[storm] referenceset and lazy attributes

[Gustavo Niemeyer]

> I think a attribute-group implementation would be the best where we can
> define which attributes are selected together. So in the example below
> url and title would be fetched together.
> 
> class A(object):
> 
>     url = Unicode(group='smalltexts')
>     title = Unicode(group='smalltexts')
>     toc = Unicode(group='hugetexgts')

One issue with that is that if you load 'toc', it wouldn't retrieve 'url'
and 'title', which are most probably wanted.  I wonder if we should just
stick with a simpler implementation.  E.g.:

  class A(object):
      url = Unicode()
      title = Unicode()
      toc = Unicode(lazy=False)
      data = Unicode(lazy=False)

Can you imagine a use case where this wouldn't be enough?

Also, I wonder if 'toc' and 'data' should be loaded together, or if they
should be fetched individually on demand.

-- 
Gustavo Niemeyer
http://niemeyer.net