=head1 NAME

Maypole::Manual::Inheritance - structure of a Maypole application


Discusses the inheritance structure of a basic and a more advanced Maypole

=over 4

=item inheritance

     +-   -+
=item notes

    target *-------- note about the target

=item association

    source ------> target


=head1 Structure of a standard Maypole application

A minimal Maypole application (such as the Beer database example from the
L<Maypole> synopsis) consists of a custom driver (or controller) class (BeerDB.pm),
a set of auto-generated model classes, and a view class:

           THE DRIVER
                                          +------- init() is a factory method,
                   1      Maypole         |           it sets up the view
   Maypole::Config <----- config();       |              classes
   model();               init(); *-------+                           THE VIEW
    |                     view_object(); -------+
    |    +--------------* setup();              |      Maypole::View::Base
    |    |                   +                  |              +
    |    |                   |                  |     1        |
    |    |    PLUGINS    Apache::MVC *-----+    +-----> Maypole::View::TT
    |    |       +           +             |             (or another view class)
    |    |       |           |             |
    |    |       +-----+-----+             |
    |    |             |                   |
    |    |           BeerDB                +----- or CGI::Maypole
    |    |                                         or MasonX:::Maypole
    |    |
    |   setup() is a factory method,
    |     it sets up the model
    |         classes
    |                                             THE MODEL
    |  Maypole::Model::Base    Class::DBI
    |             +             +      +
    |             |             |      |
    +-------> Maypole::Model::CDBI   Class::DBI::<db_driver>
                      +                     +
                      |                     |
           |            |        |       |         |
       BeerDB::Pub      |   BeerDB::Beer | BeerDB::Brewery
       beers();         |   pubs();      | beers();
                        |   brewery();   |
                        |   style();     |
          BeerDB::Handpump               |
          pub();                      BeerDB::Style
          beer();                     beers();

=head2 Ouch, that's a lot of inheritence!

Yes, that's a lot of inheritence, at some point in the future - probably Maypole 3.x we
will move to Class::C3

=head2 What about Maypole::Application - loading plugins

The main job of L<Maypole::Application> is to insert the plugins into the
hierarchy. It is also the responsibility of L<Maypole::Application> to decide
which frontend to use. It builds the list of plugins, then pushes them onto the
driver's C<@ISA>, then pushes the frontend onto the end of the driver's C<@ISA>.
So method lookup first searches all the plugins, before searching the frontend
and finally L<Maypole> itself.

From Maypole 2.11, L<Maypole::Application> makes no appearance in the
inheritance structure of a Maypole application. (In prior versions,
L<Maypole::Application> would make itself inherit the plugins, and then insert
itself in the hierarchy, but this was unnecessary).

=head2 Who builds the model?

First, remember we are talking about the standard, unmodified Maypole here. It
is possible, and common, to override some or all of this stage and build a
customised model. See below - An advanced Maypole application - for one
approach. Also, see L<Maypole's|Maypole> C<setup_model()> method. 

The standard model is built in 3 stages. 

First, C<Maypole::setup_model> calls C<setup_database> on the Maypole model
class, in this case L<Maypole::Model::CDBI>. C<setup_database> then uses
L<Class::DBI::Loader> to autogenerate individual L<Class::DBI> classes for each
of the tables in the database (C<BeerDB::Beer>, C<BeerDB::Pub> etc).
L<Class::DBI::Loader> identifies the appropriate L<Class::DBI> subclass and
inserts it into each of these table classes' C<@ISA> ( C<<
Class::DBI::<db_driver> >> in the diagrams)..

Next, C<Maypole::setup> B<pushes> L<Maypole::Model::CDBI> onto the C<@ISA> 
array of each of these classes. 

Finally, the relationships among these tables are set up. Either do this
manually, using the standard L<Class::DBI> syntax for configuring table
relationships, or try L<Class::DBI::Relationship> (which you can use via
L<Maypole::Plugin::Relationship>). If you use the plugin, you need to set up the
relationships configuration before calling C<setup()>. Be aware that some people
like the convenience of L<Class::DBI::Relationship>, others dislike the
abstraction. YMMV. 

=head1 An advanced Maypole application

We'll call it C<BeerDB2>.

Maypole is a framework, and you can replace different bits as you wish. So what 
follows is one example of good practice, other people may do things differently. 

We assume this application is being built from the ground up, but it will often
be straightforward to adapt an existing L<Class::DBI> application to this
general model.

The main idea is that the autogenerated Maypole model is used as a layer on top
of a separate L<Class::DBI> model. I am going to refer to this model as the
'Offline' model, and the Maypole classes as the 'Maypole' model. The idea is
that the Offline model can (potentially or in actuality) be used as part of
another application, perhaps a command line program or a cron script, whatever.
The Offline model does not know about the Maypole model, whereas the Maypole
model does know about the Offline model.

Let's call the offline model C<OfflineBeer>. As a traditional L<Class::DBI>
application, individual table classes in this model will inherit from a common
base (C<OfflineBeer>), which inherits from L<Class::DBI>).

