use Plack::Middleware::IIS6ScriptNameFix;
use Plack::Middleware::IIS7KeepAliveFix;
use Plack::Middleware::LighttpdScriptNameFix;
+use Plack::Util;
use Class::Load 'load_class';
BEGIN { require 5.008003; }
my $self = shift;
my %p = ( _log => $self->log );
$p{_uploadtmp} = $self->_uploadtmp if $self->_has_uploadtmp;
+ $p{data_handlers} = {$self->registered_data_handlers};
+ $p{_use_hash_multivalue} = $self->config->{use_hash_multivalue_in_request}
+ if $self->config->{use_hash_multivalue_in_request};
\%p;
}
__PACKAGE__->mk_classdata($_)
for qw/components arguments dispatcher engine log dispatcher_class
engine_loader context_class request_class response_class stats_class
- setup_finished _psgi_app loading_psgi_file run_options/;
+ setup_finished _psgi_app loading_psgi_file run_options _psgi_middleware
+ _data_handlers/;
__PACKAGE__->dispatcher_class('Catalyst::Dispatcher');
__PACKAGE__->request_class('Catalyst::Request');
# Remember to update this in Catalyst::Runtime as well!
-our $VERSION = '5.90042';
+our $VERSION = '5.90051';
sub import {
my ( $class, @arguments ) = @_;
$class->setup_log( delete $flags->{log} );
$class->setup_plugins( delete $flags->{plugins} );
+ $class->setup_middleware();
+ $class->setup_data_handlers();
$class->setup_dispatcher( delete $flags->{dispatcher} );
if (my $engine = delete $flags->{engine}) {
$class->log->warn("Specifying the engine in ->setup is no longer supported, see Catalyst::Upgrading");
EOF
}
+ # Call plugins setup, this is stupid and evil.
+ # Also screws C3 badly on 5.10, hack to avoid.
+ {
+ no warnings qw/redefine/;
+ local *setup = sub { };
+ $class->setup unless $Catalyst::__AM_RESTARTING;
+ }
+
+ # Initialize our data structure
+ $class->components( {} );
+
+ $class->setup_components;
+
if ( $class->debug ) {
my @plugins = map { "$_ " . ( $_->VERSION || '' ) } $class->registered_plugins;
$class->log->debug( "Loaded plugins:\n" . $t->draw . "\n" );
}
+ my @middleware = map {
+ ref $_ eq 'CODE' ?
+ "Inline Coderef" :
+ (ref($_) .' '. ($_->can('VERSION') ? $_->VERSION || '' : '')
+ || '') } $class->registered_middlewares;
+
+ if (@middleware) {
+ my $column_width = Catalyst::Utils::term_width() - 6;
+ my $t = Text::SimpleTable->new($column_width);
+ $t->row($_) for @middleware;
+ $class->log->debug( "Loaded PSGI Middleware:\n" . $t->draw . "\n" );
+ }
+
+ my %dh = $class->registered_data_handlers;
+ if (my @data_handlers = keys %dh) {
+ my $column_width = Catalyst::Utils::term_width() - 6;
+ my $t = Text::SimpleTable->new($column_width);
+ $t->row($_) for @data_handlers;
+ $class->log->debug( "Loaded Request Data Handlers:\n" . $t->draw . "\n" );
+ }
+
my $dispatcher = $class->dispatcher;
my $engine = $class->engine;
my $home = $class->config->{home};
? $class->log->debug(qq/Found home "$home"/)
: $class->log->debug(qq/Home "$home" doesn't exist/)
: $class->log->debug(q/Couldn't find home/);
- }
- # Call plugins setup, this is stupid and evil.
- # Also screws C3 badly on 5.10, hack to avoid.
- {
- no warnings qw/redefine/;
- local *setup = sub { };
- $class->setup unless $Catalyst::__AM_RESTARTING;
- }
-
- # Initialize our data structure
- $class->components( {} );
-
- $class->setup_components;
-
- if ( $class->debug ) {
my $column_width = Catalyst::Utils::term_width() - 8 - 9;
my $t = Text::SimpleTable->new( [ $column_width, 'Class' ], [ 8, 'Type' ] );
for my $comp ( sort keys %{ $class->components } ) {
# Support skipping finalize for psgix.io style 'jailbreak'. Used to support
# stuff like cometd and websockets
- if($c->request->has_io_fh) {
+ if($c->request->_has_io_fh) {
$c->log_response;
return;
}
return;
}
+## This exists just to supply a prebuild psgi app for mod_perl and for the
+## build in server support (back compat support for pre psgi port behavior).
