=head1 NAME

Reaction::Manual::Tutorial - Step by Step Tutorial

=head1 DESCRIPTION

This document aims at giving simple step-by-step leading to an example application
using the common functionality provided by L<Reaction>.

=head1 CREATING A NEW APPLICATION

At first we have to create a new application. For this we use the C<catalyst.pl>
script as we would for any other Catalyst application:

  $ catalyst.pl MyApp
  [lots "created ..." messages]

There is nothing to change in the application class file.

As you work through this tutorial you'll be creating several new files in
various directories. You can save some time by creating the directories now,
like this:

  mkdir -p share/skin/myapp/layout lib/MyApp/View/Site/Widget lib/MyApp/Schema lib/MyApp/InterfaceModel

=head1 THE VIEW

Since we are not just rendering templates with Reaction, but layouts and widgets,
a simple TT view won't suffice. We need to create our own C<lib/MyApp/View/Site.pm>:

  package MyApp::View::Site;
  use Reaction::Class;
  
  use namespace::autoclean;
  
  extends 'Reaction::UI::View::TT';
  
  __PACKAGE__->meta->make_immutable;
  
  1;

The C<use Reaction::Class> line will import L<Moose>, L<strict> and L<warnings> into
our file and might perform some Reaction specific setups.

We make sure that we don't provide imported functions as methods at runtime by using
L<namespace::autoclean>.

In its simplest version, our view just needs to do a C<extends 'Reaction::UI::View::TT'>
to make a new subclass of it.

We chose to call C<make_immutable> on the class' meta class instance to have it inline
methods for runtime speed improvements.

=head1 THE ROOT CONTROLLER

As usual in Catalyst, our root controller (at C<lib/MyApp/Controller/Root.pm> represents
the root namespace for our application. For this purpose, it should look like this:

  package MyApp::Controller::Root;
  use strict;
  use warnings;
  use parent 'Reaction::UI::Controller::Root';
  
  use aliased 'Reaction::UI::ViewPort';
  use aliased 'Reaction::UI::ViewPort::SiteLayout';
  
  use namespace::autoclean;
  
  __PACKAGE__->config(
      view_name       => 'Site',
      window_title    => 'MyApp Window',
      namespace       => '',
  );
  
  sub base: Chained('/') PathPart('') CaptureArgs(0) {
      my ($self, $ctx) = @_;
      $self->push_viewport(SiteLayout,
          title           => 'MyApp Test Title',
          static_base_uri => join('', $ctx->uri_for('/static')),
          meta_info       => {
              http_header   => {
                  'Content-Type' => 'text/html;charset=utf-8',
              },
          },
      );
  }
  
  sub root: Chained('base') PathPart('') Args(0) {
      my ($self, $ctx) = @_;
      $self->push_viewport(ViewPort, layout => 'root');
  }
  
  1;

The effects of L<strict>, L<warnings>, L<parent>, L<aliased> and L<namespace::autoclean>
should be clear by now. Let's take a look at the configuration.

The C<view_name> determines which view to use. We set it to C<Site>, which is our only view
by now. Be careful to set C<view_name> and not C<view>, which would fail telling you it 
expected an object.

The C<window_title> is the title given to the L<Reaction::UI::Window> instance that will be
stored in C<$ctx-E<gt>stash-E<gt>{window}> by the C<begin> action provided by
L<Reaction::UI::Controller::Root>.

The C<namespace> setting anchors the root controller at C</>.

The C<base> action here acts as a general point all other actions can chain off of. It
pushes the L<Reaction::UI::ViewPort::SiteLayout> viewport onto the 
L<focus stack|Reaction::UI::FocusStack>. As arguments we see a C<title> that will be used
as page title later. The C<static_base_uri> is used for static links like CSS and JavaScript
files. Since we didn't specify a layout C<site_layout> will be used.

We also defined a C<root> action serving as application index. It chains off the C<base>
action. It is only pushing the root viewport L<Reaction::UI::ViewPort> on the focus stack,
but this time we specified a layout named C<root>.

Reaction will try to find our layout files in C<share/skin/$skin_name/layout/*>, so the next
thing to do is to create a new skin and the layout files.

=head1 A NEW SKIN

If your version of Catalyst still creates a C<root> instead of a C<share> directory, you
might want to rename it. This is regarded as a best practice and follows the conventions
of this tutorial and other Reaction documentation.

