NAME
MooseX::Storage - A serialization framework for Moose classes
VERSION
version 0.53
SYNOPSIS
package Point;
use Moose;
use MooseX::Storage;
with Storage('format' => 'JSON', 'io' => 'File');
has 'x' => (is => 'rw', isa => 'Int');
has 'y' => (is => 'rw', isa => 'Int');
1;
my $p = Point->new(x => 10, y => 10);
## methods to pack/unpack an
## object in perl data structures
# pack the class into a hash
$p->pack(); # { __CLASS__ => 'Point-0.01', x => 10, y => 10 }
# unpack the hash into a class
my $p2 = Point->unpack({ __CLASS__ => 'Point-0.01', x => 10, y => 10 });
## methods to freeze/thaw into
## a specified serialization format
## (in this case JSON)
# pack the class into a JSON string
$p->freeze(); # { "__CLASS__" : "Point-0.01", "x" : 10, "y" : 10 }
# unpack the JSON string into a class
my $p2 = Point->thaw('{ "__CLASS__" : "Point-0.01", "x" : 10, "y" : 10 }');
## methods to load/store a class
## on the file system
$p->store('my_point.json');
my $p2 = Point->load('my_point.json');
DESCRIPTION
MooseX::Storage is a serialization framework for Moose, it provides a very flexible and highly pluggable way to serialize Moose classes to a number of different formats and styles.
Levels of Serialization
There are three levels to the serialization, each of which builds upon the other and each of which can be customized to the specific needs of your class.
- base
-
The first (base) level is
pack
andunpack
. In this level the class is serialized into a Perl HASH reference, it is tagged with the class name and each instance attribute is stored. Very simple.This level is not optional, it is the bare minimum that MooseX::Storage provides and all other levels build on top of this.
See MooseX::Storage::Basic for the fundamental implementation and options to
pack
andunpack
- format
-
The second (format) level is
freeze
andthaw
. In this level the output ofpack
is sent tofreeze
or the output ofthaw
is sent tounpack
. This levels primary role is to convert to and from the specific serialization format and Perl land.This level is optional, if you don't want/need it, you don't have to have it. You can just use
pack
/unpack
instead. - io
-
The third (io) level is
load
andstore
. In this level we are reading and writing data to file/network/database/etc.This level is also optional, in most cases it does require a
format
role to also be used, the exception being theStorableFile
role.
Behaviour modifiers
The serialization behaviour can be changed by supplying traits
to either the class or an individual attribute.
This can be done as follows:
use MooseX::Storage;
# adjust behaviour for the entire class
with Storage( traits => [Trait1, Trait2,...] );
# adjust behaviour for an attribute
has my_attr => (
traits => [Trait1, Trait2, ...],
...
);
The following class traits are currently bundled with MooseX::Storage:
- OnlyWhenBuilt
-
Only attributes that have been built (i.e., where the predicate returns 'true') will be serialized. This avoids any potentially expensive computations.
See MooseX::Storage::Traits::OnlyWhenBuilt for details.
- DisableCycleDetection
-
Disables the default checks for circular references, which is necessary if you use such references in your serialisable objects.
See MooseX::Storage::Traits::DisableCycleDetection for details.
The following attribute traits are currently bundled with MooseX::Storage:
- DoNotSerialize
-
Skip serialization entirely for this attribute.
See MooseX::Storage::Meta::Attribute::Trait::DoNotSerialize for details.
How we serialize
There are always limits to any serialization framework -- there are just some things which are really difficult to serialize properly and some things which cannot be serialized at all.
What can be serialized?
Currently only numbers, string, ARRAY refs, HASH refs and other MooseX::Storage-enabled objects are supported.
With Array and Hash references the first level down is inspected and any objects found are serialized/deserialized for you. We do not do this recursively by default, however this feature may become an option eventually.
The specific serialize/deserialize routine is determined by the Moose type constraint a specific attribute has. In most cases subtypes of the supported types are handled correctly, and there is a facility for adding handlers for custom types as well. This will get documented eventually, but it is currently still in development.
What can not be serialized?
We do not support CODE references yet, but this support might be added in using B::Deparse or some other deep magic.
Scalar refs are not supported, mostly because there is no way to know if the value being referenced will be there when the object is inflated. I highly doubt will be ever support this in a general sense, but it would be possible to add this yourself for a small specific case.
Circular references are specifically disallowed, however if you break the cycles yourself then re-assemble them later you can get around this. The reason we disallow circular refs is because they are not always supported in all formats we use, and they tend to be very tricky to do for all possible cases. It is almost always something you want to have tight control over anyway.
CAVEAT
This is not a persistence framework; changes to your object after you load or store it will not be reflected in the stored class.
EXPORTS
- Storage (%options)
-
This module will export the
Storage
method and can be used to load a specific set of MooseX::Storage roles to implement a specific combination of features. It is meant to make things easier, but it is by no means the only way. You can still compose your roles by hand if you like.By default, options are assumed to be short forms. For example, this:
Storage(format => 'JSON');
...will result in looking for MooseX::Storage::Format::JSON. To use a role that is not under the default namespace prefix, start with an equal sign:
Storage(format => '=My::Private::JSONFormat');
To use a parameterized role (for which, see MooseX::Role::Parameterized) you can pass an arrayref of the role name (in short or long form, as above) and its parameters:
Storage(format => [ JSONpm => { json_opts => { pretty => 1 } } ]);
METHODS
- import
Introspection
- meta
TODO
This module needs docs and probably a Cookbook of some kind as well. This is an early release, so that is my excuse for now :)
For the time being, please read the tests and feel free to email me if you have any questions. This module can also be discussed on IRC in the #moose channel on irc.perl.org.
SUPPORT
Bugs may be submitted through the RT bug tracker (or bug-MooseX-Storage@rt.cpan.org).
There is also a mailing list available for users of this distribution, at http://lists.perl.org/list/moose.html.
There is also an irc channel available for users of this distribution, at #moose
on irc.perl.org
.
AUTHORS
Chris Prather <chris.prather@iinteractive.com>
Stevan Little <stevan.little@iinteractive.com>
יובל קוג'מן (Yuval Kogman) <nothingmuch@woobling.org>
CONTRIBUTORS
Karen Etheridge <ether@cpan.org>
Tomas Doran <bobtfish@bobtfish.net>
Ricardo Signes <rjbs@cpan.org>
Chris Prather <chris@prather.org>
Jos Boumans <jos@dwim.org>
Shawn M Moore <sartak@gmail.com>
Jonathan Yu <frequency@cpan.org>
Dmitry Latin <dim0xff@gmail.com>
Robert Boone <robo4288@gmail.com>
Dagfinn Ilmari Mannsåker <ilmari@ilmari.org>
Cory Watson <gphat@Crankwizzah.local>
sillitoe <ian@sillit.com>
Dan Brook <dan@broquaint.com>
David Golden <dagolden@cpan.org>
David Steinbrunner <dsteinbrunner@pobox.com>
Florian Ragwitz <rafl@debian.org>
Graham Knop <haarg@haarg.org>
Jason Pope <cowholio4@gmail.com>
Johannes Plunien <plu@pqpq.de>
Jonathan Rockway <jon@jrock.us>
Yanick Champoux <yanick@babyl.dyndns.org>
COPYRIGHT AND LICENSE
This software is copyright (c) 2007 by Infinity Interactive, Inc.
This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.