Re: A trigger called when following a 'has_a' (again)

[prev] [thread] [next] [Date index for 2004/07/22]

--On Thursday, July 22, 2004 2:41 PM +0100 Peter Pimley 
<ppimley@xxxxxxxxx.xxx> wrote:

> So, back to my original problem:  There is no trigger that gets called on
> an object before it is returned from a 'has_a' method.  The docs for

I think the issue isn't so much that no trigger gets called -- for 
instance, has_a()s are implemented as triggers at the moment -- but that 
there's no guaranteed order of evaluation for triggers.  I suspect in your 
case that the bless_to_subclass() trigger end up being called before the 
inflator() trigger that's part of the has_a, which happily reblesses the 
object back into the superclass.  Since it's in a grey area, I guess it's 
not really a bug, though one can argue that the current behavior (if random 
trigger order hands you an object already blessed into a subclass of the 
has_a's "target" class, rebless it into the target class) is less intuitive 
than just letting it go by.  That said, changing C:D:R:HasA to pass the 
already-blessed object  shouldn't break anything that (reliably) works now, 
and would let an arrangement like your work reliably.  That's attractive to 
me, since only the superclass needs to know about the magic relationship 
with its subclasses.

In the meantime, custom inflators aren't subject to this problem, so you 
can avoid it by saying

  __PACKAGE__->has_a(my_thing => 'superclass',
                     inflate => \&bless_to_subclass);



(In order to explain the last call in your example, I'm making the further 
inference here that $owner owns my_thing #1 or #2, so that [courtesy of the 
Live_Objects cache] when one of the later retrieve()s calls just the 
bless_to_subclass() trigger, the single common instance of my_thing is 
reblessed into the right subclass.  If you retrieve() another owner of the 
same my_thing, does it revert to the superclass?  Alternatively, if you 
place the retrieve() call for $owner after the retrieve() calls for $sub or 
only retrieve $subs not owned by $owner, does $owner->my_thing->do_stuff 
call the superclass method?)

Please pardon the rampant speculation here.  If I'm off base, I'll have 
some time this evening or tomorrow to look at this in the debugger and try 
to sort it out.

> has_a say that:
>
>         "If the foreign class is another Class::DBI representation we
> will call retrieve() on that class with our value."
>
> But this is not exactly what happens.  I can do a retrieve and my trigger

It's not, but it's fairly close.   Other than trigger order issues, the 
only  real limitation is that a method called on the target object can't 
assume that any attributes other than the key have been fetched.  However, 
that's also true of a straight retrieve(), so it shouldn't cause major 
coding headaches.


--
Regards,
Charles Bailey  < bailey _at_ newman _dot_ upenn _dot_ edu >
Newman Center at the University of Pennsylvania