my ($first, @rest) = eval $params;
my %params = ref $first eq 'HASH' ? %$first : ($first, @rest); # both (%opts) and {%opts}
for my $key (keys %params) {
- push @{$attrs->{$key}}, $params{$key};
+ my $parameters = ref $params{$key} eq 'ARRAY' ? @{$params{$key}} : $params{$key};
+ push @{$attrs->{$key}}, $parameters;
}
+ $attrs->{_role_attributes}->{$role} = \%params;
}
if (defined(my $alias = $self->_check_for_available_import($ctx, $role))) {
}
}
+You can consume multiple action roles similarly to the way you do with the
+class or role keyword:
+
+ action user
+ with LoggedIn
+ with isSuperUser {}
+
+Or
+
+ action User
+ with (LoggedIn, isSuperUser) {}
+
+Lastly, you can pass parameters to the underlying L<Catalyst::Action> using
+a syntax that is similar to method traits:
+
+ action myaction with hasRole(opt1=>'val1', opt2=>'val2')
+
+Where C<%opts> is a hash that is used to populate $action->attributes in the
+same way you might have done the following in classic L<Catalyst>
+
+ sub myaction :Action :Does(hasRole) :opt1(val1) :opt2(val2)
+
+Here's a more detailed example:
+
+ action User
+ with hasLogger(log_engine=>'STDOUT')
+ with hasPermissions(
+ role=>['Administrator', 'Member'],
+ ) {}
+
+Think of these are classic catalyst subroutine attributes on steriods. Unlike
+subroutine attributes, you can split and format your code across multiple lines
+and you can use deep and complex data structures such as HashRefs or ArrayRefs.
+Also, since the parameters are grouped syntactically within the C<with> keyword
+this should improve readability of your code, since it will be more clear which
+parameters belong to with roles. This should give L<CatalystX::Declare> greater
+compatibility with legacy L<Catalyst> code but offer us a way forward from
+needing subroutine attributes, which suffer from significant drawbacks.
+
+A few caveats and differences from method traits. First of all, unlike method
+traits, parameters are not passed to the L<Catalyst::Action> constructor, but
+instead used to populate the C<attributes> attribute, which is to preserve
+compatibility with how subroutine attributes work in classic L<Catalyst>.
+
+Additionally, since subroutines attributes supported a very limited syntax for
+supplying values, we follow the convention where parameter values are pushed
+onto an arrayref. In other words the following:
+
+ action User with hasLogger(engine=>'STDOUT')
+
+would create the following data structure:
+
+ $action->attributes->{engine} = ['STDOUT']
+
+The one exception is that if the value is an arrayref, those will be merged:
+
+ action User with Permissions(roles=>[qw/admin member/]) {}
+ ## Creates: $action->attributes->{roles} = ['admin','member']
+
+My feeling is that this gives better backward compatibility with classic sub
+attributes:
+
+ sub User :Action :Does(Permissions) :roles(admin) :roles(member)
+
+However, I realize this method could lead to namespace collisions. So in
+addition to putting paramters into $action->attributes, we also populate a
+special key "_role_attributes", which will preserve parameters by role:
+
+ $action->attributes->{_role_attributes}->{$role} = \%params;
+
+So the following:
+
+ action User with Permissions(roles=>[qw/admin member/]) {}
+
+Creates:
+
+ $action->attributes->{_role_attributes}->{Permissions}
+ = {roles=>[qw/admin member/]};
+
+For now you should only use this for your private code.
+
=head2 Action Classes
B<This option is even more experimental>