accessor.pm

来自「视频监控网络部分的协议ddns,的模块的实现代码,请大家大胆指正.」· PM 代码 · 共 751 行 · 第 1/2 页

package Object::Accessor;

use strict;
use Carp            qw[carp croak];
use vars            qw[$FATAL $DEBUG $AUTOLOAD $VERSION];
use Params::Check   qw[allow];
use Data::Dumper;

### some objects might have overload enabled, we'll need to
### disable string overloading for callbacks
require overload;

$VERSION    = '0.32';
$FATAL      = 0;
$DEBUG      = 0;

use constant VALUE => 0;    # array index in the hash value
use constant ALLOW => 1;    # array index in the hash value

=head1 NAME

Object::Accessor

=head1 SYNOPSIS

    ### using the object
    $obj = Object::Accessor->new;        # create object
    $obj = Object::Accessor->new(@list); # create object with accessors
    $obj = Object::Accessor->new(\%h);   # create object with accessors
                                         # and their allow handlers

    $bool   = $obj->mk_accessors('foo'); # create accessors
    $bool   = $obj->mk_accessors(        # create accessors with input
               {foo => ALLOW_HANDLER} ); # validation
                
    $clone  = $obj->mk_clone;            # create a clone of original
                                         # object without data
    $bool   = $obj->mk_flush;            # clean out all data

    @list   = $obj->ls_accessors;        # retrieves a list of all
                                         # accessors for this object

    $bar    = $obj->foo('bar');          # set 'foo' to 'bar'
    $bar    = $obj->foo();               # retrieve 'bar' again

    $sub    = $obj->can('foo');          # retrieve coderef for
                                         # 'foo' accessor
    $bar    = $sub->('bar');             # set 'foo' via coderef
    $bar    = $sub->();                  # retrieve 'bar' by coderef

    ### using the object as base class
    package My::Class;
    use base 'Object::Accessor';

    $obj    = My::Class->new;               # create base object
    $bool   = $obj->mk_accessors('foo');    # create accessors, etc...

    ### make all attempted access to non-existant accessors fatal
    ### (defaults to false)
    $Object::Accessor::FATAL = 1;

    ### enable debugging
    $Object::Accessor::DEBUG = 1;

    ### advanced usage -- callbacks
    {   my $obj = Object::Accessor->new('foo');
        $obj->register_callback( sub { ... } );
        
        $obj->foo( 1 ); # these calls invoke the callback you registered
        $obj->foo()     # which allows you to change the get/set 
                        # behaviour and what is returned to the caller.
    }        

    ### advanced usage -- lvalue attributes
    {   my $obj = Object::Accessor::Lvalue->new('foo');
        print $obj->foo = 1;            # will print 1
    }

    ### advanced usage -- scoped attribute values
    {   my $obj = Object::Accessor->new('foo');
        
        $obj->foo( 1 );
        print $obj->foo;                # will print 1

        ### bind the scope of the value of attribute 'foo'
        ### to the scope of '$x' -- when $x goes out of 
        ### scope, 'foo's previous value will be restored
        {   $obj->foo( 2 => \my $x );
            print $obj->foo, ' ', $x;   # will print '2 2'
        }
        print $obj->foo;                # will print 1
    }


=head1 DESCRIPTION

C<Object::Accessor> provides an interface to create per object
accessors (as opposed to per C<Class> accessors, as, for example,
C<Class::Accessor> provides).

You can choose to either subclass this module, and thus using its
accessors on your own module, or to store an C<Object::Accessor>
object inside your own object, and access the accessors from there.
See the C<SYNOPSIS> for examples.

=head1 METHODS

=head2 $object = Object::Accessor->new( [ARGS] );

Creates a new (and empty) C<Object::Accessor> object. This method is
inheritable.

Any arguments given to C<new> are passed straight to C<mk_accessors>.

If you want to be able to assign to your accessors as if they
were C<lvalue>s, you should create your object in the 
C<Object::Acccessor::Lvalue> namespace instead. See the section
on C<LVALUE ACCESSORS> below.

=cut

sub new {
    my $class   = shift;
    my $obj     = bless {}, $class;
    
    $obj->mk_accessors( @_ ) if @_;
    
    return $obj;
}

=head2 $bool = $object->mk_accessors( @ACCESSORS | \%ACCESSOR_MAP );

Creates a list of accessors for this object (and C<NOT> for other ones
in the same class!).
Will not clobber existing data, so if an accessor already exists,
requesting to create again is effectively a C<no-op>.

When providing a C<hashref> as argument, rather than a normal list,
you can specify a list of key/value pairs of accessors and their
respective input validators. The validators can be anything that
C<Params::Check>'s C<allow> function accepts. Please see its manpage
for details.

For example:

    $object->mk_accessors( {
        foo     => qr/^\d+$/,       # digits only
        bar     => [0,1],           # booleans
        zot     => \&my_sub         # a custom verification sub
    } );        

Returns true on success, false on failure.

Accessors that are called on an object, that do not exist return
C<undef> by default, but you can make this a fatal error by setting the
global variable C<$FATAL> to true. See the section on C<GLOBAL
VARIABLES> for details.

Note that you can bind the values of attributes to a scope. This allows
you to C<temporarily> change a value of an attribute, and have it's 
original value restored up on the end of it's bound variable's scope;

For example, in this snippet of code, the attribute C<foo> will 
temporarily be set to C<2>, until the end of the scope of C<$x>, at 
which point the original value of C<1> will be restored.

    my $obj = Object::Accessor->new;
    
    $obj->mk_accessors('foo');
    $obj->foo( 1 );
    print $obj->foo;                # will print 1

    ### bind the scope of the value of attribute 'foo'
    ### to the scope of '$x' -- when $x goes out of 
    ### scope, 'foo' previous value will be restored
    {   $obj->foo( 2 => \my $x );
        print $obj->foo, ' ', $x;   # will print '2 2'
    }
    print $obj->foo;                # will print 1
    

Note that all accessors are read/write for everyone. See the C<TODO>
section for details.