One advantage of this approach is that you can still use Maypole's autogenerated
model. Another is that you do not mix online and offline code in the same

=head2 Building it

Build a driver in a similar way as for the basic app, calling C<setup()> after
setting up all the configuration. 

It is a good habit to use a custom Maypole model class for each application, as
it's a likely first target for customisation. Start it like this:

    package BeerDB2::Maypole::Model;
    use strict;
    use warnings;
    use base 'Maypole::Model::CDBI';
You can add methods which should be shared by all table classes to this package 
as and when required.
Configure it like this, before the C<setup()> call in the driver class:

    # in package BeerDB2

The C<setup()> call will ensure your custom model is loaded via C<require>.

B<Note>: by default, this will create Maypole/CDBI classes for all the tables in
the database. You can control this by passing options for L<Class::DBI::Loader>
in the call to C<setup()>.

For each class in the model, you need to create a separate file. So for
C<BeerDB2::Beer>, you would write:

    package BeerDB2::Beer;
    use strict;
    use warnings;
    use base 'OfflineBeer::Beer';
From Maypole 2.11, this package will be loaded automatically during C<setup()>,
and C<BeerDB2::Maypole::Model> is B<pushed> onto it's C<@ISA>.

Configure relationships either in the individual C<OfflineBeer::*> classes, or
else all together in C<OfflineBeer> itself i.e. not in the Maypole model. This 
way, you only define the relationships in one place.

The resulting model looks like this:

    MAYPOLE 'MODEL'                       |
   Maypole::Model::Base                   |
           +                              |
           |       +-----------------+----+-----------------+
           |       |                 |                      |
           |       |                 |                      |
     Maypole::Model::CDBI            |                      |     OFFLINE
             +                       |                      |        MODEL
             |                       |                      |
     BeerDB2::Maypole::Model  Class::DBI::<db_driver>  OfflineBeer
       +                             +                      +
       |                             |                      |
       +-----------------------------+                      |
       |                                                    |
       +--- BeerDB2::Pub --------+ OfflineBeer::Pub --------+
       |                           beers();                 |
       |                                                    |
       |                           OfflineBeer::Handpump ---+
       |                           beer();                  |
       |                           pub();                   |
       |                                                    |
       +--- BeerDB2::Beer -------+ OfflineBeer::Beer -------+
       |                           pubs();                  |
       |                           brewery();               |
       |                           style();                 |
       |                                                    |
       +--- BeerDB2::Style ------+ OfflineBeer::Style ------+
       |                           beers();                 |
       |                                                    |
       +--- BeerDB2::Brewery ----+ OfflineBeer::Brewery ----+

=head3 Features

1. Non-Maypole applications using the Offline model are completely isolated from
the Maypole application, and need not know it exists at all.

2. Methods defined in the Maypole table classes, override methods defined in the
Offline table classes, because C<BeerDB2::Maypole::Model> was pushed onto the
end of each Maypole table class's C<@ISA>. Perl's depth first,
left-to-right method lookup from e.g. C<BeerDB2::Beer> starts in
C<BeerDB2::Beer>, then C<BeerDB2::Maypole::Model>, C<Maypole::Model::CDBI>,
C<Maypole::Model::Base>, and C<Class::DBI>, before moving on to
C<OfflineBeer::Beer> and finally C<OfflineBeer>.

B<CAVEAT> - if your Offline model overrides L<Class::DBI> methods, these methods
will B<not> be overridden when called from the Maypole application, because the
Maypole model provides an alternative path to L<Class::DBI> which is searched
first. The solution is to place such methods in a separate package, e.g.
C<OfflineBeer::CDBI>. Place this B<first> in the C<@ISA> of both
C<BeerDB2::Maypole::Model> and C<OfflineBeer>. Note that C<OfflineBeer::CDBI>
does not itself need to inherit from L<Class::DBI>.

3. Methods defined in the Maypole model base class (C<BeerDB2::Maypole::Model>),
override methods in the individual Offline table classes, and in the Offline
model base class (C<Offline>). 

4. Relationships defined in the Offline classes are inherited by the Maypole

5. The Maypole model has full access to the underlying Offline model. 

=head3 Theory 

This layout illustrates more clearly why the Maypole model may be thought of as
part of the controller, rather than part of the model of MVC. Its function is to 
mediate web requests, translating them into method calls on the Offline model, 
munging the results, and returning them via the Maypole request object. 

Another way of thinking about it is that Maypole implements a two-layer
controller. The first layer translates a raw request into a single method call
on the Maypole model layer, which then translates that call into one or more
calls on the underlying model.

Whatever label you prefer to use, this approach provides for clear separation of
concerns between the underlying model and the web/user interface, and that's
what it's all about.

=head1 Advanced applications - building the model by hand ** TODO

- using Maypole::Model::CDBI::Plain or Maypole::FormBuilder::Model::Plain
- setup_model() and load_model_subclass()
- cutting out all those separate paths to CDBI - they're confusing 

=head1 Method inheritance ** TODO

More description of Perl's left-to-right, depth-first method lookup, and where
it's particularly important in Maypole.

=head1 AUTHOR

David Baird, C<< <cpan@riverside-cms.co.uk> >>


Copyright 2005 David Baird, All Rights Reserved.

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