NAME

Oryx - Meta-Model Driven Object Persistance with Multiple Inheritance

SYNOPSIS

#===========================================================================
# connect to storage
$storage = Oryx->connect(['dbi:Pg:dbname=cms', $usname, $passwd]);

# or specify a schema
$storage = Oryx->connect(
   ["dbi:Pg:dbname=cms", $usname, $passwd], 'CMS::Schema'
);

# for DBM::Deep back-end
Oryx->connect(['dbm:Deep:datapath=/path/to/data'], 'CMS::Schema');

$storage->deployClass('CMS::Page');
$storage->deploySchema('CMS::Schema');

#===========================================================================
# deploy the schema
$storage->deploySchema();

# or automatically build tables as needed
Oryx::Class->auto_deploy(1);

#===========================================================================
# define a persistent class
package CMS::Page;
use base qw(Oryx::Class);

# ... class meta-data here (see DEFINING CLASS META-DATA below) ...

1;

#===========================================================================
# use a persistent class
use CMS::Page;

$page = CMS::Page->create({title => 'Life in the Metaverse'});
$page = CMS::Page->retrieve($id);

$page->update;
$page->delete;

@pages = CMS::Page->search({author => 'Richard Hun%'});

#===========================================================================
# commit your changes
$page->dbh->commit; # or simply ...
$page->commit;

#===========================================================================
# attribute mutator
$page->title('The Metamanic Mechanic');
$tite = $page->title;

#===========================================================================
# reference association mutator
$template_obj = $page->template;
$page->template( $template_obj );

#===========================================================================
# array association accessor
$page->paragraphs->[0] = $intro_para;
$paragraph = $page->paragraphs->[42];

#===========================================================================
# array association operators
$concl = pop   @{$page->paragraphs};
$intro = shift @{$page->paragraphs};
push    @{$page->paragraphs}, $concl;
unshift @{$page->paragraphs}, $new_intro;
splice  @{$page->paragraphs}, 1, 4, ($summary);

#===========================================================================
# hash association accessor
$image_obj = $page->images->{logo};
$page->images->{mug_shot} = $my_ugly_mug;
@keys   = keys   %{$page->images};
@values = values %{$page->images};

DESCRIPTION

Oryx is an Object Persistence framework which is not coupled to any particular storage back-end. In other words, you should be able to swap out an RDMBS with a DBM style database (and vice versa) without changing your persistent classes at all.

This is achieved with the use a meta model which fits in as closely with Perl's own as possible - and due to Perl's excellent introspection capabilities and enormous flexibility - this is very close indeed. For this reason Hash, Array and Reference association types are implemented with liberal use of `tie'. The use of a meta model, albeit a very transparent one, conceptually supports the de-coupling of storage back-end from persistent classes, and, for the most part, beside a really small amout of meta-data, you would use persistent classes in a way that is virtually indistinguishable from ordinary perl classes.

INTRODUCTION

This documentation applies to classes persisted in DBM::Deep style storage as well except insofar as the implementation details are concerned where tables and columns are mentioned - separate files are used for DBM::Deep based storage instead of tables (see Oryx::DBM for details).

This is still a very early release and only supports DBM::Deep and Postgres back-ends at the moment. Having said this, however, Oryx is already quite usable. It needs to be thrashed a lot more and support for the rest of the popular RDBMS needs to be added. Things will change (for the better, one hopes); if you're interested in helping to precipitate that change... let me know, you'd be most welcome.

CREATING PERSISTENT CLASSES

Creating persistent classes is simple, there is no need to create any database tables by hand as the DB schema is deployed automatically as needed (see "AUTOMATIC TABLE CREATION" below).

The following three steps illustrate how this is done:

Inherit from Oryx::Class or subclass thereof (see "INHERITANCE" below):

package CMS::Page;
use base qw(Oryx::Class);

Define meta-data (see "DEFINING CLASS META-DATA" below):

our $schema = {
    attributes => [{
        name  => 'title',
        type  => 'String',
    },{
        name  => 'author',
        type  => 'String',
    },{
        name  => 'number',
        type  => 'Integer',
    }],
    associations => [{
        role  => 'paragraphs',
        class => 'CMS::Paragraph',
        type  => 'Array',
    },{
        role  => 'template',
        class => 'CMS::Template',
        type  => 'Reference',
    }],
};

1;

Connect to storage (see "CONNECTING TO STORAGE" below):

...far away in another piece of code...

use CMS::Page;

use Oryx;
Oryx->connect(["dbi:Pg:dbname=cms", $usname, $passwd]);

...

