Describes a class attribute, including its name and data type. Attribute objects are perhaps the most useful Class::Meta objects, in that they can provide a great deal of information about the structure of a class. The most interesting methods are:

name: Returns the name of the attribute.
type: Returns the name of the attribute's data type.
required: Returns true if the attribute is required to have a value.
once: Returns true if the attribute value can be set to a defined value only once.
set: Sets the value of an attribute on an object.
get: Returns the value of an attribute on an object.

Class::Meta::Method

Describes a method of a class, including its name and context (class vs. instance). The relevant methods are:

name: The method name.
context: The context of the method indicated by a value corresponding to either Class::Meta::OBJECT or Class::Meta::CLASS.
call: Calls the method, passing in the arguments passed to call() itself.

Consult the documentation of the individual classes for a complete description of their interfaces.

INTERFACE

Class Methods

default_error_handler

  Class::Meta->default_error_handler($code);
  my $default_error_handler = Class::Meta->default_error_handler;

Sets the default error handler for Class::Meta classes. If no error_handler attribute is passed to new, then this error handler will be associated with the new class. The default default error handler uses Carp::croak() to handle errors.

Note that if other modules are using Class::Meta that they will use your default error handler unless you reset the default error handler to its original value before loading them.

handle_error

  Class::Meta->handle_error($err);

Uses the code reference returned by default_error_handler() to handle an error. Used internally Class::Meta classes when no Class::Meta::Class object is available. Probably not useful outside of Class::Meta unless you're creating your own accessor generation class. Use the handle_error() instance method in Class::Meta::Class, instead.

for_key

  my $class = Class::Meta->for_key($key);

Returns the Class::Meta::Class object for a class by its key name. This can be useful in circumstances where the key has been used to track a class, and you need to get a handle on that class. With the class package name, you can of course simply call $pkg->my_class; this method is the solution for getting the class object for a class key.

keys

  my @keys = Class::Meta->keys;

Returns the keys for all Class::Meta::Class objects. The order of keys is not guaranteed. In scalar context, this method returns an array reference containing the keys.

clear

  Class::Meta->clear;
  Class::Meta->clear($key);

Called without arguments, clear will remove all Class::Meta::Class objects from memory. Called with an argument, clear attempts to remove only that key from memory. Calling it with a non-existent key is a no-op.

In general, you probably won't want to use this method, except perhaps in tests, when you might need to do funky things with your classes.

Constructors

new

  my $cm = Class::Meta->new( key => $key );

Constructs and returns a new Class::Meta object that can then be used to define and build the complete interface of a class. The supported parameters are:

package

The package that defines the class. Defaults to the package of the code that calls new().

key

A key name that uniquely identifies a class within an application. Defaults to the value of the package parameter if not specified.

abstract

A boolean indicating whether the class being defined is an abstract class. An abstract class, also known as a "virtual" class, is not intended to be used directly. No objects of an abstract class should every be created. Instead, classes that inherit from an abstract class must be implemented.

trust

An array reference of key names or packages that are trusted by the class.

  trust => ['Foo::Bar', 'Foo::Bat'],

Trusted packages and the classes that inherit from them can retrieve trusted attributes and methods of the class. Trusted packages need not be Class::Meta classes. Trusted classes do not include the declaring class by default, so if you want the class that declares an attribute to be able to use trusted attribute accessors, be sure to include it in the list of trusted packages:

  trust => [__PACKAGE__, 'Foo::Bar', 'Foo::Bat'],

If you need to trust a single class, you may pass in the key name or package of that class rather than an array reference:

  trust => 'Foo::Bar',

class_class

The name of a class that inherits from Class::Meta::Class to be used to create all of the class objects for the class. Defaults to Class::Meta::Class.

constructor_class

The name of a class that inherits from Class::Meta::Constructor to be used to create all of the constructor objects for the class. Defaults to Class::Meta::Constructor.

attribute_class

The name of a class that inherits from Class::Meta::Attribute to be used to create all of the attribute objects for the class. Defaults to Class::Meta::Attribute.

method_class

The name of a class that inherits from Class::Meta::Method to be used to create all of the method objects for the class. Defaults to Class::Meta::Method.

error_handler

A code reference that will be used to handle errors thrown by the methods created for the new class. Defaults to the value returned by Class::Meta->default_error_handler.

add_constructor

  $cm->add_constructor(
      name   => 'construct',
      create => 1,
  );

