[Grok-dev] Persistence issue with grok.Model

[Brandon Craig Rhodes]

"Paul Wilson" <paulalexwilson at gmail.com> writes:

>>>> class C(object):
> ...     a = 0
> ...     def __getattribute__(*args):
> ...             print "Intercepting", args
> ...             return object.__getattribute__(*args)
> ...
>>>> c = C()
>>>> c.a
> Intercepting (<__main__.C object at 0x4024948c>, 'a')
> 0
>>>> c.a = 0
>>>> c.__dict__
> Intercepting (<__main__.C object at 0x4024948c>, '__dict__')
> {'a': 0}
>
> My interpretation of what you said was that the "Intercepting" line
> would be omitted from the c.__dict__ lookup.

You were correctly interpreting me, and it looks like I was wrong!  The
rule as I should have stated it is:

    "When object.__getattribute__ is asked for the attribute '__dict__',
    then instead of looking inside the object's dictionary for the name
    '__dict__' or else invoking the method __getattr__ to retrieve the
    value, it short-circuits and immediately returns the object's
    attribute dictionary without invoking any of the __getattr__
    machinery that intercepts any other attribute."

Turning to the ZODB, I should really have perused its documentation
before trying to answer you. :-/ Take a look at:

http://svn.zope.org/ZODB/trunk/src/persistent/cPersistence.c?rev=91565&view=auto

and look at the comment above the function named "unghost_getattr".  It
looks (from my cursory inspection) as though an object freshly pulled
from the ZODB is a "ghost" without object attributes, until you ask for
one of the attributes that is not in the list shown in that method.

But, as you've found already, I could be wrong. :-)

-- 
Brandon Craig Rhodes   brandon at rhodesmill.org   http://rhodesmill.org/brandon