#========================================================================
#
# Badger::Class::Methods
#
# DESCRIPTION
#   Class mixin module for adding methods to a class.
#
# AUTHOR
#   Andy Wardley   <abw@wardley.org>
#
#========================================================================

package Badger::Class::Methods;

use Carp;
use Badger::Class
    version   => 0.01,
    debug     => 0,
    base      => 'Badger::Base',
    import    => 'class BCLASS',
    constants => 'DELIMITER ARRAY HASH PKG CODE',
    utils     => 'is_object',
    exports   => {
        hooks => {
            init => \&initialiser,
            map { $_ => [\&generate, 1] }
            qw( accessors mutators get set slots hash auto_can )
        },
    },
    messages  => {
        no_target  => 'No target class specified to generate methods for',
        no_type    => 'No method type specified to generate',
        no_methods => 'No %s specified to generate',
        bad_method => 'Invalid %s method: %s',
        bad_type   => 'Invalid method generator specified: %s',
    };

# method aliases
*get = \&accessors;
*set = \&mutators;

our $AUTOLOAD;

sub generate {
    my $class   = shift;
    my $target  = shift
        || return $class->error_msg('no_target');
    my $type    = shift
        || return $class->error_msg('no_type');
    my $methods = shift
        || return $class->error_msg( no_methods => $type );
    my $code    = $class->can($type)
        || return $class->error_msg( bad_type => $type );

    $class->debug("generate($target, $type, $methods)") if DEBUG;
    
    $code->($class, $target, $methods);
}

sub accessors {
    my ($class, $target, $methods) = shift->args(@_);

    $target->import_symbol( 
        $_ => $class->accessor($_) 
    ) for @$methods;
}

sub accessor {
    my ($self, $name) = @_;
    return sub {
        $_[0]->{ $name };
    };
}

sub mutators {
    my ($class, $target, $methods) = shift->args(@_);

    $target->import_symbol( 
        $_ => $class->mutator($_) 
    ) for @$methods;
}

sub mutator {
    my ($self, $name) = @_;
    return sub {
        # You wouldn't ever want to write a real subroutine like this.
        # But that's OK, because we're here to do it for you.  You get
        # the efficiency without having to ever look at code like this:
        @_ == 2 
            ? ($_[0]->{ $name } = $_[1])
            :  $_[0]->{ $name };
    };
}

sub hash {
    my ($class, $target, $methods) = shift->args(@_);

    foreach (@$methods) {
        my $name = $_;              # new lexical var for closure
        $target->import_symbol(
            $name => sub {
                # return hash ref when called without args
                return $_[0]->{ $name } if @_ == 1;
                
                # return hash item when called with one non-ref arg
                return $_[0]->{ $name }->{ $_[1] } if @_ == 2 && ! ref $_[1];
                
                # add items to hash when called with hash ref or multiple args
                my $self  = shift;
                my $items = @_ == 1 && ref $_[0] eq HASH ? shift : { @_ };
                my $hash  = $self->{ $name };
                @$hash{ keys %$items } = values %$items;
                return $hash;
            }
        );
    }
}

sub initialiser {
    my ($class, $target, $methods) = shift->args(@_);

    $target->import_symbol(
        init => sub {
            my ($self, $config) = @_;
            $self->{ config } = $config;
            foreach my $name (@$methods) {
                $self->$name($config);
            }
            return $self;
        }
    );
}

sub slots {
    my ($class, $target, $methods) = shift->args(@_);
    my $index = 0;

    foreach my $method (@$methods) {
        my $i = $index++;           # new lexical var for closure
        $target->import_symbol(
            $method => sub {
                return @_ > 1
                    ? ($_[0]->[$i] = $_[1])
                    :  $_[0]->[$i];
            }
        );
    }
}

sub auto_can {
    my ($class, $target, $methods) = shift->args(@_);

    die "auto_can only support a single method at this time\n"
        if @$methods != 1;
        
    my $method = shift @$methods;

    croak "Invalid auto_can method specified: $method\n"
        if ref $method eq CODE;
        
    # avoid runaways
    my $seen = { };
    
    $class->debug("installing AUTOLOAD and can() in $target") if DEBUG;