+## This is so that we don't build a new psgi app for each request when using
+## the mod_perl handler or the built in servers (http and fcgi, etc).
+
sub _finalized_psgi_app {
my ($app) = @_;
return $app->_psgi_app;
}
+## Look for a psgi file like 'myapp_web.psgi' (if the app is MyApp::Web) in the
+## home directory and load that and return it (just assume it is doing the
+## right thing :) ). If that does not exist, call $app->psgi_app, wrap that
+## in default_middleware and return it ( this is for backward compatibility
+## with pre psgi port behavior ).
+
sub _setup_psgi_app {
my ($app) = @_;
sub psgi_app {
my ($app) = @_;
- return $app->engine->build_psgi_app($app);
+ my $psgi = $app->engine->build_psgi_app($app);
+ return $app->Catalyst::Utils::apply_registered_middleware($psgi);
}
=head2 $c->setup_home
$class => @roles
) if @roles;
}
+}
+
+=head2 registered_middlewares
+
+Read only accessor that returns an array of all the middleware in the order
+that they were added (which is the REVERSE of the order they will be applied).
+
+The values returned will be either instances of L<Plack::Middleware> or of a
+compatible interface, or a coderef, which is assumed to be inlined middleware
+
+=head2 setup_middleware (?@middleware)
+
+Read configuration information stored in configuration key C<psgi_middleware> or
+from passed @args.
+
+See under L</CONFIGURATION> information regarding C<psgi_middleware> and how
+to use it to enable L<Plack::Middleware>
+
+This method is automatically called during 'setup' of your application, so
+you really don't need to invoke it.
+
+When we read middleware definitions from configuration, we reverse the list
+which sounds odd but is likely how you expect it to work if you have prior
+experience with L<Plack::Builder> or if you previously used the plugin
+L<Catalyst::Plugin::EnableMiddleware> (which is now considered deprecated)
+
+=cut
+
+sub registered_middlewares {
+ my $class = shift;
+ if(my $middleware = $class->_psgi_middleware) {
+ return @$middleware;
+ } else {
+ die "You cannot call ->registered_middlewares until middleware has been setup";
+ }
+}
+
+sub setup_middleware {
+ my ($class, @middleware_definitions) = @_;
+ push @middleware_definitions, reverse(
+ @{$class->config->{'psgi_middleware'}||[]});
+
+ my @middleware = ();
+ while(my $next = shift(@middleware_definitions)) {
+ if(ref $next) {
+ if(Scalar::Util::blessed $next && $next->can('wrap')) {
+ push @middleware, $next;
+ } elsif(ref $next eq 'CODE') {
+ push @middleware, $next;
+ } elsif(ref $next eq 'HASH') {
+ my $namespace = shift @middleware_definitions;
+ my $mw = $class->Catalyst::Utils::build_middleware($namespace, %$next);
+ push @middleware, $mw;
+ } else {
+ die "I can't handle middleware definition ${\ref $next}";
+ }
+ } else {
+ my $mw = $class->Catalyst::Utils::build_middleware($next);
+ push @middleware, $mw;
+ }
+ }
+
+ $class->_psgi_middleware(\@middleware);
+}
+
+=head2 registered_data_handlers
+
+A read only copy of registered Data Handlers returned as a Hash, where each key
+is a content type and each value is a subref that attempts to decode that content
+type.
+
+=head2 setup_data_handlers (?@data_handler)
+
+Read configuration information stored in configuration key C<data_handlers> or
+from passed @args.
+
+See under L</CONFIGURATION> information regarding C<data_handlers>.
+
+This method is automatically called during 'setup' of your application, so
+you really don't need to invoke it.
+
+=head2 default_data_handlers
+
+Default Data Handlers that come bundled with L<Catalyst>. Currently there are
+only two default data handlers, for 'application/json' and an alternative to
+'application/x-www-form-urlencoded' which supposed nested form parameters via
+L<CGI::Struct> or via L<CGI::Struct::XS> IF you've installed it.