First we need to create a directory for our new skin:

  $ mkdir -p share/skin/myapp/layout

Next we need to configure our new skin. This is done in the C<share/skin/myapp/skin.conf>
file. At the moment, all it should contain is

  extends /Reaction/default

Note that this C<extends> specification contains the distribution name of the
library or application of which to use the templates as base. You can also give it
a relative name like

  extends foo

and it would try to extend a skin named C<foo> in your own application's C<share/skin>
directory.

In general, the Reaction distribution provides two skin choices to build upon:
C</Reaction/default> and C</Reaction/base>.  C<base> is Reaction's most basic skin
while the C<default> skin extends the base by embellishing the layouts with more XHTML.  
The default skin is designed to be a working skin.

In other words, if you want to write most of the low-level XHTML yourself
then put the following in your skin.conf:

  extends /Reaction/base
  
If you want to use more XHTML out of the box, then configure your application to extend 
the default skin.  

It's worth noting that you are not bound to the XHTML of the default or base skin layouts,
because you are able to override layout fragments in your own skin.

Next we create C<share/skin/defaults.conf> to allow settings that concern all skins of
the application. It should contain only this:

  widget_search_path MyApp::View::Site::Widget
  widget_search_path Reaction::UI::Widget

This will tell Reaction to look in C<Reaction::UI::Widget::*> and 
C<MyApp::View::Site::Widget::*> for widget classes. That means that our layout named
C<root> will check for C<MyApp::View::Site::Widget::Root> first and then look if
C<Reaction::UI::Widget> exists.

We want the first line to be able to create our own widgets and the second line to
have Reaction find its own widgets.

Now we need to tell Reaction what skin it should use. We do this by adding this section
to our C<myapp.conf>:

  <View Site>
      skin_name myapp
  </View>

The value should be the name of the target directory under C<share/skin/>.

=head1 LAYOUTS

We will need two layout files to begin with. One controlling the site layout and one
for the root action.

The first will be created as C<share/skin/myapp/layout/site_layout.tt>:

  =extends NEXT

  =for layout body

      <h1>Welcome to MyApp</h1>

      <div id="content">
          [% inner %]
      </div>

  =cut

The C<=extends> directive specifies that this layout file is an extension of another
layout file. The C<NEXT> value here tells Reaction that this extends the C<site_layout>
layout in the base skin, which we have defined as C</Reaction/default>. That means, you
can take a look at the layout we are extending at C<share/skin/default/layout/site_layout.tt>
in the L<Reaction> distribution.

The C<=for layout> directives allows us to set a layout fragment. We define a C<body> fragment
containing the common C<body> for all pages using this site layout. The C<[% inner %]> is
where the deeper parts of the stack will be included, in the case of our C<root> action that
would be the C<Reaction::UI::ViewPort> with the C<root> layout.

If we want to override a specific fragment, we can do just that. There are two choices
when overriding a specific fragment.  One can disregard the parent fragment all together
and start from scratch or they could reuse the parent fragment and add to it.  The key to 
re-using a parent (extended) layout fragment is the keyword C<call_next>.  Specifically, 
we insert C<[% call_next %]> to include the layout fragment of the parent.  

The parent layout fragment is found either in another layout template named in the 
C<=extends> directive or it's in a subclass of Reaction::UI::Widget.
In the latter case, the Widget subclass name matches the name of the template or
the widget name posed at the top of the layout file via the C<=widget> directive.

Once you know the hierarchy C<call_next> follows, then you know where to examine
the details of the parent layouts or widgets.

The layout representing the root action is called C<share/skin/myapp/layout/root.tt>:

  =for layout widget

      <p>Hello, World!</p>

  =cut

This one is rather simple. The C<=for layout widget> directive is special in that the
C<widget> fragment will always be where the rendering starts. In fact, our C<site_layout>
layout too contains a C<widget> fragment, you just don't see it because you inherited it from
your base skin (or your base skin's base skin, for that matter) instead of defining it yourself.

=head1 A SIMPLE WIDGET

If we wanted to use a different kind of widget than that assumed automatically by Reaction, we 
could add a

  =widget ClassName

directive at the top of the layout file. But for now, we will instead create our own
widget at C<lib/MyApp/View/Site/Widget/Root.pm>:

  package MyApp::View::Site::Widget::Root;
  use Reaction::UI::WidgetClass;
  
  use namespace::autoclean;
  
  __PACKAGE__->meta->make_immutable;
  
  1;

This adds no new functionality at the moment. It just uses C<Reaction::UI::WidgetClass> to ease
and automate the setup of a new widget class. The widget can provide functionality and fragments
to the layout. In a way, it can be seen as the Perl code backend to the layout file.

