On Tuesday 10 November 2009 00:31:57 Dave Rolsky wrote: > On Tue, 10 Nov 2009, Nathan Gass wrote: > > They are mainly useful for overwriting. My current code for > > MooseX::Attribute::Delegated at > > http://github.com/xabbu42/MooseX-Attribute-Delegated > > has to use a custom Method::Accessor and a custom Method::Constructor > > with two mostly copy-pasted methods (see > > http://github.com/xabbu42/MooseX-Attribute- > > Delegated/blob/master/lib/MooseX/Attribute/Delegated.pm) > > Not to get too off-topic, but what is the point of what you're doing? > > If you want to extract a bunch of attributes from a hashref passed to the > constructor as one of its arguments, there are lots of simpler ways to do > this. > > Notably, a parameterized role that added a method modifier on BUILD would > probably be a _much_ simpler solution. You could also just use a lazy > builder.
I mainly want an attribute which works with normal object introspection (- >meta->get_all_attributes and company). Later it could eventual be interesting to add other Extensions which change the behavior of attributes like for example MX::TrackDirty::Attributes. Both of this only work if I have a real attribute meta-object. What you describe is actually currently not implemented in MX::Delegated::Attribute, since this point is not very important to me. > > > as I see it, Method::Constructor defines the behavior of a constructor > > and should not presuppose the way a specific attribute gets stored in the > > object. How an attribute gets stored should instead be decided by the > > attribute meta- object. Same story for Method::Accessor. I'll try to find > > an example on CPAN for this point. > > The Constructor doesn't assume anything about how attributes are stored, > it asks the meta-instance to handle this detail. This is all really > confusing without adding yet another layer of indirection. The Constructor assumes that every attribute gets stored in a single slot directly in the object. > > > My changes streamline some inconsistencies in the current code too. The > > current code uses sometimes $attr->name and sometimes $attr->slots to > > decide on the slot name. Overwriting the slots method of Meta::Attribute > > therefore currently gives you a buggy attribute, since some code > > incorrectly still uses the name of the attribute as slot name. This bug > > can of course be fixed without my changes. The point is that bugs like > > this are easier avoided if all code concerning inline attribute storage > > is concentrated in one place. > > I'd like to see that bug fixed too. > > However, attribute storage is already in one place, the meta instance. Yeah, but the meta instance is the same for all attributes. So there is no way to change how a single attribute gets stored. - xabbu42 > > > -dave > > /*============================================================ > http://VegGuide.org http://blog.urth.org > Your guide to all that's veg House Absolute(ly Pointless) > ============================================================*/ >