Creates and returns a Class::Meta::Constructor object that describes a constructor for the class. The supported parameters are:

name: The name of the constructor. The name must consist of only alphanumeric characters or "_". Required.
create: When true, Class::Meta::Constructor will automatically create and install a constructor named for the name parameter. Defaults to true unless code is passed. In general you won't need to specify this parameter unless you've written your own constructor in the package, in which case you'll want to specify create => 0.
label: A label for the constructor. Generally used for displaying its name in a user interface. Optional.
desc: A description of the constructor. Possibly useful for displaying help text in a user interface. Optional.
code: You can implicitly define the constructor in your class by passing a code reference via the code parameter. Once build() is called, Kinetic::Meta::Constructor will install the constructor into the package for which the Class::Meta object was defined, and with the name specified via the name parameter. Note that if the constructor view is PRIVATE or PROTECTED, the constructor will be wrapped in extra code to envocde the view. Optional.
view: The visibility of the constructor. The possible values are defined by the following constants:
caller: A code reference that calls the constructor. Defaults to a code reference that calls a method with the name provided by the name attribute on the class being defined.

add_attribute

  $cm->add_attribute(
      name => 'tail',
      type => 'scalar',
  );

Creates and returns a Class::Meta::Attribute object that describes an attribute of the class. The supported parameters are:

name: The name of the attribute. The name must consist of only alphanumeric characters or "_". Required.
type
is: The data type of the attribute. See Data Types for some possible values for this parameter. If the type name corresponds to a data type in a package in the Class::Meta::Types namespae, that package will automatically be loaded and configured with Perl-style accessors, so that the data type can simply be used. Required. If both type and is are passed, is will be used.
required: A boolean value indicating whether the attribute is required to have a value. Defaults to false.
once: A boolean value indicating whether the attribute can be set to a defined value only once. Defaults to false.
label: A label for the attribute. Generally used for displaying its name in a user interface. Optional.
desc: A description of the attribute. Possibly useful for displaying help text in a user interface. Optional.
view: The visibility of the attribute. See the description of the view parameter to add_constructor for a description of its value.
authz: The authorization of the attribute. This value indicates whether it is read-only, write-only, read/write, or inaccessible. The possible values are defined by the following constants:
create: Indicates what type of accessor or accessors are to be created for the attribute.
context: The context of the attribute. This indicates whether it's a class attribute or an object attribute. The possible values are defined by the following constants:
default: The default value for the attribute, if any. This may be either a literal value or a code reference that will be executed to generate a default value.
override: If an attribute being added to a class has the same name as an attribute in a parent class, Class::Meta will normally throw an exception. However, in some cases you might want to override an attribute in a parent class to change its properties. In such a case, pass a true value to the override parameter to override the attribute and avoid the exception.

add_method

  $cm->add_method( name => 'wag' );

Creates and returns a Class::Meta::Method object that describes a method of the class. The supported parameters are:

name: The name of the method. The name must consist of only alphanumeric characters or "_".
label: A label for the method. Generally used for displaying its name in a user interface. Optional.
desc: A description of the method. Possibly useful for displaying help text in a user interface. Optional.
view: The visibility of the method. See the description of the view parameter to add_constructor for a description of its value.
code: You can implicitly define the method in your class by passing a code reference via the code parameter. Once build() is called, Kinetic::Meta::Method will install the method into the package for which the Class::Meta object was defined, and with the name specified via the name parameter.
context: The context of the method. This indicates whether it's a class method or an object method. See the description of the context parameter to add_attribute for a description of its value.
caller: A code reference that calls the method. This code reference will be be used by the call() method of Class::Meta::Method to execute the method on behalf of an object. Defaults to a code reference that calls a method with the name provided by the name attribute on the class being defined.
args: A description of the arguments to the method. This can be anything you like, but I recommend something like a string for a single argument, an array reference for a list of arguments, or a hash reference for parameter arguments.
returns: A string describing the return value or values of the method.

Instance Methods

class

  my $class = $cm->class;

Returns the instance of the Class::Meta::Class object that will be used to provide the introspection API for the class being generated.

build

  $cm->build;

Builds the class defined by the Class::Meta object, including the my_class() class method, and all requisite constructors and accessors.

TO DO

Make class attribute accessors behave as they do in Class::Data::Inheritable.
Modify class attribute accessors so that they are thread safe. This will involve sharing the attributes across threads, and locking them before changing their values. If they've also been made to behave as they do in Class::Data::Inheritable, we'll have to figure out a way to make it so that newly generated accessors for subclasses are shared between threads, too. This may not be easy.

BUGS

Please send bug reports to <bug-class-meta@rt.cpan.org> or report them via the CPAN Request Tracker at http://rt.cpan.org/NoAuth/Bugs.html.

AUTHOR

David Wheeler <david@kineticode.com>

COPYRIGHT AND LICENSE

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

Class::Meta - Class automation, introspection, and data validation