Package twisted.popsicle
Persistent Object Partitioning Strategy Involving Concurrent
Lightweight Events
POPSICLE is designed to persist objects as close to transparently as
possible. While complete transparency is impossible with python's
semantics, popsicle strives to preserve three things:
-
preserve a modicum of referential transparency by using Deferreds for
all references which can not (or should not!) be immediately
resolved
-
allow for the use of multiple persistence strategies with the same
object. Specific back-ends we wish to target: filesystem databases,
BSDDB, ReiserFS, ZODB, and arbitrary RDBMS schemas.
-
minimize memory usage by paging unnecessary objects out whenever
possible, transparently to everything involved. We want to use
Python's normal garbage collection mechanisms to accomplish this.
There is a top level persistence manager (the 'freezer') which keeps a
weak key dictionary of {persistent_object: [PersistentReference,
[savers]]}
The persistent_object should be 'trained' to expect callable objects
that return Deferreds to populate part of it. Whenever a persistence
mechanism encounters an ID for a different object that could be
loaded...
Each Storage is responsible for its own cache management wrt IDs.
Objects that are dirty should call popsicle.dirty(self) to be added to
the dirty list. when popsicle.clean() is called, the dirty list will be
walked, notifying all savers for each dirty object.
Modules