package MyLazyObject;
use Object::Realize::Later
becomes => 'MyRealObject',
realize => 'load';
On demand realization is all about performance gain. Why should you spent costly time on realizing an object, when the data on the object is never (or not yet) used? In interactive programs, postponed realization may boost start-up: the realization of objects is triggered by the use, so spread over time.
-Option --Default becomes <required> believe_caller <false> realize <required> source_module <becomes> warn_realization <false> warn_realize_again <false>
After realization, you may still have your hands on the lazy object on various places. Be sure that your realization method is coping with that, for instance by using Memoize. See examples below.
example:
MyLazyObject->can('lazyWork') # true
MyLazyObject->can('realWork') # true
my $lazy = MyLazyObject->new;
$lazy->can('lazyWork'); # true
$lazy->can('realWork'); # true
example:
MyLazyObject->isa('MyRealObject') # true
MyLazyObject->isa('SuperClassOfLazy'); # true
MyLazyObject->isa('SuperClassOfReal'); # true
my $lazy = MyLazyObject->new;
$lazy->isa('MyRealObject'); # true
$lazy->isa('SuperClassOfLazy'); # true
$lazy->isa('SuperClassOfReal'); # true
An implementation of the first solution is:
sub realize {
my $self = shift;
return $self unless $self->{_is_realized};
# read the data from file, or whatever
$self->{data} = ....;
$self->{_is_realized} = 1;
$self;
}
sub getData() {
my $self = shift;
return $self->realize->{data};
}
The above implementation is error-prone, where you can easily forget to call realize(). The tests cannot cover all ordenings of method-calls to detect the mistakes.
The second approach uses autoloading, and is supported by this package. First we create a stub-object, which will be transformable into a realized object later. This transformation is triggered by AUTOLOAD.
This stub-object may contain some methods from the realized object, to reduce the need for realization. The stub will also contain some information which is required for the creation of the real object.
"Object::Realize::Later" solves the inheritance problems (especially the isa() and can() methods) and supplies the AUTOLOAD method. Class methods which are not defined in the stub object are forwarded as class methods without realization.
The two versions of realization:
This is the safe version. The realized object is the same object as the delayed one, but reblessed in a different package. When multiple references to the delayed object exists, they will all be updated at the same, because the bless information is stored within the refered variable.
This is the nicest way of realization, but also quite more dangerous. Consider this:
package Delayed;
use Object::Realize::Later
becomes => 'Realized',
realize => 'load';
sub new($) {my($class,$v)=@_; bless {label=>$v}, $class}
sub setLabel($) {my $self = shift; $self->{label} = shift}
sub load() {$_[0] = Realized->new($_[0]->{label}) }
package Realized; # file Realized.pm or use use(source_module)
sub new($) {my($class,$v)=@_; bless {label=>$v}, $class}
sub setLabel($) {my $self = shift; $self->{label} = shift}
sub getLabel() {my $self = shift; $self->{label}}
package main;
my $original = Delayed->new('original');
my $copy = $original;
print $original->getLabel; # prints 'original'
print ref $original; # prints 'Realized'
print ref $copy; # prints 'Delayed'
$original->setLabel('changed');
print $original->getLabel; # prints 'changed'
print $copy->getLabel; # prints 'original'
In the first example, we delay-load a message. On the moment the message is defined, we only take the location. When the data of the message is taken (header or body), the data is autoloaded.
package Mail::Message::Delayed;
use Object::Realize::Later
( becomes => 'Mail::Message::Real'
, realize => 'loadMessage'
);
sub new($) {
my ($class, $file) = @_;
bless { filename => $file }, $class;
}
sub loadMessage() {
my $self = shift;
Mail::Message::Real->new($self->{filename});
}
In the main program:
package main;
use Mail::Message::Delayed;
my $msg = Mail::Message::Delayed->new('/home/user/mh/1');
$msg->body->print; # this will trigger autoload.
Example 2
Your realization may also be done by reblessing. In that case to change the type of your object into a different type which stores the same information. Is that right? Are you sure? For simple cases, this may be possible:
package Alive;
use Object::Realize::Later
becomes => 'Dead',
realize => 'kill';
sub new() {my $class = shift; bless {@_}, $class}
sub jump() {print "Jump!\n"}
sub showAntlers() {print "Fight!\n"}
sub kill() {bless(shift, 'Dead')}
package Dead;
sub takeAntlers() {...}
In the main program:
my $deer = Alive->new(Animal => 'deer'); my $trophy = $deer->takeAntlers();
In this situation, the object (reference) is not changed but is reblessed. There is no danger that the un-realized version of the object is kept somewhere: all variable which know about this partical deer see the change.
Example 3
This module is especially useful for larger projects, which there is a need for speed or memory reduction. In this case, you may have an extra overview on which objects have been realized (transformed), and which not. This example is taken from the MailBox modules:
The Mail::Box module tries to boost the access-time to a folder. If you only need the messages of the last day, why shall all be read? So, MailBox only creates an invertory of messages at first. It takes the headers of all messages, but leaves the body (content) of the message in the file.
In MailBox' case, the Mail::Message-object has the choice between a number of Mail::Message::Body's, one of which has only be prepared to read the body when needed. A code snippet:
package Mail::Message;
sub new($$)
{ my ($class, $head, $body) = @_;
my $self = bless {head => $head, body => $body}, $class;
$body->message($self); # tell body about the message
}
sub head() { shift->{head} }
sub body() { shift->{body} }
sub loadBody()
{ my $self = shift;
my $body = $self->body;
# Catch re-invocations of the loading. If anywhere was still
# a reference to the old (unrealized) body of this message, we
# return the new-one directly.
return $body unless $body->can('forceRealize');
# Load the body (change it to anything which really is of
# the promised type, or a sub-class of it.
my ($lines, $size) = .......; # get the data
$self->{body} = Mail::Message::Body::Lines
->new($lines, $size, $self);
# Return the realized object.
return $self->{body};
}
package Mail::Message::Body::Lines;
use base 'Mail::Message::Body';
sub new($$$)
{ my ($class, $lines, $size, $message) = @_;
bless { lines => $lines, size => $size
, message => $message }, $class;
}
sub size() { shift->{size} }
sub lines() { shift->{lines} }
sub message() { shift->{message);
package Mail::Message::Body::Delayed;
use Object::Realize::Later
becomes => 'Mail::Message::Body',
realize => sub {shift->message->loadBody};
sub new($)
{ my ($class, $size) = @_;
bless {size => $size}, $class;
}
sub size() { shift->{size} }
sub message(;$)
{ my $self = shift;
@_ ? ($self->{message} = shift) : $self->{messages};
}
package main;
use Mail::Message;
use Mail::Message::Body::Delayed;
my $body = Mail::Message::Body::Delayed->new(42);
my $message = Mail::Message->new($head, $body);
print $message->size; # will not trigger realization!
print $message->can('lines'); # true, but no realization yet.
print $message->lines; # realizes automatically.
This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself. See http://dev.perl.org/licenses/