+
+The 'application/json' data handler is used to parse incoming JSON into a Perl
+data structure. It used either L<JSON::MaybeXS> or L<JSON>, depending on which
+is installed. This allows you to fail back to L<JSON:PP>, which is a Pure Perl
+JSON decoder, and has the smallest dependency impact.
+
+Because we don't wish to add more dependencies to L<Catalyst>, if you wish to
+use this new feature we recommend installing L<JSON> or L<JSON::MaybeXS> in
+order to get the best performance. You should add either to your dependency
+list (Makefile.PL, dist.ini, cpanfile, etc.)
+
+=cut
+
+sub registered_data_handlers {
+ my $class = shift;
+ if(my $data_handlers = $class->_data_handlers) {
+ return %$data_handlers;
+ } else {
+ die "You cannot call ->registered_data_handlers until data_handers has been setup";
+ }
+}
+
+sub setup_data_handlers {
+ my ($class, %data_handler_callbacks) = @_;
+ %data_handler_callbacks = (
+ %{$class->default_data_handlers},
+ %{$class->config->{'data_handlers'}||+{}},
+ %data_handler_callbacks);
+
+ $class->_data_handlers(\%data_handler_callbacks);
+}
+
+sub default_data_handlers {
+ my ($class) = @_;
+ return +{
+ 'application/x-www-form-urlencoded' => sub {
+ my ($fh, $req) = @_;
+ my $params = $req->_use_hash_multivalue ? $req->body_parameters->mixed : $req->body_parameters;
+ Class::Load::load_first_existing_class('CGI::Struct::XS', 'CGI::Struct')
+ ->can('build_cgi_struct')->($params);
+ },
+ 'application/json' => sub {
+ Class::Load::load_first_existing_class('JSON::MaybeXS', 'JSON')
+ ->can('decode_json')->(do { local $/; $_->getline });
+ },
+ };
}
=head2 $c->stack
In the future this might become the default behavior.
+=item *
+
+C<use_hash_multivalue_in_request>
+
+In L<Catalyst::Request> the methods C<query_parameters>, C<body_parametes>
+and C<parameters> return a hashref where values might be scalar or an arrayref
+depending on the incoming data. In many cases this can be undesirable as it
+leads one to writing defensive code like the following:
+
+ my ($val) = ref($c->req->parameters->{a}) ?
+ @{$c->req->parameters->{a}} :
+ $c->req->parameters->{a};
+
+Setting this configuration item to true will make L<Catalyst> populate the
+attributes underlying these methods with an instance of L<Hash::MultiValue>
+which is used by L<Plack::Request> and others to solve this very issue. You
+may prefer this behavior to the default, if so enable this option (be warned
+if you enable it in a legacy application we are not sure if it is completely
+backwardly compatible).
+
+=item *
+
+C<psgi_middleware> - See L<PSGI MIDDLEWARE>.
+
+=item *
+
+C<data_handlers> - See L<DATA HANDLERS>.
+
=back
=head1 INTERNAL ACTIONS
modules you are using must also be thread-safe. Some modules, most notably
L<DBD::SQLite>, are not thread-safe.
+=head1 DATA HANDLERS
+
+The L<Catalyst::Request> object uses L<HTTP::Body> to populate 'classic' HTML
+form parameters and URL search query fields. However it has become common
+for various alternative content types to be PUT or POSTed to your controllers
+and actions. People working on RESTful APIs, or using AJAX often use JSON,
+XML and other content types when communicating with an application server. In
+order to better support this use case, L<Catalyst> defines a global configuration
+option, C<data_handlers>, which lets you associate a content type with a coderef
+that parses that content type into something Perl can readily access.
+
+ package MyApp::Web;
+
+ use Catalyst;
+ use JSON::Maybe;
+
+ __PACKAGE__->config(
+ data_handlers => {
+ 'application/json' => sub { local $/; decode_json $_->getline },
+ },
+ ## Any other configuration.
+ );
+
+ __PACKAGE__->setup;
+
+By default L<Catalyst> comes with a generic JSON data handler similar to the
+example given above, which uses L<JSON::Maybe> to provide either L<JSON::PP>
+(a pure Perl, dependency free JSON parser) or L<Cpanel::JSON::XS> if you have
+it installed (if you want the faster XS parser, add it to you project Makefile.PL
+or dist.ini, cpanfile, etc.)