    $target->import_symbol( 
        can => sub {
            my ($this, $name, @args) = @_;
            $class->debug("looking to see if $this can $name()") if DEBUG;

            # This avoids runaways where can() calls itself repeatedly, but 
            # doesn't prevent can() from being called several times for the
            # same item. 
            return if $seen->{ $name };
            local $seen->{ $name } = 1;

            return $this->SUPER::can($name)
                || $this->$method($name, @args);
        }
    );

    $target->import_symbol( 
        AUTOLOAD => sub {
            my ($this, @args) = @_;
            my ($name) = ($AUTOLOAD =~ /([^:]+)$/ );
            return if $name eq 'DESTROY';
            if (my $method = $this->can($name, @args)) {
                my $that = class($this);
                $class->debug("$class installing $name method in $that") if DEBUG;
                $that->method( $name => $method );
                return $method->($this, @args);
            }

            # Hmmm... what if $this isn't a subclass of Badger::Base?
            return $this->error_msg( bad_method => $name, ref $this, (caller())[1,2] );
        }
    );

    $class->debug("installed AUTOLOAD and can() in $target") if DEBUG;
}

sub args {
    my $class   = shift;
    my $target  = shift;
    my $methods = @_ == 1 ? shift : [ @_ ];

    # update $target to a Badger::Class object if not already one
    $target  = class($target)
        unless is_object(BCLASS, $target);

    # split text string into list ref of method names
    $methods = [ split(DELIMITER, $methods) ] 
        unless ref $methods eq ARRAY;
    
    return ($class, $target, $methods);
}
        


1;

__END__

=head1 NAME

Badger::Class::Method - metaprogramming module for adding methods to a class

=head1 SYNOPSIS

    package My::Module;
    
    # using the module directly
    use Badger::Class::Methods
        accessors => 'foo bar',
        mutators  => 'wiz bang';
    
    # or via Badger::Class
    use Badger::Class
        accessors => 'foo bar',
        mutators  => 'wiz bang';

=head1 DESCRIPTION

This module can be used to generate methods for a class. It can be used
directly, or via the L<accessors|Badger::Class/accessors>, 
L<accessors|Badger::Class/accessors> and L<slots|Badger::Class/slots>
export hooks in L<Badger::Class>.

=head1 METHODS

=head2 generate($class,$type,$methods)

This method is a central dispatcher to other methods.

    Badger::Class::Methods->generate(
        accessors => 'foo bar',
    );

=head2 accessors($class,$methods) / get($class,$methods)

This method can be used to generate accessor (read-only) methods for a class
(L<Badger::Class> object) or package name. You can pass a list, reference to a
list, or a whitespace delimited string of method names as arguments.

    # these all do the same thing
    Badger::Class::Methods->accessors('My::Module', 'foo bar');
    Badger::Class::Methods->accessors('My::Module', 'foo', 'bar');
    Badger::Class::Methods->accessors('My::Module', ['foo', 'bar']);

A method will be generated in the target class for each that returns the
object member data of the same name.  The method itself is generated by 
calling the L<accessor()> method.

=head2 accessor($name)

This method generates an accessor method for accessing the item in an object
denoted by C<$name>.  The method is returned as a code reference.  It is not
installed in the symbol table of any package - that's up to you (or use the
L<accessors()> method). 

    my $coderef = Badger::Class::Method->accessor('foo');

The code generated is equivalent to this:

    sub foo {
        $_[0]->{ foo };
    }

=head2 mutators($class,$methods) / set($class,$methods)

This method can be used to generate mutator (read/write) methods for a class
(L<Badger::Class> object) or package name. You can pass a list, reference to a
list, or a whitespace delimited string of method names as arguments.

    # these all do the same thing
    Badger::Class::Methods->mutators('My::Module', 'foo bar');
    Badger::Class::Methods->mutators('My::Module', 'foo', 'bar');
    Badger::Class::Methods->mutators('My::Module', ['foo', 'bar']);

A method will be generated in the target class for each that returns the
object member data of the same name. If an argument is passed to the method
then the member data is updated and the new value returned.

The method itself is generated by calling the L<mutator()> method.