Now we're ready to start using persistent CMS::Page objects (and friends).

CREATING AND USING OBJECTS

Oryx::Class defines a create method (see Oryx::Class for more) which takes a hash reference as a constructor for setting up the object's initial state:

use CMS::Page;
my $page = CMS::Schema::Page->create({
    title  => 'Meta Model Mania',
    author => 'Sam Vilain',
});

Once an object has been instatiated, attribute mutators can be used to get and set attributes on the object (see "ATTRIBUTES" below):

$page->number(42);

Associations are similar except that we associate one object with another (see "ASSOCIATIONS" below), so we create an instance of the target class:

my $paragraph1 = CMS::Paragraph->create({
    content => $some_block_of_text,
});

And then, because the association mutator returns a reference to a tied object (an ARRAY in this case), we can:

$page->paragraphs->[0] = $paragraph1;

Then update your object when done:

$page->update;

Or if you no longer need it:

$page->delete;

Finally, commit your changes:

$page->commit;

DEFINING CLASS META-DATA

Three ways of defining meta data for your persistent classes are supported as follows:

Tangram style using a $schema class variable:

package CMS::Page;
use base qw(Oryx::Class);

our $schema = {
    attributes => [{
        name  => 'title',
        type  => 'String',
    },{
        name  => 'author',
        type  => 'String',
    }],
    associations => [{
        role  => 'paragraphs',
        class => 'CMS::Paragraph',
        type  => 'Array',
    },{
        role  => 'template',
        class => 'CMS::Template',
        type  => 'Reference',
    }],
};

1;

Class::DBI style adding members dynamically:

package CMS::Paragraph;
use base qw(Oryx::Class);

__PACKAGE__->addAttribute({
    name  => 'content',
    type  => 'Text',
});

__PACKAGE__->addAttribute({
    name  => 'formatted',
    type  => 'Boolean',
});

__PACKAGE__->addAssociation({
    role  => 'images',
    class => 'CMS::Image',
    type  => 'Hash',
});

1;

If you have XML::DOM::Lite, put it in the DATA section:

package CMS::Image;
use base qw(Oryx::Class);

1;
__DATA__
<Class>
  <Attribute name="alt_text" type="String" />
  <Attribute name="path_to_file" type="String" />
</Class>

AUTOMATIC TABLE CREATION

With Oryx, you never need to write a single line of SQL although you can if you want to in exactly the same way as you would when using Class::DBI (actually it's a ImA::DBI feature). Tables are named sensibly as pluralised versions of the class name with link table names equally intuitive.

Enabling auto_deploy for all classes

To enable automatic table creation, you need to do the following near the top of your application before you use any of your classes:

use Oryx::Class;
Oryx::Class->auto_deploy( 1 );

Because the check to see if a table exists is made once when the class is first use'ed, the performance penalty for this is minimal in long running process environments such as mod perl. Otherwise when running in an environment where your code is recompiled each time the program is run, or you would like more control, you can leave auto_deploy turned off at the top level (which it is by default) and simply turn it on for each new class that you're adding to the schema as this method is inherited.

ATTRIBUTES

Attributes are declared as having a name and a type and as such are simply tied Oryx::Value derivatives (see Oryx::Value for details) which are generally associated with a field (or column) in the underlying database, and which have mutators which are automatically created in the class for getting and setting these values.

Certain attributes may also be declared with additional properties as relevant, for instance, attributes declared as type => "Float" support a precision property which describes the valid number of decimal places.

Attribute Value Input Checking

Input is checked when assigning values to attributes and return values are cast to the correct type using a combination of regular expressions, the Data::Types module, YAML or Class::Date where relevant. Where additional properties are set such as size or precision, these are checked also and your program will croak if types mismatch or overflow.

Supported Attribute Value Types

Several basic value data types are supported:

String

Varying character type (VARCHAR for most RDBMS). Input is checked using Data::Types::is_string and if the attribute is declared with a size property, the length is also checked.

Text

Corresponds to a SQL TEXT type; type checking is done using Data::Types::is_string, but no length checking is performed.

Boolean

Corresponds to a a SQL TINYINT or INT type and is checked for the values 0 or 1.