=cut

sub mk_accessors {
    my $self    = $_[0];
    my $is_hash = UNIVERSAL::isa( $_[1], 'HASH' );
    
    ### first argument is a hashref, which means key/val pairs
    ### as keys + allow handlers
    for my $acc ( $is_hash ? keys %{$_[1]} : @_[1..$#_] ) {
    
        ### already created apparently
        if( exists $self->{$acc} ) {
            __PACKAGE__->___debug( "Accessor '$acc' already exists");
            next;
        }

        __PACKAGE__->___debug( "Creating accessor '$acc'");

        ### explicitly vivify it, so that exists works in ls_accessors()
        $self->{$acc}->[VALUE] = undef;
        
        ### set the allow handler only if one was specified
        $self->{$acc}->[ALLOW] = $_[1]->{$acc} if $is_hash;
    }

    return 1;
}

=head2 @list = $self->ls_accessors;

Returns a list of accessors that are supported by the current object.
The corresponding coderefs can be retrieved by passing this list one
by one to the C<can> method.

=cut

sub ls_accessors {
    ### metainformation is stored in the stringified 
    ### key of the object, so skip that when listing accessors
    return sort grep { $_ ne "$_[0]" } keys %{$_[0]};
}

=head2 $ref = $self->ls_allow(KEY)

Returns the allow handler for the given key, which can be used with
C<Params::Check>'s C<allow()> handler. If there was no allow handler
specified, an allow handler that always returns true will be returned.

=cut

sub ls_allow {
    my $self = shift;
    my $key  = shift or return;
    return exists $self->{$key}->[ALLOW]
                ? $self->{$key}->[ALLOW] 
                : sub { 1 };
}

=head2 $clone = $self->mk_clone;

Makes a clone of the current object, which will have the exact same
accessors as the current object, but without the data stored in them.

=cut

### XXX this creates an object WITH allow handlers at all times.
### even if the original didnt
sub mk_clone {
    my $self    = $_[0];
    my $class   = ref $self;

    my $clone   = $class->new;
    
    ### split out accessors with and without allow handlers, so we
    ### don't install dummy allow handers (which makes O::A::lvalue
    ### warn for exampel)
    my %hash; my @list;
    for my $acc ( $self->ls_accessors ) {
        my $allow = $self->{$acc}->[ALLOW];
        $allow ? $hash{$acc} = $allow : push @list, $acc;
    }

    ### copy the accessors from $self to $clone
    $clone->mk_accessors( \%hash ) if %hash;
    $clone->mk_accessors( @list  ) if @list;

    ### copy callbacks
    #$clone->{"$clone"} = $self->{"$self"} if $self->{"$self"};
    $clone->___callback( $self->___callback );

    return $clone;
}

=head2 $bool = $self->mk_flush;

Flushes all the data from the current object; all accessors will be
set back to their default state of C<undef>.

Returns true on success and false on failure.

=cut

sub mk_flush {
    my $self = $_[0];

    # set each accessor's data to undef
    $self->{$_}->[VALUE] = undef for $self->ls_accessors;

    return 1;
}

=head2 $bool = $self->mk_verify;

Checks if all values in the current object are in accordance with their
own allow handler. Specifically useful to check if an empty initialised
object has been filled with values satisfying their own allow criteria.

=cut

sub mk_verify {
    my $self = $_[0];
    
    my $fail;
    for my $name ( $self->ls_accessors ) {
        unless( allow( $self->$name, $self->ls_allow( $name ) ) ) {
            my $val = defined $self->$name ? $self->$name : '<undef>';

            __PACKAGE__->___error("'$name' ($val) is invalid");
            $fail++;
        }
    }

    return if $fail;
    return 1;
}   

=head2 $bool = $self->register_callback( sub { ... } );

This method allows you to register a callback, that is invoked
every time an accessor is called. This allows you to munge input
data, access external data stores, etc.

You are free to return whatever you wish. On a C<set> call, the
data is even stored in the object.

Below is an example of the use of a callback.
    
    $object->some_method( "some_value" );
    
    my $callback = sub {
        my $self    = shift; # the object
        my $meth    = shift; # "some_method"
        my $val     = shift; # ["some_value"]  
                             # could be undef -- check 'exists';
                             # if scalar @$val is empty, it was a 'get'
    
        # your code here

        return $new_val;     # the value you want to be set/returned
    }        

To access the values stored in the object, circumventing the
callback structure, you should use the C<___get> and C<___set> methods
documented further down. 

=cut

sub register_callback {
    my $self    = shift;
    my $sub     = shift or return;
    
    ### use the memory address as key, it's not used EVER as an
    ### accessor --kane
    $self->___callback( $sub );

    return 1;
}


=head2 $bool = $self->can( METHOD_NAME )

This method overrides C<UNIVERAL::can> in order to provide coderefs to
accessors which are loaded on demand. It will behave just like
C<UNIVERSAL::can> where it can -- returning a class method if it exists,
or a closure pointing to a valid accessor of this particular object.

You can use it as follows:

    $sub = $object->can('some_accessor');   # retrieve the coderef
    $sub->('foo');                          # 'some_accessor' now set
                                            # to 'foo' for $object
    $foo = $sub->();                        # retrieve the contents