=head2 mutator($name)

This method generates a mutator method for accessing and updating the item in
an object denoted by C<$name>. The method is returned as a code reference. It
is not installed in the symbol table of any package - that's up to you (or use
the L<mutators()> method).

    my $coderef = Badger::Class::Method->mutator('foo');

The code generated is equivalent to this:

    sub foo {
        @_ == 2 
            ? ($_[0]->{ foo } = $_[1])
            :  $_[0]->{ foo };
    }

Ugly isn't it?   But of course you wouldn't ever write it like that, being 
a conscientious Perl programmer concerned about the future readability and
maintainability of your code.  Instead you might write it something like
this:

    sub foo {
        my $self = shift;
        if (@_) {
            # an argument implies a set
            return ($self->{ foo } = shift);
        }
        else {
            # no argument implies a get
            return $self->{ foo };
        }
    }

Or perhaps like this:

    sub foo {
        my $self = shift;
        # update value if an argument was passed
        $self->{ foo } = shift if @_;
        return $self->{ foo };
    }

Or even like this (my personal favourite):

    sub foo {
        my $self = shift;
        return @_
            ? ($self->{ foo } = shift)
            :  $self->{ foo };
    }

Whichever way you do it is a waste of time, both for you and anyone who has to
read your code at a later. Seriously, give it up! Let us generate the methods
for you. We'll not only save you the effort of typing pages of code that
no-one will ever read (or want to read), but we'll also generate the most
efficient code for you. The kind that you wouldn't normally want to handle by
yourself.

So in summary, using this method will keep your code clean, your code 
efficient, and will free up the rest of the afternoon so you can go out 
skateboarding.  Tell your boss I said it was OK.

=head2 hash($class, $methods)

This method generates methods for accessing or updating items in a hash
reference stored in an object.  In the following example we create a 
C<users()> method for accessing the internal C<users> hash reference.

    package Your::Module;
    
    use base 'Badger::Base';
    use Badger::Class::Methods
        hash => 'users';
    
    sub init {
        my ($self, $config) = @_;
        $self->{ users } = $config->{ users } || { };
        return $self;
    }

The C<init()> method copies any C<users> passed as a configuration
parameter or creates an empty hash reference.

    my $object = Your::Module->new(
        users => {
            tom => 'tom@badgerpower.com',
        }
    );

When called without any arguments, the generated C<users()> method returns a
reference to the C<users> hash array.

    print $object->users->{ tom };  # tom@badgerpower.com

When called with a single non-reference argument, it returns the entry
in the hash corresponding to that key.

    print $object->users('tom');    # tom@badgerpower.com

When called with a single reference to a hash array, or a list of named 
parameters, the method will add the new items to the internal hash array.
A reference to the hash array is returned.

    $object->users({                        # single hash ref
        dick  => 'richard@badgerpower.com', 
        harry => 'harold@badgerpower.com',
    });
    
    $object->users(                         # list of amed parameters
        dick  => 'richard@badgerpower.com', 
        harry => 'harold@badgerpower.com',
    );

=head2 initialiser($class,$methods)

This method can be used to create a custom C<init()> method for your object
class. A list, reference to a list, or string of whitespace delimited method
names should be passed an argument(s). A method will be generated which 
calls each in turn, passing a reference to a hash array of configuration
parameters.

    use Badger::Class::Methods->initialiaser(
        'My::Module', 
        'init_foo init_bar'
    )

The above example will generate an C<init()> method in C<My::Module>
equivalent to:

    sub init {
        my ($self, $config) = @_;
        $self->{ config } = $config;
        $self->init_foo($config);
        $self->init_bar($config);
        return $self;
    }

It's up to you to implement the C<init_foo()> and C<init_bar()> methods,
or to inherit them from a base class or mixin.

=head2 slots($class,$methods)

This method can be used to define methods for list-based object classes.
A list, reference to a list, or string of whitespace delimited method
names should be passed an argument(s).  A method will be generated for
each item specified.  The first method will reference the first (0th) item
in the list, the second method will reference the second (1st), and so on.

