Allow fields to be pickled
|Reported by:||akaariai||Owned by:||nobody|
|Component:||Database layer (models, ORM)||Version:||master|
|Has patch:||yes||Needs documentation:||no|
|Needs tests:||no||Patch needs improvement:||no|
Model field instances are not pickleable currently because the field's model isn't necessarily available (autocreated m2m models for example). This causes some ugly code in django.db.models.sql.query.Query __getstate__ and __setstate__ methods as we need to avoid pickling the fields found from the various structs in the Query.
To avoid this problem we can use __reduce__ to make fields pickleable. The idea is that we save the field's identifier (app_lable, model_name, field_name) in pickling, then use app cache to load the model, and then the field from the model in unpickling. This makes the Query code much nicer, and allows for extended use of fields in the Query structures.
Patch available from https://github.com/akaariai/django/tree/field_pickle