use strict; #-*-cperl,hi-lock,auto-fill-*- use warnings; use lib qw( ../../../../lib ); =head1 NAME Algorithm::Evolutionary::Individual::Base - Base class for chromosomes that knows how to build them, and has some helper methods. =head1 SYNOPSIS use Algorithm::Evolutionary::Individual::Base; my $xmlStr="1010"; my $ref = XMLin($xmlStr); my $binIndi2 = Algorithm::Evolutionary::Individual::Base->fromXML( $ref ); #From XML fragment print $binIndi2->asXML(); my $indi = Algorithm::Evolutionary::Individual::Base->fromParam( $ref->{initial}{section}{indi}{param} ); #From parametric description $binIndi2->Fitness( 3.5 ); #Sets or gets fitness print $binIndi2->Fitness(); my $emptyIndi = new Algorithm::Evolutionary::Individual::Base; =head1 DESCRIPTION Base class for individuals, that is, "chromosomes" in evolutionary computation algorithms. However, chromosomes needn't be bitstrings, so the name is a bit misleading. This is, however, an "empty" base class, that acts as a boilerplate for deriving others. =cut package Algorithm::Evolutionary::Individual::Base; use Algorithm::Evolutionary::Utils qw(parse_xml); use YAML qw(Dump Load LoadFile); use Carp; our $VERSION = sprintf "%d.%03d", q$Revision: 3.1 $ =~ /(\d+)\.(\d+)/g; use constant MY_OPERATORS => qw(None); =head1 METHODS =head2 new( $options ) Creates a new Base individual of the required class, with a fitness, and sets fitnes to undef. Takes as params a hash to the options of the individual, that will be passed on to the object of the class when it iss initialized. =cut sub new { my $class = shift; if ( $class !~ /Algorithm::Evolutionary/ ) { $class = "Algorithm::Evolutionary::Individual::$class"; } my $options = shift; my $self = { _fitness => undef }; # Avoid error bless $self, $class; # And bless it #If the class is not loaded, we load it. The if ( !$INC{"$class\.pm"} ) { eval "require $class" || croak "Can't find $class Module"; } if ( $options ) { $self->set( $options ); } return $self; } =head2 create( $ref_to_hash ) Creates a new individual, but uses a different interface: takes a ref-to-hash, with named parameters, which gives it a common interface to all the hierarchy. The main difference with respect to new is that after creation, it is initialized with random values. =cut sub create { my $class = shift; my $ref = shift || croak "Can't find the parameters hash"; my $self = Algorithm::Evolutionary::Individual::Base::new( $class, $ref ); $self->randomize(); return $self; } =head2 set( $ref_to_hash ) Sets values of an individual; takes a hash as input. Keys are prepended an underscore and turn into instance variables =cut sub set { my $self = shift; my $hash = shift || croak "No params here"; for ( keys %{$hash} ) { $self->{"_$_"} = $hash->{$_}; } } =head2 fromXML( $xml_string ) Takes a definition in the shape .... and turns it into a bitstring, if it knows how to do it. The definition must have been processed using XML::Simple. It forwards stuff it does not know about to the corresponding subclass, which should implement the C method. The class it refers about is Cd in runtime. =cut sub fromXML { my $class = shift; my $xml = shift || croak "XML fragment missing "; my $fragment; # Inner part of the XML if ( ref $xml eq '' ) { #We are receiving a string, parse it $xml = parse_xml ($xml ); croak "Incorrect XML fragment" if !$xml->{'indi'}; # $fragment = $xml->{'indi'}; } elsif ( $xml->{'indi'} ) { # parsed externally, as in general.t $fragment = $xml->{'indi'}; } else { #parsed fragment $fragment = $xml; } my $thisClassName = $fragment->{'-type'}; if ( $class eq __PACKAGE__ ) { #Deduct class from the XML $class = $thisClassName || shift || croak "Class name missing"; } #Calls new, adds preffix if it's not there my $self = Algorithm::Evolutionary::Individual::Base::new( $class ); ($self->Fitness( $fragment->{'fitness'} ) ) if defined $fragment->{'fitness'}; $class = ref $self; eval "require $class" || croak "Can't find $class\.pm