You can now start your C<script/myapp_server.pl> and visit

  http://localhost:3000/

to view your "Hello, World" page.

=head1 ADDING A SCHEMA

The next part of the tutorial will be about adding data storage to our application. While most
L<Catalyst> web applications today (or at least they should) abstract their database schema
with L<DBIx::Class::Schema> into a separate module separated from the webapplication, Reaction
takes this one step further by introducing so called interface models. The interface model
defines the layer between your application and your domain model (in this case, the L<DBIx::Class>
schema).

The first thing we will need is a schema class in C<lib/MyApp/Schema.pm>:

  package MyApp::Schema;
  use strict;
  use warnings;
  
  use parent 'DBIx::Class::Schema';
  
  __PACKAGE__->load_classes;
  
  1;

The schema class itself is built like a typical L<DBIx::Class::Schema>. The difference in class
definition starts at the result classes. For the example's sake, let's make a SQLite database
called C<example.sqlite>:

  $ cat > example.sqlite.sql
  CREATE TABLE foo (
    id          INTEGER PRIMARY KEY AUTOINCREMENT,
    first_name  VARCHAR NOT NULL,
    last_name   VARCHAR NOT NULL
  );
  <Ctrl-D>

  $ sqlite3 example.sqlite < example.sqlite.sql
  $

The result class for this table combines the usual style of L<DBIx::Class> with L<Moose> meta
data additions in C<lib/MyApp/Schema/Foo.pm>:

  package MyApp::Schema::Foo;
  use Moose;
  use MooseX::Types::Moose  qw( Int );
  use Reaction::Types::Core qw( NonEmptySimpleStr );
  
  use namespace::autoclean;
  
  extends 'DBIx::Class';
  
  has id => 
      (is => 'ro', isa => Int, required => 1);
  
  has first_name => 
      (is => 'rw', isa => NonEmptySimpleStr, required => 1);
  
  has last_name => 
      (is => 'rw', isa => NonEmptySimpleStr, required => 1);
  
  __PACKAGE__->load_components(qw( IntrospectableM2M Core ));
  __PACKAGE__->table('foo');
  
  __PACKAGE__->add_columns(
      id => { 
          data_type         => 'integer', 
          is_auto_increment => 1,
      },
      first_name => { data_type => 'varchar' },
      last_name  => { data_type => 'varchar' },
  );
  
  __PACKAGE__->set_primary_key('id');
  
  1;

The L<MooseX::Types::Moose> and L<Reaction::Types::Core> modules export L<Moose> type
constraints (See also L<MooseX::Types> and L<Moose::Util::TypeConstraints>). Note that
we are using L<Moose/extends> here instead of L<base> or L<parent> to extend 
L<DBIx::Class>.

Next we see our columns in form of Moose attribute definitions. The C<is>, C<isa> and
C<required> attribute parameters will all be used for introspection and interface
building later. The C<required> is rather straight-forward. The C<is> will decide whether
this attribute (or column) can be edited (C<ro> means that it can't, C<rw> means it can).
The C<isa> attribute will be used for validation and rendering of input fields.

The imported C<NonEmptySimpleStr> for example gives us a simple single-line input box,
while a C<Str> from L<MooseX::Types::Moose> would give us a textbox.

Following that, we have the usual L<DBIx::Class> result class definitions. The only thing
different might be the new L<DBIx::Class::IntrospectableM2M> which will allow us to
inspect many-to-many relations later on.

=head1 CREATING AN INTERFACE MODEL

The interface model should be separated from the application and the schema, since it
will tie both together. In this case, we will use a reflector to set up the usual interface
model actions for our schema (C<Create>, C<Update>, C<Delete>, C<DeleteAll>) in
C<lib/MyApp/InterfaceModel/DBIC.pm>:

  package MyApp::InterfaceModel::DBIC;
  
  # keep this on top
  use parent 'Reaction::InterfaceModel::Object';
  
  use Reaction::Class;
  use Reaction::InterfaceModel::Reflector::DBIC;
  
  use namespace::autoclean;
  
  my $reflector = Reaction::InterfaceModel::Reflector::DBIC->new;
  
  $reflector->reflect_schema(
      model_class     => __PACKAGE__,
      schema_class    => 'MyApp::Schema',
  );
  
  __PACKAGE__->meta->make_immutable;
  
  1;

The L<parent> import must happen before the L<Reaction::Class> one in this case. Other
than that, the only thing we do here is create a new L<Reaction::InterfaceModel::Reflector::DBIC>
and call C<reflect_schema> to build our C<MyApp::InterfaceModel::DBIC::*> namespace out of our
C<MyApp::Schema>.