Binary

No checking is done here, but a BLOB or BYTEA or equivalent column type is created when the class is deployed.

Complex

This can be anything that can be (de)serialized using YAML and is stored internally in the DB in a column with a TEXT type.

DateTime

Uses Class::Date objects. You can pass either a Class::Date instance to the mutator as follows:

use Class::Date qw(date);
$page->date_created( date(localtime) );

or any value which is valid input to the Class::Date::new constructor this includes ARRAY refs etc. (see Class::Date for details).

Attributes declared as DateTime types additionaly support a format property which is used to set Class::Date::DATE_FORMAT for date formatting.

Float

Floating point number checked using Data::Types::is_float. Return value is done with Data::Types::to_float and precision checks are made if the attribute is declared with such.

Integer

Corresponds to INT or INTEGER SQL type. Input checks are performed using Data::Types::is_int.

Oid

This is also an integer type, but with the distinction that when a class is deployed to an RDBMS the column is constrained as a PRIMARY KEY.

ASSOCIATIONS

Oryx implements the three most common ways in which associations between classes can be achieved natively with Perl. An object can be associated with another by simple reference, or we can use either ordered (ARRAY), or keyed (HASH) associations - so a field in one object (usually a blessed HASH reference) can be an ARRAY reference, for example, which could be filled with references to other objects (which themselves are persistent).

In RDBMS terms, this sort of to-many ordered relationship requires a link table with a column holding ordering information, which is exactly what happens under the hood, but Oryx makes it transparent for you using Perl's tie mechanism while managing the link table automagically.

Furthermore one can also have to-many ordered (Array) or to-many keyed (Hash) associations which are mixed - in other words one class can have an ARRAY (or HASH) reference which can contain instances of different classes (see "ABSTRACT CLASSES" below).

Reference

getting :

my $a_template = $page->template;

setting :

$page->template($another_template);

Array

getting :

my $para42 = $page->paragraphs->[42];

setting :

$page->paragraph->[0] = $intro_para;

as well as all the usual push, pop, shift, unshift and splice.

Hash

getting :

my $image_obj = $page->images->{logo};

setting :

$page->images->{mug_shot} = $my_ugly_mug;

RETRIEVING AND SEARCHING FOR OBJECTS

Retrieval is simple, just pass in the id (primary key) :

my $page = CMS::Page->retrieve($page_id);

Searching uses 'LIKE' (assuming an RDBMS storage) :

my @pages = CMS::Page->search({ author => '%Hundt%'});

NOTE : Searches don't search through superclass fields yet...

INHERITANCE

Inheritance works as you would expect.

So if we have the following :

package CMS::Section;
use base qw(Oryx::Class);

# ... schema definition here ...

1;

package CMS::Paragraph;
use base qw(CMS::Section);

# ... schema definition here ...

1;

You get exactly what you would normally get in Perl, that is :

UNIVERSAL::isa('CMS::Paragraph', 'Oryx::Class')

holds true and attributes and associations defined in CMS::Section are available to CMS::Paragraph instances. So any class which has persistant class as an ancestor, can be treated and persisted in the same way as the ancestor. However, it is important to note that it gets its own table in the database.

For multiple persistent base classes :

package Orange;
use base qw(Food Fruit);

As long as Food and Fruit are Oryx::Class derivatives, the Force That Into the Database Drives the Object will make sure the proverbial Right Thing is Done.

Oryx uses a multiple table inheritance model (as opposed to putting all the instances for classes in an inheritance chain into the same table), each subclass instance has a corresponding superclass instance for each superclass (assuming said superclass is a derivative of Oryx::Class), so that attributes which exists in the superclass are stored (as a row) in the superclass' table, and are therefore fully fledged instances of the superclass.

You can access these superclass instances with the PARENT method as follows:

my $parent_section_instance = $paragraph->PARENT('CMS::Section');

and then use this instance normally.

Updates and deletes cascade up the inheritance chain, as you'd expect.

ABSTRACT CLASSES

Abstract classes to Oryx are simply classes which do not define any attributes, but may have associations. The effect is automatic.

Abstract classes behave slightly differently to concrete classes (which define attributes) in that if you retrieve an instance of an abstract class (by id or by accessing a member of an association), you get an instance of the sub class (the one which created the row in the abstract class's table).

This is particularly useful where you have an Array or Hash association between two classes and need to mix instances of different types in that association. As long as all the members of the array (or hash) inherit from the same abstract class, accessing them produces the expected result.

