# PODNAME: Moose::Cookbook::Basics::Genome_OverloadingSubtypesAndCoercion # ABSTRACT: Operator overloading, subtypes, and coercion __END__ =pod =encoding UTF-8 =head1 NAME Moose::Cookbook::Basics::Genome_OverloadingSubtypesAndCoercion - Operator overloading, subtypes, and coercion =head1 VERSION version 2.2207 =head1 SYNOPSIS package Human; use Moose; use Moose::Util::TypeConstraints; subtype 'Sex' => as 'Str' => where { $_ =~ m{^[mf]$}s }; has 'sex' => ( is => 'ro', isa => 'Sex', required => 1 ); has 'mother' => ( is => 'ro', isa => 'Human' ); has 'father' => ( is => 'ro', isa => 'Human' ); use overload '+' => \&_overload_add, fallback => 1; sub _overload_add { my ( $one, $two ) = @_; die('Only male and female humans may create children') if ( $one->sex() eq $two->sex() ); my ( $mother, $father ) = ( $one->sex eq 'f' ? ( $one, $two ) : ( $two, $one ) ); my $sex = 'f'; $sex = 'm' if ( rand() >= 0.5 ); return Human->new( sex => $sex, mother => $mother, father => $father, ); } =head1 DESCRIPTION This Moose cookbook recipe shows how operator overloading, coercion, and subtypes can be used to mimic the human reproductive system (well, the selection of genes at least). =head1 INTRODUCTION Our C class uses operator overloading to allow us to "add" two humans together and produce a child. Our implementation does require that the two objects be of opposite sex. Remember, we're talking about biological reproduction, not marriage. While this example works as-is, we can take it a lot further by adding genes into the mix. We'll add the two genes that control eye color, and use overloading to combine the genes from the parent to model the biology. =head2 What is Operator Overloading? Overloading is I a Moose-specific feature. It's a general OO concept that is implemented in Perl with the C pragma. Overloading lets objects do something sane when used with Perl's built in operators, like addition (C<+>) or when used as a string. In this example we overload addition so we can write code like C<$child = $mother + $father>. =head1 GENES There are many genes which affect eye color, but there are two which are most important, I and I. We will start by making a class for each gene. =head2 Human::Gene::bey2 package Human::Gene::bey2; use Moose; use Moose::Util::TypeConstraints; type 'bey2_color' => where { $_ =~ m{^(?:brown|blue)$} }; has 'color' => ( is => 'ro', isa => 'bey2_color' ); This class is trivial. We have a type constraint for the allowed colors, and a C attribute. =head2 Human::Gene::gey package Human::Gene::gey; use Moose; use Moose::Util::TypeConstraints; type 'gey_color' => where { $_ =~ m{^(?:green|blue)$} }; has 'color' => ( is => 'ro', isa => 'gey_color' ); This is nearly identical to the C class, except that the I gene allows for different colors. =head1 EYE COLOR We could just give four attributes (two of each gene) to the C class, but this is a bit messy. Instead, we'll abstract the genes into a container class, C. Then a C can have a single C attribute. package Human::EyeColor; use Moose; use Moose::Util::TypeConstraints; coerce 'Human::Gene::bey2' => from 'Str' => via { Human::Gene::bey2->new( color => $_ ) }; coerce 'Human::Gene::gey' => from 'Str' => via { Human::Gene::gey->new( color => $_ ) }; has [qw( bey2_1 bey2_2 )] => ( is => 'ro', isa => 'Human::Gene::bey2', coerce => 1 ); has [qw( gey_1 gey_2 )] => ( is => 'ro', isa => 'Human::Gene::gey', coerce => 1 ); The eye color class has two of each type of gene. We've also created a coercion for each class that coerces a string into a new object. Note that a coercion will fail if it attempts to coerce a string like "indigo", because that is not a valid color for either type of gene. As an aside, you can see that we can define several identical attributes at once by supplying an array reference of names as the first argument to C. We also need a method to calculate the actual eye color that results from a set of genes. The I brown gene is dominant over both blue and green. The I green gene is dominant over blue. sub color { my ($self) = @_; return 'brown' if ( $self->bey2_1->color() eq 'brown' or $self->bey2_2->color() eq 'brown' ); return 'green' if ( $self->gey_1->color() eq 'green' or $self->gey_2->color() eq 'green' ); return 'blue'; } We'd like to be able to treat a C object as a string, so we define a string overloading for the class: use overload '""' => \&color, fallback => 1; Finally, we need to define overloading for addition. That way we can add together two C objects and get a new one with a new (genetically correct) eye color. use overload '+' => \&_overload_add, fallback => 1; sub _overload_add { my ( $one, $two ) = @_; my $one_bey2 = 'bey2_' . _rand2(); my $two_bey2 = 'bey2_' . _rand2(); my $one_gey = 'gey_' . _rand2(); my $two_gey = 'gey_' . _rand2(); return Human::EyeColor->new( bey2_1 => $one->$one_bey2->color(), bey2_2 => $two->$two_bey2->color(), gey_1 => $one->$one_gey->color(), gey_2 => $two->$two_gey->color(), ); } sub _rand2 { return 1 + int( rand(2) ); } When two eye color objects are added together, the C<_overload_add()> method will be passed two C objects. These are the left and right side operands for the C<+> operator. This method returns a new C object. =head1 ADDING EYE COLOR TO Cs Our original C class requires just a few changes to incorporate our new C class. use List::Util 1.56 qw( mesh ); coerce 'Human::EyeColor' => from 'ArrayRef' => via { my @genes = qw( bey2_1 bey2_2 gey_1 gey_2 ); return Human::EyeColor->new( mesh ( \@genes, $_ ) ); }; has 'eye_color' => ( is => 'ro', isa => 'Human::EyeColor', coerce => 1, required => 1, ); We also need to modify C<_overload_add()> in the C class to account for eye color: return Human->new( sex => $sex, eye_color => ( $one->eye_color() + $two->eye_color() ), mother => $mother, father => $father, ); =head1 CONCLUSION The three techniques we used, overloading, subtypes, and coercion, combine to provide a powerful interface. If you'd like to learn more about overloading, please read the documentation for the L pragma. To see all the code we created together, take a look at F. =head1 NEXT STEPS Had this been a real project we'd probably want: =over 4 =item Better Randomization with Crypt::Random =item Characteristic Base Class =item Mutating Genes =item More Characteristics =item Artificial Life =back =head1 AUTHORS =over 4 =item * Stevan Little =item * Dave Rolsky =item * Jesse Luehrs =item * Shawn M Moore =item * יובל קוג'מן (Yuval Kogman) =item * Karen Etheridge =item * Florian Ragwitz =item * Hans Dieter Pearcey =item * Chris Prather =item * Matt S Trout =back =head1 COPYRIGHT AND LICENSE This work is licensed under a Creative Commons Attribution 3.0 Unported License. License details are at: L =cut