=head1 TIEING THE INTERFACE MODEL TO THE APPLICATION

Next on the list is the integration of our new interface model into our application. For
this we create a simple catalyst model in C<lib/MyApp/Model/DBIC.pm>:

  package MyApp::Model::DBIC;
  use Reaction::Class;
  
  use namespace::autoclean;
  
  extends 'Catalyst::Model::Reaction::InterfaceModel::DBIC';
  
  __PACKAGE__->meta->make_immutable;
  
  __PACKAGE__->config(
      im_class    => 'MyApp::InterfaceModel::DBIC',
      db_dsn      => 'dbi:SQLite:example.sqlite',
  );
  
  1;

This model L<extends|Moose> the L<Catalyst::Model::Reaction::InterfaceModel::DBIC> base class
shipped with Reaction. It's configuration must contain the C<im_class> naming our interface
model and the C<db_dsn>. If you're using a different kind of database then you
may neeb to add config for C<db_user>, C<db_password>, and C<db_params>.
All these get passed to the schema's connect() method.

Of course, since this is Catalyst, all this can also easily be specified 
via your application config file under the C<Model::DBIC> key.

=head1 BUILDING A SIMPLE CRUD CONTROLLER

Now, since we have defined our interface model as well as our domain model including meta
data, it isn't very hard (at least not for us) to build a basic (but extendable)
CRUD controller in C<lib/MyApp/Controller/Foo.pm>:

  package MyApp::Controller::Foo;
  use strict;
  use warnings;
  
  use parent 'Reaction::UI::Controller::Collection::CRUD';
  use Reaction::Class;
  
  use namespace::autoclean;
  
  __PACKAGE__->config(
      model_name      => 'DBIC',
      collection_name => 'Foo',
      actions => {
          base => { Chained => '/base', PathPart => 'foo' },
      },
  );
  
  1;

This controller subclasses L<Reaction::UI::Controller::Collection::CRUD>, which is itself a
subclass of L<Reaction::UI::Controller::Collection>, a class to ease the creation of controllers
who act on collections of things.

As you can see, for the simplest case we don't need any code; we simply configure our controller.

The C<model_name> is the name of our interface model sans the C<MyApp::Model::> prefix. This means
this entry points to C<MyApp::Model::DBIC> in this case. The C<collection_name> is the name of
the collection in the specified interface model. For us, this would be C<Foo>, to match the result
class we created above and want to manage.

The C<actions> part of the configuration is not Reaction, but rather Catalyst specific. This
configures the actions inherited from L<Reaction::UI::Controller::Collection::CRUD>. For it to
work, we only need to tell the C<base> action where to chain off from, and what C<PathPart>
to use. We chain it to the C<base> action in our root controller. The C<foo> path part is rather
obvious.

Now you can restart your application and visit

  http://localhost:3000/foo

and you have a complete CRUD interface for C<Foo> with listing, viewing, creating, updating
and deleting capabilities.

=head1 WHERE TO GO NEXT

=over

=item L<Reaction::Manual::Templates>

When a viewport tries to render a layout, it will involve the view to figure out the corresponding
template (or any other kind of GUI description) and render it. The template documentation is concerned
mostly with the C<Reaction::UI::View::TT> implementation allowing the developer to use the L<Template>
engine to render the layouts.

=item L<Reaction::Manual::Widgets>

A widget is the backend Perl object providing the Perl view logic to a layout. What the rendered
output actually looks like is determined by the layout. The widget is concerned with storing, providing
and managing data used and rendered by the layout.

=back

=head1 SEE ALSO

=over 

=item * L<Reaction::Manual>

=item * L<Reaction::Manual::Intro>

=item * L<Reaction::Manual::Overview>

=item * L<Reaction::Manual::Templates>

=item * L<Reaction::Manual::RenderPage>

=item * L<Reaction::UI::View::TT>

=item * L<Reaction::UI::Controller::Root>

=item * L<Reaction::UI::WidgetClass>

=back

=head1 AUTHORS

See L<Reaction::Class> for authors.

=head1 LICENSE

See L<Reaction::Class> for the license.

=cut