optimize flush() and/or locate_dirty() and/or autoflush behavior

looks like now that we have autoflush=True, flushes on every query are expensive, and complaints are coming in. the locate_dirty() call is pretty expensive and we should figure out a way to not need its normal behavior on every hit.

we probably want to keep a log of all mappers which include "mutable" objects and localize the is_modifed() call to just those objects...in all other cases, we can just hit "modified" flags directly on instances.

Comments (2)