package Catalyst::Component; use Moose; use Class::MOP; use Class::MOP::Object; use Catalyst::Utils; use Class::C3::Adopt::NEXT; use Devel::InnerPackage (); use MRO::Compat; use mro 'c3'; use Scalar::Util 'blessed'; use Class::Load 'is_class_loaded'; use Moose::Util 'find_meta'; use namespace::clean -except => 'meta'; with 'MooseX::Emulate::Class::Accessor::Fast'; with 'Catalyst::ClassData'; =head1 NAME Catalyst::Component - Catalyst Component Base Class =head1 SYNOPSIS # lib/MyApp/Model/Something.pm package MyApp::Model::Something; use base 'Catalyst::Component'; __PACKAGE__->config( foo => 'bar' ); has foo => ( is => 'ro', ); sub test { my $self = shift; return $self->foo; } sub forward_to_me { my ( $self, $c ) = @_; $c->response->output( $self->foo ); } 1; # Methods can be a request step $c->forward(qw/MyApp::Model::Something forward_to_me/); # Or just methods print $c->comp('MyApp::Model::Something')->test; print $c->comp('MyApp::Model::Something')->foo; =head1 DESCRIPTION This is the universal base class for Catalyst components (Model/View/Controller). It provides you with a generic new() for component construction through Catalyst's component loader with config() support and a process() method placeholder. B that calling C<< $self->config >> inside a component is strongly not recommended - the correctly merged config should have already been passed to the constructor and stored in attributes - accessing the config accessor directly from an instance is likely to get the wrong values (as it only holds the class wide config, not things loaded from the config file!) =cut __PACKAGE__->mk_classdata('_plugins'); __PACKAGE__->mk_classdata('_config'); has catalyst_component_name => ( is => 'ro' ); # Cannot be required => 1 as context # class @ISA component - HATE # Make accessor callable as a class method, as we need to call setup_actions # on the application class, which we don't have an instance of, ewwwww # Also, naughty modules like Catalyst::View::JSON try to write to _everything_, # so spit a warning, ignore that (and try to do the right thing anyway) here.. around catalyst_component_name => sub { my ($orig, $self) = (shift, shift); Carp::cluck("Tried to write to the catalyst_component_name accessor - is your component broken or just mad? (Write ignored - using default value.)") if scalar @_; blessed($self) ? $self->$orig() || blessed($self) : $self; }; sub BUILDARGS { my $class = shift; my $args = {}; if (@_ == 1) { $args = $_[0] if ref($_[0]) eq 'HASH'; } elsif (@_ == 2) { # is it ($app, $args) or foo => 'bar' ? if (blessed($_[0])) { $args = $_[1] if ref($_[1]) eq 'HASH'; } elsif (is_class_loaded($_[0]) && $_[0]->isa('Catalyst') && ref($_[1]) eq 'HASH') { $args = $_[1]; } else { $args = +{ @_ }; } } elsif (@_ % 2 == 0) { $args = +{ @_ }; } return $class->merge_config_hashes( $class->config, $args ); } sub COMPONENT { my ( $class, $c ) = @_; # Temporary fix, some components does not pass context to constructor my $arguments = ( ref( $_[-1] ) eq 'HASH' ) ? $_[-1] : {}; if ( my $next = $class->next::can ) { my ($next_package) = Class::MOP::get_code_info($next); warn "There is a COMPONENT method resolving after Catalyst::Component in ${next_package}.\n"; warn "This behavior can no longer be supported, and so your application is probably broken.\n"; warn "Your linearized isa hierarchy is: " . join(', ', @{ mro::get_linear_isa($class) }) . "\n"; warn "Please see perldoc Catalyst::Upgrading for more information about this issue.\n"; } return $class->new($c, $arguments); } sub config { my $self = shift; # Uncomment once sane to do so #Carp::cluck("config method called on instance") if ref $self; my $config = $self->_config || {}; if (@_) { my $newconfig = { %{@_ > 1 ? {@_} : $_[0]} }; $self->_config( $self->merge_config_hashes( $config, $newconfig ) ); } else { # this is a bit of a kludge, required to make # __PACKAGE__->config->{foo} = 'bar'; # work in a subclass. # TODO maybe this should be a ClassData option? my $class = blessed($self) || $self; my $meta = find_meta($class); unless (${ $meta->get_or_add_package_symbol('$_config') }) { # Call merge_hashes to ensure we deep copy the parent # config onto the subclass $self->_config( Catalyst::Utils::merge_hashes($config, {}) ); } } return $self->_config; } sub merge_config_hashes { my ( $self, $lefthash, $righthash ) = @_; return Catalyst::Utils::merge_hashes( $lefthash, $righthash ); } sub process { Catalyst::Exception->throw( message => ( ref $_[0] || $_[0] ) . " did not override Catalyst::Component::process" ); } sub expand_modules { my ($class, $component) = @_; return Devel::InnerPackage::list_packages( $component ); } __PACKAGE__->meta->make_immutable; 