[postmodern-devel] Can't redefine slots as columns - why?

[Marijn Haverbeke]

Hi Luke,

I have to admit that I don't really remember how direct slots and
effective slots are supposed to work. If everything works as intended
for you, I'll drop the test and allow slots from superclasses to be
defined as column slots.

Best,
Marijn

On Tue, Mar 10, 2009 at 8:00 PM, Lucas Hope <lucas.r.hope at gmail.com> wrote:
> Hi Marijn,
>
> I guess I suffered from working 12 hours and then emailing you.
>
> My little example had a fatal flaw: I hadn't defined an :initarg for
> radius!  Some preliminary (re-)testing indicates that your initial
> fix/alteration works.
>
> Thanks for your help!
>
> -Luke
>
>
> On Tue, Mar 10, 2009 at 9:51 PM, Marijn Haverbeke <marijnh at gmail.com> wrote:
>> Ah, you are right. If the slot object has already been instantiated as
>> a regular slot, I don't know of any safe, portable way to make it a
>> column slot. If you can think of a hack to make this work, let me
>> know. But for now, you'll have to solve this in some other way.
>>
>> Best,
>> Marijn
>>
>>
>> On Tue, Mar 10, 2009 at 10:10 AM, Lucas Hope <lucas.r.hope at gmail.com> wrote:
>>> Hi Marijn,
>>>
>>> I stubbed out the form and the dao-circle was successfully compiled (of course).
>>>
>>> However the example circle I tried to make did not have its slot value
>>> set.  I understand (I think) what you were trying to do with the
>>> *direct-column-slot* special variable, but maybe in this case (on
>>> Allegro CL, btw),  while the effective-slot-definition-class works
>>> correctly (indicated by the correct CREATE TABLE below),  the actual
>>> effective-column-slot object is instantiated OUTSIDE the scope of the
>>> binding in compute-effective-slot-definitio which sets
>>> *direct-column-slot* to non-nil.
>>>
>>> There's no reason for the lisp implementation to instantiate the class
>>> returned by effective-column-slot immediately, as far as I can tell.
>>>
>>> (dao-table-definition 'dao-circle) ;; success!
>>> "CREATE TABLE dao_circle (radius REAL NOT NULL)"
>>>
>>> (setf circle (make-instance 'dao-circle :radius 5.0))
>>>
>>> (slot-boundp circle 'radius)
>>> => NIL ;; fail. :(
>>>
>>> -Luke
>>>
>>> On Tue, Mar 10, 2009 at 6:19 PM, Marijn Haverbeke <marijnh at gmail.com> wrote:
>>>> Hello Luke,
>>>>
>>>> I think (though I'm never entirely sure with this MOP stuff) that that
>>>> error is only there becuase I hadn't thought of the way you are trying
>>>> to use this metaclass, and assumed redefining a slot as a column was
>>>> always an accident. Could you try killing the WHEN-form on line 97 of
>>>> postmodern/table.lisp, doing what you're trying to do, and reporting
>>>> back whether it works as hoped?
>>>>
>>>> Best,
>>>> Marijn
>>>>
>>>>
>>>> On Tue, Mar 10, 2009 at 1:11 AM, Lucas Hope <lucas.r.hope at gmail.com> wrote:
>>>>> Hi there,
>>>>>
>>>>> I am just getting my feet wet with postmodern.  It looks like I will
>>>>> be doing some heavy development with it for the next couple of weeks.
>>>>> Wish me luck!
>>>>>
>>>>> My (first?) issue is that I want to do something similar to the
>>>>> following example code.  Here CIRCLE is an in-memory implementation of
>>>>> a circle, and it is desired to create a DAO object which moves storage
>>>>> to a postgres backend. (We have lots of circles...)
>>>>>
>>>>> Pretend CIRCLE has a host of useful functionality attached and I can't
>>>>> just point all the defmethods to DAO-CIRCLE.
>>>>>
>>>>> (defclass SHAPE () ())
>>>>>
>>>>> (defclass CIRCLE (SHAPE)
>>>>>  (radius))
>>>>>
>>>>> (defclass DAO-CIRCLE (CIRCLE)
>>>>>  ((radius :col-type real))
>>>>>  (:metaclass dao-class))
>>>>>
>>>>> (dao-table-definition 'dao-circle)
>>>>> => ERROR: Slot RADIUS in class #<DAO-CLASS DAO-CIRCLE> is both a
>>>>> column slot and a regular slot.
>>>>>
>>>>> Why doesn't this work, and how can I make it work?  It seems like this
>>>>> is useful functionality to have.  I want the DAO-CIRCLE's radius to be
>>>>> a column slot.
>>>>>
>>>>> Cheers,
>>>>>
>>>>> -Luke
>
> _______________________________________________
> postmodern-devel mailing list
> postmodern-devel at common-lisp.net
> http://common-lisp.net/cgi-bin/mailman/listinfo/postmodern-devel
>