    Badger::Class::Methods->slots('My::Module', 'foo bar');
    Badger::Class::Methods->slots('My::Module', 'foo', 'bar');
    Badger::Class::Methods->slots('My::Module', ['foo', 'bar']);

It is usually called indirectly via the L<slots|Badger::Class/slots>
export hook in L<Badger::Class>.

    package Badger::Example;
    
    use Badger::Class
        slots => 'size colour object';
    
    sub new {
        my ($class, @stuff) = @_;
        bless \@stuff, $class;
    }

The above example defines a simple list-based object class with three
slots: C<size>, C<colour> and C<object>.  You can use it like this:

    my $bus = Badger::Test::Slots->new(qw( big red bus ));
    
    print $bus->size;       # big
    print $bus->colour;     # red
    print $bus->object;     # bus

The methods generated are mutators.  That is, you can pass an argument
to update the slot value.

    $bus->size('large');

=head2 auto_can($class,$method)

This can be used to define a method that automatically generates other
methods on demand.  

Suppose you have a view class that renders a view of a tree. In classic
I<double dispatch> style, each node in the tree calls a method against the
view object corresponding to the node's type. A C<text> node calls
C<$view-E<gt>view_text($self)>, a C<bold> node calls
C<$view-E<gt>view_bold($self)>, and so on (we're assuming that this is some
kind of document object model we're rendering, but it could apply to
anything).

Our view methods might look something like this:

    sub view_text {
        my ($self, $node) = @_;
        print "TEXT: $node\n";
    }

    sub view_bold {
        my ($self, $node) = @_;
        print "BOLD: $node\n";
    }

This can get rather repetitive and boring if you've got lots of different
node types.  So instead of defining all the methods manually, you can declare
an C<auto_can> method that will create methods on demand.

    use Badger::Class
        auto_can => 'can_view';
        
    sub can_view {
        my ($self, $name) = @_;
        my $NAME = uc $name;
        
        return sub {
            my ($self, $node) = @_;
            print "$NAME: $node";
        }
    }

The method should return a subroutine reference or any false value if it
declines to generate a method.  For example, you might want to limit the 
generator method to only creating methods that match a particular format.

    sub can_view {
        my ($self, $name) = @_;
        
        # only create methods that are prefixed with 'view_'
        if ($name =~ s/^view_//) {
            my $NAME = uc $name;
            
            return sub {
                my ($self, $node) = @_;
                print "$NAME: $node";
            }
        }
        else {
            return undef;
        }
    }

The C<auto_can()> method adds C<AUTOLOAD()> and C<can()> methods to your 
class.  The C<can()> method first looks to see if the method is pre-defined
(i.e. it does what the default C<can()> method does).  If it isn't, it then
calls the C<can_view()> method that we've declared using the C<auto_can> 
option (you can call your method C<auto_can()> if you like, but in this 
case we're calling it C<can_view()> just to be different).  The end result
is that you can call C<can()> and it will generate any missing methods on
demand.

    # this calls can_view() which returns a CODE sub 
    my $method = $object->can('view_italic');

The C<AUTOLOAD()> method is invoked whenever you call a method that 
doesn't exist.  It calls the C<can()> method to automatically generate 
the method and then installs the new method in the package's symbol table.
The next time you call the method it will be there waiting for you.  There's
no need for the C<AUTOLOAD()> method to get involved from that point on.

    # this calls can_view() to create the method and then calls it
    $object->view_cheese('Camembert');      # CHEESE: Camembert
    
    # this directly calls the new method
    $object->view_cheese('Cheddar');        # CHEESE: Cheddar

If your C<can_view()> method returns a false value then C<AUTOLOAD()> 
will raise the familiar "Invalid method..." error that you would normally
get from calling a non-existent method.

=head1 INTERNAL METHODS

=head2 args(@args)

This methods inspect the arguments and performs the necessary validation
for the L<accessors()>, L<mutators()> and L<slots()> methods.

=head1 AUTHOR

Andy Wardley L<http://wardley.org/>

=head1 COPYRIGHT

Copyright (C) 2008-2009 Andy Wardley.  All Rights Reserved.

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

=cut

# Local Variables:
# mode: perl
# perl-indent-level: 4
# indent-tabs-mode: nil
# End:
#
# vim: expandtab shiftwidth=4: