Add a mention of load order issues

[gitmo/Moose.git] / lib / Moose / Manual / Concepts.pod

=head1 NAME

Moose::Manual::Concepts - Moose OO Concepts

=head1 MOOSE CONCEPTS (VS "OLD SCHOOL" Perl)

In the past, you may not have thought too much about the difference
between packages and classes, attributes and methods, constructors and
methods, etc. With Moose, these are all conceptually separate things,
even though under the hood they're implemented with plain old Perl.

Our meta-object protocol (aka MOP) provides a well-defined
introspection features for each of those concepts, and Moose in turn
provides I<distinct> sugar for each of them. Moose also introduces
additional concepts such as roles, method modifiers, and declarative
delegation.

Knowing what these concepts mean in Moose-speak, and how they used to
be done in old school Perl 5 OO is a good way to start learning to use
Moose.

=head2 Class

When you say "use Moose" in a package, you are making your package a
class. At its simplest, a class will consist simply of attributes
and/or methods. It can also include roles, method modifiers, and more.

A class I<has> zero or more B<attributes>.

A class I<has> zero or more B<methods>.

A class I<has> zero or more superclasses (aka parent classes). A
class inherits from its superclass(es).

A class I<has> zero or more B<method modifiers>. These modifiers can
apply to its own methods or methods that are inherited from its
ancestors.

A class I<does> zero or more B<roles>.

A class I<has> a B<constructor> and a B<destructor>. These are
provided for you "for free" by Moose.

The B<constructor> accepts named parameters corresponding to the
class's attributes and uses them to initialize an B<object instance>.

A class I<has> a B<metaclass>, which in turn has B<meta-attributes>,
B<meta-methods>, and B<meta-roles>. This metaclass I<describes> the
class.

A class is usually analogous to a category of nouns, like "People" or
"Users".

  package Person;

  use Moose;
  # now it's a Moose class!

=head2 Attribute

An attribute is a property of the class that defines it. It I<always>
has a name, and it I<may have> a number of other defining
characteristics.

These characteristics may include a read/write flag, a B<type>,
accessor method names, B<delegations>, a default value, and more.

Attributes I<are not> methods, but defining them causes various
accessor methods to be created. At a minimum, a normal attribute will
always have a reader accessor method. Many attributes have things like
a writer method, clearer method, and predicate method ("has it been
set?").

An attribute may also define B<delegation>s, which will create
additional methods based on the delegation specification.

By default, Moose stores attributes in the object instance, which is a
hashref, I<but this is invisible to the author of a Moose-base class>!
It is best to think of Moose attributes as "properties" of the
I<opaque> B<object instance>. These properties are accessed through
well-defined accessor methods.

An attribute is something that the class's members have. For example,
People have first and last names. Users have passwords and last login
datetimes.

  has 'first_name' => (
      is  => 'rw',
      isa => 'Str',
  );

=head2 Method

A method is very straightforward. Any subroutine you define in your
class is a method.

Methods correspond to verbs, and are what your objects can do. For
example, a User can login.

  sub login { ... }

=head2 Roles

A role is something that a class I<does>. For example, a Machine class
might do the Breakable role, and so could a Bone class. A role is
used to define some concept that cuts across multiple unrelated
classes, like "breakability", or "has a color".

A role I<has> zero or more B<attributes>.

A role I<has> zero or more B<methods>.

A role I<has> zero or more B<method modifiers>.

A role I<has> zero or more B<required methods>.

A required method is not implemented by the role. Required methods say
"to use this Role you must implement this method".

Roles are I<composed> into classes (or other roles). When a role is
composed into a class, its attributes and methods are "flattened" into
the class. Roles I<do not> show up in the inheritance hierarchy. When
a role is composed, its attributes and methods appear as if they were
defined I<in the consuming class>.

Role are somewhat like mixins or interfaces in other OO languages.

  package Breakable;

  use Moose::Role;

  has is_broken => (
      is  => 'rw',
      isa => 'Bool',
  );

  requires 'break';

  before 'break' => {
      my $self = shift;

      $self->is_broken(1);
  };

=head2 Method Modifiers

A method modifier is a hook that is called when a named method is
called. For example, you could say "before calling C<login()>, call
this modifier first". Modifiers come in different flavors like
"before", "after", "around", and "augment", and you can apply more
than one modifier to a single method.

Method modifiers are often used as an alternative to overriding a
method in a parent class. They are also used in roles as a way of
modifying methods in the consuming class.

Under the hood, a method modifier is just a plain old Perl subroutine
that gets called before or after (or around, etc.) some named method.

  before 'login' => sub {
      my $self = shift;
      my $pw   = shift;

      warn "Called login() with $pw\n";
  };

=head2 Type

Moose also comes with a (miniature) type system. This allows you to
define types for attributes. Moose has a set of built-in types based
on what Perl provides, such as "Str", "Num", "Bool", "HashRef", etc.

In addition, every class name in your application can also be used as
a type name. We saw an example using "DateTime" earlier.

Finally, you can define your own types, either as subtypes or entirely
new types, with their own constraints. For example, you could define a
type "PosInt", a subtype of "Int" which only allows positive numbers.

=head2 Delegation

Moose attributes provide declarative syntax for defining
delegations. A delegation is a method which calls some method on an
attribute to do its real work.

