Re: "automatically" wrapping methods.

Or you can use B::Hooks::EndOfScope to do the wrapping at the end of the compilation stage:

# untested code
package MooseX::MethodValidation;
use B::Hooks::EndOfScope;
sub import {
  on_scope_end {
    my $caller = caller;
    my $meta = Moose::Meta::Class->initialize($caller);
    foreach my $method ( $meta->get_all_methods )  {
      ... # wrapping code goes here
    }
  }
}

package MyModel::User;
use MooseX::MethodValidation;
... # your subs go here

Cheers,

--
Peter



>________________________________
> From: Chris Weyl <[email protected]>
>To: [email protected] 
>Cc: Bill Moseley <[email protected]>; moose <[email protected]> 
>Sent: Monday, 9 January 2012, 20:26
>Subject: Re: "automatically" wrapping methods.
> 
>Parameterized roles are nifty (and I've probably used them more than is
>good for my health), but in this scenario a more elegant approach (IMHO)
>would be to create a trait to be applied to the method metaclass that will
>do the validation you need, and syntactic sugar (see Moose::Exporter) to
>handle the work of it.  Ideally you'd end up with something like:
>
>    use Moose;
>    use MooseX::MethodValidation;
>
>    validated_method foo => sub { ... };
>
>validated_method() being the sugar that will take the class metaclass,
>method name, and method body; then create/apply the method using the
>trait'ed method metaclass.  You may need to do a little experimenting to
>see what works best here vis-a-vis how much to do in the sugar and how much
>to do in the trait.
>
>This approach has the advantage that you can easily create validated
>methods in the same manner as method modifiers (around, before, etc), and
>you can also easily determine which methods are validated by checking their
>metaclasses to see if they do the validation trait.  It also tends to make
>it pretty obvious about what's going on (not how, but what) with a minimum
>of fuss.
>
>                                 -Chris
>
>On Sun, Jan 8, 2012 at 11:57 PM, <[email protected]> wrote:
>
>> I'd look to moving this logic into a parameterised role, with the
>> parameters being the methods you are looking at validating. Have a look at
>> MooseX::Role::Parameterized for details on how to do this.
>>
>> Take care,
>> Mark.
>>
>> On 9 Jan 2012, at 05:14, Bill Moseley <[email protected]> wrote:
>>
>> > I have have a number of classes that include methods that need input
>> > validation.  I use a validation class for each method which means
>> > validation is called the same way every time its used.  That is, I have a
>> > common pattern I'm using on a number of methods and wondering if anyone
>> can
>> > suggest a more clever approach so I don't end up repeating similar code.
>> >
>> > A code example might make it clearer:
>> >
>> > package MyModel::User;
>> > use Moose;
>> > extends 'MyModel::Base';
>> >
>> >
>> > # This method is called by the consumer to create a new user,
>> > # and the $user_data passed here is assumed to have been validated.
>> >
>> > sub new_user {
>> >    my ( $self, $user_data ) = @_;
>> >
>> >    return $self->ORM_layer( 'User' )->create( $user_data );
>> > }
>> >
>> >
>> > # Here's the code that gets repeated for every method that requires
>> > validation.
>> >
>> > around new_user => sub {
>> >    my ( $orig, $self, $args ) = @_;
>> >
>> >    # Fetch validation instance for
>> >    my $validation_class = $self->validation_class_for_method( 'new_user'
>> );
>> >    my $validation_instance = $validation_class->new( $args );
>> >
>> >    # Throw exception object with validation failures unless data passes
>> > validation.
>> >    $validation_instance->thow_exception unless
>> > $validation_instance->validates_ok;
>> >
>> >    return;
>> > };
>> >
>> >
>> > So, I'm looking for a suggestion how to avoid the explicit wrapper.  One
>> > option I was considering was something Catalyst-like with subroutine
>> > attributes:
>> >
>> > sub new_user : Validate {
>> >    ...
>> > }
>> >
>> > Then all methods with the "Validate" attribute would automatically get
>> > wrapped, although I'm not clear if that's the best approach or even its
>> > implementation.
>> >
>> > Thanks,
>> >
>> >
>> >
>> >
>> >
>> > --
>> > Bill Moseley
>> > [email protected]
>>
>
>
>
>-- 
>Chris Weyl
>Ex astris scientia
>
>
>