Consider the following case :

                   <ABSTRACT>
+------+  <Array> +----------+
| Page |----------| Fragment |
+------+  frags   +----------+
|______|          |__________|
                       /_\
                        |
              +---------+------+
              |                |
        +-----------+      +-------+
        | Paragraph |      | Image |
        +-----------+      +-------+
        |___________|      |_______|

Here the Paragraph and Image both inherit from the abstract Fragment class. When the frags Array association is accessed it may contain a mixture of both Paragraph and Image instances.

Thus you can say:

$my_para = Paragraph->create({ ... });
$my_page->frags->[42] = $my_para;

$my_img = Image->create({ ... });
$my_page->frags->[69] = $my_img;

pretty neat huh?

CONNECTING TO STORAGE

The call to Oryx->connect(...) specifies the dsn and connection credentials to use when connecting where applicable. For DBM::Deep style storage, the connection arguments look like this:

Oryx->connect(['dbm:Deep:datapath=/path/to/data'], 'CMS::Schema');

For RDBMS (Postgres in this case) it may look like this:

Oryx->connect(["dbi:Pg:dbname=cms", $usname, $passwd], 'CMS::Schema');

The Schema defaults to 'Oryx::Schema' and is therefore optional, so we could say Oryx->connect([ ... dsn ... ]), and forget about passing in a Schema.

One advantage to using separate Schema classes is that this gives you namespace separation where you need to connect several sets of classes to different storage back ends (especially where these are mixed in such a way where the same classes exist in different stores). Another advantage is that the Schema class may define a prefix method which simply returns a string to prefix table names with, for those of us who only get a single database with our hosting package and need to have some namespace separation.

Here's an example of a Schema class :

package CMS::Schema;
use base qw(Oryx::Schema);

sub prefix { 'cms' }

1;

OBJECT CACHING

In the interest of consistency, objects are cached and are unique in memory. Therefore, if you retrieve an object more than once, each subsequent retrieve until the reference count on it has dropped to zero and has been eaten by the garbage collector, will return a reference to the same object.

This has a performance gain in certain situations too.

TODO

test test test: Tests are a bit sparse at the moment.
Support for MySQL, Oracle, etc.: Only PostgreSQL and DBM::Deep are supported currently. It should be fairly trivial to add support for the other RDBMS' - MySQL doesn't have sequences so we'll probably have to create an additional table to manage sequences, c'est la vie.
More documentation: References are made to seeing also Oryx::Class and Oryx::Value in this document, but they don't have any pod yet

BUGS

I'm sure there are some... if I had more tests, I know I'd find 'em

ACKNOWLEDGEMENTS

Special thanks to:

Sam Vilain; Aaron Trevena

AUTHOR

LICENSE

Oryx may be used under the same terms as Perl itself.

To install Oryx, copy and paste the appropriate command in to your terminal.

cpanm

cpanm Oryx

CPAN shell

perl -MCPAN -e shell
install Oryx

For more information on module installation, please visit the detailed CPAN module installation guide.

	Global
`s`	Focus search bar
`?`	Bring up this help dialog

	GitHub
`g` `p`	Go to pull requests
`g` `i`	go to github issues (only if github is preferred repository)

	POD
`g` `a`	Go to author
`g` `c`	Go to changes
`g` `i`	Go to issues
`g` `d`	Go to dist
`g` `r`	Go to repository/SCM
`g` `s`	Go to source
`g` `b`	Go to file browse

	Search terms
module: (e.g. module:Plugin)
distribution: (e.g. distribution:Dancer auth)
author: (e.g. author:SONGMU Redis)
version: (e.g. version:1.00)

NAME

SYNOPSIS

DESCRIPTION

INTRODUCTION

CREATING PERSISTENT CLASSES

CREATING AND USING OBJECTS

DEFINING CLASS META-DATA

AUTOMATIC TABLE CREATION

Enabling auto_deploy for all classes

ATTRIBUTES

Attribute Value Input Checking

Supported Attribute Value Types

ASSOCIATIONS

Reference

Array

Hash

RETRIEVING AND SEARCHING FOR OBJECTS

INHERITANCE

ABSTRACT CLASSES

CONNECTING TO STORAGE

OBJECT CACHING

TODO

BUGS

ACKNOWLEDGEMENTS

SEE ALSO

AUTHOR

LICENSE

Module Install Instructions