Module"; no strict qw(refs); # To be able to check if a ref exists or not for (@{$fragment->{'atom'}} ) { $self->addAtom($_); #roundabout way of adding the content of the stuff } return $self; } =head2 fromParam( $xml_fragment ) Takes an array of params that describe the individual, and builds it, with random initial values. Params have this shape: The 'type' will show the class of the individuals that are going to be created, and the rest will be type-specific, and left to the particular object to interpret. =cut sub fromParam { my $class = shift; my $xml = shift || croak "XML fragment missing "; my $thisClass; my %params; for ( @{$xml->{'param'}} ) { if ( $_->{'-name'} eq 'type' ) { $thisClass = $_->{'-value'} } else { $params{ $_->{'-name'} } = $_->{'-value'}; } } $thisClass = "Algorithm::Evolutionary::Individual::$thisClass" if $thisClass !~ /Algorithm::Evolutionary/; eval "require $thisClass" || croak "Can't find $class\.pm Module"; my $self = $thisClass->new(); $self->set( \%params ); $self->randomize(); return $self; } =head2 asXML() Prints it as XML. The caller must close the tags. =cut sub asXML { my $self = shift; my ($opName) = ( ( ref $self) =~ /::(\w+)$/ ); my $str = "{_fitness} ) { $str.= "fitness='$self->{_fitness}'"; } $str.=" />\n\t"; return $str; } =head2 as_yaml() Prints it as YAML. =cut sub as_yaml { my $self = shift; return Dump($self); } =head2 as_string() Prints it as a string in the most meaningful representation possible =cut sub as_string { croak "This function is not defined at this level, you should override it in a subclass\n"; } =head2 as_string_with_fitness( [$separator] ) Prints it as a string followed by fitness. Separator by default is C<;> =cut sub as_string_with_fitness { my $self = shift; my $separator = shift || "; "; return $self->as_string().$separator.$self->Fitness(); } =head2 Atom( $index [, $value ) Sets or gets the value of an atom. Each individual is divided in atoms, which can be accessed sequentially. If that does not apply, Atom can simply return the whole individual =cut sub Atom { croak "This function is not defined at this level, you should override it in a subclass\n"; } =head2 Fitness( [$value] ) Sets or gets fitness =cut sub Fitness { my $self = shift; if ( defined $_[0] ) { $self->{_fitness} = shift; } return $self->{_fitness}; } =head2 my_operators() Operators that can act on this data structure. Returns an array with the names of the known operators =cut sub my_operators { my $self = shift; return $self->MY_OPERATORS; } =head2 evaluate( $fitness ) Evaluates using the $fitness thingy given. Can be a L object or a ref-to-sub =cut sub evaluate { my $self = shift; my $fitness_func = shift || croak "Need a fitness function"; if ( ref $fitness_func eq 'CODE' ) { return $self->Fitness( $fitness_func->($self) ); } elsif ( ( ref $fitness_func ) =~ 'Fitness' ) { return $self->Fitness( $fitness_func->apply($self) ); } else { croak "$fitness_func can't be used to evaluate"; } } =head2 Chrom() Sets or gets the chromosome itself, that is, the data structure evolved. Since each derived class has its own data structure, and its own name, it is left to them to return it =cut sub Chrom { my $self = shift; croak "To be implemented in derived classes!"; } =head2 size() OK, OK, this is utter inconsistence, but I'll re-consistence it eventually. Returns a meaningful size; but should be reimplemented by siblings =cut sub size() { croak "To be implemented in derived classes!"; } =head1 Known subclasses =over 4 =item * L =item * L =item * L =item * L =back =head1 Copyright This file is released under the GPL. See the LICENSE file included in this distribution, or go to http://www.fsf.org/licenses/gpl.txt CVS Info: $Date: 2009/07/24 08:53:13 $ $Header: /cvsroot/opeal/Algorithm-Evolutionary/lib/Algorithm/Evolutionary/Individual/Base.pm,v 3.1 2009/07/24 08:53:13 jmerelo Exp $ $Author: jmerelo $ $Revision: 3.1 $ $Name $ =cut "The plain truth";