If I understand your question correctly, then I am seeing a big
problem with your design.
If you have an instance of a DBIx::Class::Row subclass stored in your
attribute, which can potentially be from different resultsets, then
you should not alter your *class* based on a value which is specific
to the *instance*. The value of that attribute, and therefore the
methods you can to delegate too, can change from instance to instance.
You do not want to be altering the set of delegated methods at the
class level.
What you want to do is to create different classes for each possible
DBIx::Class::Row subclass, they can share behavior through a common
base class or roles, but you should have a specific class for each
possible type of row object. This could even be done with anon-classes
if you want.
- Stevan
On Dec 28, 2009, at 11:43 AM, Rolf Schaufelberger wrote:
Hi,
I have some classes, that have a DBIx::Class::Row object a one of
it's attributes.
What is the best way to delegate all methods (acessing columns an
relations ) to this object ?
In my current, non Moose implementation I'm doing this with
AUTOLOAD, there I'm testing if the object can the required method
and call it.
The docs mention that either a role or a subref can be specified as
the value for the delegates param.
Now, since my row objects are from different resultsets, the
methods aren't always the same, so I cannot see, how a role could
help but, but in a sub I could introspect the row object. How will
that subref be called ? Or is there better solution ?
Rolf Schaufelberger