+
+The C<data_handlers> configuation is a hashref whose keys are HTTP Content-Types
+(matched against the incoming request type using a regexp such as to be case
+insensitive) and whose values are coderefs that receive a localized version of
+C<$_> which is a filehandle object pointing to received body.
+
+This feature is considered an early access release and we reserve the right
+to alter the interface in order to provide a performant and secure solution to
+alternative request body content. Your reports welcomed!
+
+=head1 PSGI MIDDLEWARE
+
+You can define middleware, defined as L<Plack::Middleware> or a compatible
+interface in configuration. Your middleware definitions are in the form of an
+arrayref under the configuration key C<psgi_middleware>. Here's an example
+with details to follow:
+
+ package MyApp::Web;
+
+ use Catalyst;
+ use Plack::Middleware::StackTrace;
+
+ my $stacktrace_middleware = Plack::Middleware::StackTrace->new;
+
+ __PACKAGE__->config(
+ 'psgi_middleware', [
+ 'Debug',
+ '+MyApp::Custom',
+ $stacktrace_middleware,
+ 'Session' => {store => 'File'},
+ sub {
+ my $app = shift;
+ return sub {
+ my $env = shift;
+ $env->{myapp.customkey} = 'helloworld';
+ $app->($env);
+ },
+ },
+ ],
+ );
+
+ __PACKAGE__->setup;
+
+So the general form is:
+
+ __PACKAGE__->config(psgi_middleware => \@middleware_definitions);
+
+Where C<@middleware> is one or more of the following, applied in the REVERSE of
+the order listed (to make it function similarly to L<Plack::Builder>:
+
+=over 4
+
+=item Middleware Object
+
+An already initialized object that conforms to the L<Plack::Middleware>
+specification:
+
+ my $stacktrace_middleware = Plack::Middleware::StackTrace->new;
+
+ __PACKAGE__->config(
+ 'psgi_middleware', [
+ $stacktrace_middleware,
+ ]);
+
+
+=item coderef
+
+A coderef that is an inlined middleware:
+
+ __PACKAGE__->config(
+ 'psgi_middleware', [
+ sub {
+ my $app = shift;
+ return sub {
+ my $env = shift;
+ if($env->{PATH_INFO} =~m/forced/) {
+ Plack::App::File
+ ->new(file=>TestApp->path_to(qw/share static forced.txt/))
+ ->call($env);
+ } else {
+ return $app->($env);
+ }
+ },
+ },
+ ]);
+
+
+
+=item a scalar
+
+We assume the scalar refers to a namespace after normalizing it using the
+following rules:
+
+(1) If the scalar is prefixed with a "+" (as in C<+MyApp::Foo>) then the full string
+is assumed to be 'as is', and we just install and use the middleware.
+
+(2) If the scalar begins with "Plack::Middleware" or your application namespace
+(the package name of your Catalyst application subclass), we also assume then
+that it is a full namespace, and use it.
+
+(3) Lastly, we then assume that the scalar is a partial namespace, and attempt to
+resolve it first by looking for it under your application namespace (for example
+if you application is "MyApp::Web" and the scalar is "MyMiddleware", we'd look
+under "MyApp::Web::Middleware::MyMiddleware") and if we don't find it there, we
+will then look under the regular L<Plack::Middleware> namespace (i.e. for the
+previous we'd try "Plack::Middleware::MyMiddleware"). We look under your application
+namespace first to let you 'override' common L<Plack::Middleware> locally, should
+you find that a good idea.
+
+Examples:
+
+ package MyApp::Web;
+
+ __PACKAGE__->config(
+ 'psgi_middleware', [
+ 'Debug', ## MyAppWeb::Middleware::Debug->wrap or Plack::Middleware::Debug->wrap
+ 'Plack::Middleware::Stacktrace', ## Plack::Middleware::Stacktrace->wrap
+ '+MyApp::Custom', ## MyApp::Custom->wrap
+ ],
+ );
+
+=item a scalar followed by a hashref
+
+Just like the previous, except the following C<HashRef> is used as arguments
+to initialize the middleware object.
+
+ __PACKAGE__->config(
+ 'psgi_middleware', [
+ 'Session' => {store => 'File'},
+ ]);
+
+=back
+
+Please see L<PSGI> for more on middleware.
+
=head1 ENCODING
On request, decodes all params from encoding into a sequence of
projects / catagits/Catalyst-Runtime.git / blobdiff