1; __END__ =head1 METHODS =head2 new($app, $arguments) Called by COMPONENT to instantiate the component; should return an object to be stored in the application's component hash. =head2 COMPONENT C<< my $component_instance = $component->COMPONENT($app, $arguments); >> If this method is present (as it is on all Catalyst::Component subclasses), it is called by Catalyst during setup_components with the application class as $app and any config entry on the application for this component (for example, in the case of MyApp::Controller::Foo this would be C<< MyApp->config('Controller::Foo' => \%conf >>). The arguments are expected to be a hashref and are merged with the C<< __PACKAGE__->config >> hashref before calling C<< ->new >> to instantiate the component. You can override it in your components to do custom construction, using something like this: sub COMPONENT { my ($class, $app, $args) = @_; $args = $class->merge_config_hashes($class->config, $args); return $class->new($app, $args); } B Generally when L starts, it initializes all the components and passes the hashref present in any configuration information to the COMPONENT method. For example MyApp->config( 'Model::Foo' => { bar => 'baz', }); You would expect COMPONENT to be called like this ->COMPONENT( 'MyApp', +{ bar=>'baz'}); This would happen ONCE during setup. =head2 $c->config =head2 $c->config($hashref) =head2 $c->config($key, $value, ...) Accessor for this component's config hash. Config values can be set as key value pair, or you can specify a hashref. In either case the keys will be merged with any existing config settings. Each component in a Catalyst application has its own config hash. The component's config hash is merged with any config entry on the application for this component and passed to C (as mentioned above at L). The recommended practice to access the merged config is to use a Moose attribute for each config entry on the receiving component. =head2 $c->process() This is the default method called on a Catalyst component in the dispatcher. For instance, Views implement this action to render the response body when you forward to them. The default is an abstract method. =head2 $c->merge_config_hashes( $hashref, $hashref ) Merges two hashes together recursively, giving right-hand precedence. Alias for the method in L. =head2 $c->expand_modules( $setup_component_config ) Return a list of extra components that this component has created. By default, it just looks for a list of inner packages of this component =cut =head1 OPTIONAL METHODS =head2 ACCEPT_CONTEXT($c, @args) Catalyst components are normally initialized during server startup, either as a Class or a Instance. However, some components require information about the current request. To do so, they can implement an ACCEPT_CONTEXT method. If this method is present, it is called during $c->comp/controller/model/view with the current $c and any additional args (e.g. $c->model('Foo', qw/bar baz/) would cause your MyApp::Model::Foo instance's ACCEPT_CONTEXT to be called with ($c, 'bar', 'baz')) and the return value of this method is returned to the calling code in the application rather than the component itself. B All classes that are Ls will have a COMPONENT method, but classes that are intended to be factories or generators will have ACCEPT_CONTEXT. If you have initialization arguments (such as from configuration) that you wish to expose to the ACCEPT_CONTEXT you should proxy them in the factory instance. For example: MyApp::Model::FooFactory; use Moose; extends 'Catalyst::Model'; has type => (is=>'ro', required=>1); sub ACCEPT_CONTEXT { my ($self, $c, @args) = @_; return bless { args=>\@args }, $self->type; } MyApp::Model::Foo->meta->make_immutable; MyApp::Model::Foo->config( type => 'Type1' ); And in a controller: my $type = $c->model('FooFactory', 1,2,3,4): # $type->isa('Type1') B If you define a ACCEPT_CONTEXT method it MUST check to see if the second argument is blessed (is a context) or not (is an application class name) and it MUST return something valid for the case when the scope is application. This is required because a component maybe be called from the application scope even if it requires a context and you must prevent errors from being issued if this happens. Remember not all components that ACCEPT_CONTEXT actually need or use context information (and there is a school of thought that suggestions doing so is a design error anyway...) =head1 SEE ALSO L, L, L, L. =head1 AUTHORS Catalyst Contributors, see Catalyst.pm =head1 COPYRIGHT This library is free software. You can redistribute it and/or modify it under the same terms as Perl itself. =cut