=head2 Constructor

A constructor creates an B<object instance> for the class. In old
school Perl, this was usually done by defining a method called
C<new()> which in turn called C<bless> on a reference.

With Moose, this C<new()> method is created for you, and it simply
does the right thing. You should never need to define your own
constructor!

Sometimes you want to do something whenever an object is created. In
those cases, you can provide a C<BUILD()> method in your class. Moose
will call this for you after creating a new object.

=head2 Destructor

This is a special method called when an object instance goes out of
scope. You can specialize what your class does in this method if you
need to, but you usually don't.

With old school Perl 5, this is the C<DESTROY()> method, but with
Moose it is the C<DEMOLISH()> method.

=head2 Object Instance

An object instance is a specific noun in the class's "category". For
example, one specific Person or User. An instance is created by the
class's B<constructor>.

An instance has values for its attributes. For example, a specific
person has a first and last name.

In old school Perl 5, this is often a blessed hash reference. With
Moose, you should never need to know what your object instance
actually is. (ok, it's usually a blessed hashref with Moose too)

=head2 Moose VS Old School Summary

=over 4

=item * Class

A package with no introspection other than mucking about in the symbol
table.

With Moose, you get well-defined declaration and introspection.

=item * Attributes

Hand-written accessor methods, symbol table hackery, or a helper
module like C<Class::Accessor>.

With Moose, these are declaratively defined, and distinct from
methods.

=item * Method

These are pretty much the same in Moose as in old school Perl.

=item * Roles

C<Class::Trait> or C<Class::Role>, or maybe C<mixin.pm>.

With Moose, they're part of the core feature set, and are
introspectable like everything else.

=item * Method Modifiers

Could only be done through serious symbol table wizardry, and you
probably never saw this before (at least in Perl 5).

=item * Type

Hand-written parameter checking in your C<new()> method and accessors.

With Moose, you define types declaratively, and then use them by name
in your attributes.

=item * Delegation

C<Class::Delegation> or C<Class::Delegator>, but probably even more
hand-written code.

With Moose, this is also declarative.

=item * Constructor

A C<new()> method which calls C<bless> on a reference.

Comes for free when you define a class with Moose.

=item * Destructor

A C<DESTROY()> method.

With Moose, this is called C<DEMOLISH()>.

=item * Object Instance

A blessed reference, usually a hash reference.

With Moose, this is an opaque thing which has a bunch of attributes
and methods, as defined by its class.

=item * Immutabilization

Moose comes with a feature called "immutabilization". When you make
your class immutable, it means you're done adding methods, attributes,
roles, etc. This lets Moose optimize your class with a bunch of
extremely dirty in-place code generation tricks that speed up things
like object construction and so on.

=back

=head1 META WHAT?

A metaclass is a class that describes classes. With Moose, every class
you define gets a C<meta()> method. It returns a L<Moose::Meta::Class>
object, which has an introspection API that can tell you about the
class it represents.

  my $meta = User->meta();

  for my $attribute ( $meta->compute_all_applicable_attributes ) {
      print $attribute->name(), "\n";

      if ( $attribute->has_type_constraint ) {
          print "  type: ", $attribute->type_constraint->name, "\n";
      }
  }

  for my $method ( $meta->compute_all_applicable_methods ) {
      print $method->name, "\n";
  }

Almost every concept we defined earlier has a meta class, so we have
L<Moose::Meta::Class>, L<Moose::Meta::Attribute>,
L<Moose::Meta::Method>, L<Moose::Meta::Role>,
L<Moose::Meta::TypeConstraint>, L<Moose::Meta::Instance>, and so on.

=head1 BUT I NEED TO DO IT MY WAY!

One of the great things about Moose is that if you dig down and find
that it does something the "wrong way", you can change it by extending
a metaclass. For example, you can have arrayref based objects, you can
make your constructors strict (no unknown params allowed!), you can
define a naming scheme for attribute accessors, you can make a class a
Singleton, and much, much more.

Many of these extensions require surprisingly small amounts of code,
and once you've done it once, you'll never have to hand-code "your way
of doing things" again. Instead you'll just load your favorite
extensions.

  package MyWay::User;

  use Moose;
  use MooseX::StrictConstructor
  use MooseX::MyWay;

  has ...;

=head1 WHAT NEXT?

So you're sold on Moose. Time to learn how to really use it.

If you want to see how Moose would translate directly old school Perl
5 OO code, check out L<Moose::Unsweetened>. This might be helpful for
quickly wrapping your brain around some aspects of "the Moose way".

Obviously, the next thing to read is the rest of the L<Moose::Manual>.

After that we recommend that you start with the L<Moose::Cookbook>. If
you work your way through all the recipes under the basics section,
you should have a pretty good sense of how Moose works, and all of its
basic OO features.

After that, check out the Role recipes. If you're really curious, go
on and read the Meta and Extending recipes, but those are mostly there
for people who want to be Moose wizards and change how Moose works.

=head1 AUTHOR

Dave Rolsky E<lt>autarch@urth.orgE<gt>

=head1 COPYRIGHT AND LICENSE

Copyright 2008 by Infinity Interactive, Inc.

L<http://www.iinteractive.com>

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

=cut