On 9/15/05, Sune Kirkeby <[EMAIL PROTECTED]> wrote: > Couldn't django auto-fill much of module_constants > from co_freevars in the code-objects of the methods? That > would be nice.
That's possible, but I've hesitated to do so because models have a lot of global variables available to them -- basically, a lot of random crap such as the other model classes. Is it better to grab more or fewer global variables? I've had a preference for fewer, but I'd love to get other opinions on it. > Hmm... And, how about a "MODULE_LEVEL"-class inside > the model-classes, whose attributes are copied into the > genereated module? I think I like that more than class-attributes, > since it emphasises that black magic is occuring behind the > scenes (I think it helped me to understand what was really > going on, that the behind-the-scenes magic was not too > hidden...) > > Then one could put "def gnah" in MODULE_LEVEL instead of > "def _module_gnah" in the model, also exceptions and the > module_constants that cannot be auto-deduced could > go in there... That's a great idea. It would make it more explicit that module-level magic is happening, and it would remove the need for the ugly "_module" prefixes. And it can be backwards-compatible (for at least a short while), because we could support both approaches. What do people think? Adrian -- Adrian Holovaty holovaty.com | djangoproject.com | chicagocrime.org
