:class:`.AttributeImpl` as an "initiator" token has been changed;
the object is now an event-specific object called :class:`.attributes.Event`.
Additionally, the attribute system no longer halts events based
on a matching "initiator" token; this logic has been moved to be
specific to ORM backref event handlers, which are the typical source
of the re-propagation of an attribute event onto subsequent append/set/remove
operations. End user code which emulates the behavior of backrefs
must now ensure that recursive event propagation schemes are halted,
if the scheme does not use the backref handlers. Using this new system,
backref handlers can now peform a
"two-hop" operation when an object is appended to a collection,
associated with a new many-to-one, de-associated with the previous
many-to-one, and then removed from a previous collection. Before this
change, the last step of removal from the previous collection would
not occur.
[ticket:2789]
become an externally usable package but still remains within the main sqlalchemy parent package.
in this system, we use kind of an ugly hack to get the noseplugin imported outside of the
"sqlalchemy" package, while still making it available within sqlalchemy for usage by
third party libraries.
- remove() on a scalar object will raise if the object
removed is not what was present.
- InstanceState can be pickled if obj() is None; this
to support the other changes in this commit
- only use trackparent flag on attributes if
single_parent or ONETOMANY; otherwise we can skip this
overhead
- attribute hasparent()/sethasparent() check that trackparent
is set, else their usage is invalid
- [bug] Fixed backref behavior when "popping" the
value off of a many-to-one in response to
a removal from a stale one-to-many - the operation
is skipped, since the many-to-one has since
been updated. [ticket:2315]
- [bug] After some years of not doing this, added
more granularity to the "is X a parent of Y"
functionality, which is used when determining
if the FK on "Y" needs to be "nulled out" as well
as if "Y" should be deleted with delete-orphan
cascade. The test now takes into account the
Python identity of the parent as well its identity
key, to see if the last known parent of Y is
definitely X. If a decision
can't be made, a StaleDataError is raised. The
conditions where this error is raised are fairly
rare, requiring that the previous parent was
garbage collected, and previously
could very well inappropriately update/delete
a record that's since moved onto a new parent,
though there may be some cases where
"silent success" occurred previously that will now
raise in the face of ambiguity.
Expiring "Y" resets the "parent" tracker, meaning
X.remove(Y) could then end up deleting Y even
if X is stale, but this is the same behavior
as before; it's advised to expire X also in that
case. [ticket:2264]
backrefs, typically when autoflush=False, where
the back-referenced collection wouldn't
properly handle add/removes with no net
change. Thanks to Richard Murri for the
test case + patch. [ticket:2123]
(also in 0.6.7).
access to the cls/self.tables/classes registries
- express orm/_base.py ORMTest in terms of engine/_base.py TablesTest,
factor out common steps into TablesTest, remove AltEngineTest as a
separate class. will further consolidate these base classes
outside of "sqlalchemy" and under "test/".
Rationale:
- coverage plugin works without issue, without need for an awkward
additional package install
- command line for "nosetests" isn't polluted with SQLAlchemy options
[ticket:1949]
relationship(), to eliminate confusion over the relational
algebra term. relation() however will remain available
in equal capacity for the foreseeable future. [ticket:1740]
- renamed PASSIVE_NO_CALLABLES to PASSIVE_NO_FETCH
- passive now propagates all the way through lazy callables,
all the way into query._get(), so that many-to-one lazy load
can load the instance via the local session but not trigger
any SQL if not available, fixes [ticket:1298] without
messing up consistency of tests added in r6201
- many-to-one also handles returning PASSIVE_NO_RESULT
for the "old" value thus eliminating the need for the
previous value even if the new value is None
- query._get() uses identity_map.get(), which has been
changed to no longer raise KeyError, thus providing
mythical time savings that didn't seem to make any
difference in how fast the unit tests ran.
relation(). When a collection is mutated, many-to-one
backrefs on the other side will not fire off to load
the "old" value, unless "single_parent=True" is set.
A direct assignment of a many-to-one still loads
the "old" value in order to update backref collections
on that value, which may be present in the session
already, thus maintaining the 0.5 behavioral contract.
[ticket:1483]
Alex Gaynor