]> git.ipfire.org Git - thirdparty/sqlalchemy/sqlalchemy.git/log
thirdparty/sqlalchemy/sqlalchemy.git
15 years agomerge default
Mike Bayer [Sat, 10 Apr 2010 23:45:34 +0000 (19:45 -0400)] 
merge default

15 years ago- starting to groom the branch for its inclusion
Mike Bayer [Sat, 10 Apr 2010 23:21:54 +0000 (19:21 -0400)] 
- starting to groom the branch for its inclusion
- one-to-many relationships now maintain a list of positive
parent-child associations within the flush, preventing
previous parents marked as deleted from cascading a
delete or NULL foreign key set on those child objects,
despite the end-user not removing the child from the old
association. [ticket:1764]
- re-established Preprocess as unique on their arguments,
as they were definitely duped in inheritance scenarios
- added a "memo" feature to UOWTransaction which represents the usual
pattern of using the .attributes collection
- added the test case from [ticket:1081] into perf/

15 years agocomment
Mike Bayer [Fri, 9 Apr 2010 22:40:28 +0000 (18:40 -0400)] 
comment

15 years agoattempt to reduce redundant calc of pk_switched
Mike Bayer [Fri, 9 Apr 2010 22:17:06 +0000 (18:17 -0400)] 
attempt to reduce redundant calc of pk_switched

15 years agofix bug + add coverage to ensure unneeded SaveUpdateAll/DeleteAll plus extra
Mike Bayer [Fri, 9 Apr 2010 22:04:18 +0000 (18:04 -0400)] 
fix bug + add coverage to ensure unneeded SaveUpdateAll/DeleteAll plus extra
work doesn't occur during per-state usage

15 years agogot DetectKeySwitch into the fold, can now greatly collapse a lot of the preprocess...
Mike Bayer [Fri, 9 Apr 2010 21:43:50 +0000 (17:43 -0400)] 
got DetectKeySwitch into the fold, can now greatly collapse a lot of the preprocess/process down

15 years agoremoves some unneeded methods, initial DetectKeySwitch not present unnecessarily
Mike Bayer [Fri, 9 Apr 2010 20:36:58 +0000 (16:36 -0400)] 
removes some unneeded methods, initial DetectKeySwitch not present unnecessarily

15 years agorefined system bywhich dependencyprocessor per-state actions
Mike Bayer [Fri, 9 Apr 2010 17:57:49 +0000 (13:57 -0400)] 
refined system bywhich dependencyprocessor per-state actions
are established

15 years agomade this easier to read
Chris Withers [Fri, 9 Apr 2010 17:02:06 +0000 (18:02 +0100)] 
made this easier to read

15 years ago- fixed numeric test for pg8000, factored out decimal/float codes
Mike Bayer [Fri, 9 Apr 2010 17:01:17 +0000 (13:01 -0400)] 
- fixed numeric test for pg8000, factored out decimal/float codes

15 years agoignore egg-info directories
Chris Withers [Fri, 9 Apr 2010 15:54:08 +0000 (16:54 +0100)] 
ignore egg-info directories

15 years agostart adding tests to ensure the size of the uow
Mike Bayer [Thu, 8 Apr 2010 22:52:04 +0000 (18:52 -0400)] 
start adding tests to ensure the size of the uow

15 years agostarting to arrange things such that unneeded executors aren't getting
Mike Bayer [Thu, 8 Apr 2010 22:21:02 +0000 (18:21 -0400)] 
starting to arrange things such that unneeded executors aren't getting
pulled into the unit of work at all.   this involves dancing around lists
of states, seeing if child objects exist, not adding excessive callcounts
while doing that, etc.

15 years agowe can load this sum ahead of time, and if there is none, we dont need a per-state...
Mike Bayer [Thu, 8 Apr 2010 18:15:11 +0000 (14:15 -0400)] 
we can load this sum ahead of time, and if there is none, we dont need a per-state/proc for this at all.
this greatly reduces unnecessary crap in the UOW for complex models.

15 years agoduh
Mike Bayer [Thu, 8 Apr 2010 01:16:11 +0000 (21:16 -0400)] 
duh

15 years ago- make it exceedlingly obvious that all topological/unitofwork code is
Mike Bayer [Thu, 8 Apr 2010 01:00:16 +0000 (21:00 -0400)] 
- make it exceedlingly obvious that all topological/unitofwork code is
being rewritten, and nothing here should be consulted for any
future activity.
- underscore current topological methods as their API behavior
is changing, possibly in 0.6.1 if [ticket:1742] remains on track

15 years agofix this test for oracle
Mike Bayer [Thu, 8 Apr 2010 00:25:37 +0000 (20:25 -0400)] 
fix this test for oracle

15 years agoadd example for changing TypeEngine compilation
Mike Bayer [Wed, 7 Apr 2010 20:21:01 +0000 (16:21 -0400)] 
add example for changing TypeEngine compilation

15 years ago- Repaired missing import in psycopg2._PGNumeric type when
Mike Bayer [Wed, 7 Apr 2010 20:07:55 +0000 (16:07 -0400)] 
- Repaired missing import in psycopg2._PGNumeric type when
unknown numeric is received.

- psycopg2/pg8000 dialects now aware of REAL[], FLOAT[],
DOUBLE_PRECISION[], NUMERIC[] return types without
raising an exception.

- introducing testing.provide_metadata for all these stupid little
create/drop tests

15 years ago- ah. oursql didn't have "extra steps" here, the previous system within execution_op...
Mike Bayer [Wed, 7 Apr 2010 19:20:20 +0000 (15:20 -0400)] 
- ah.  oursql didn't have "extra steps" here, the previous system within execution_options()
used by oursql would generate a proxied connection from within the dialect.initialize() phase.  the new
clone system bypasses that.

15 years agozoomark has additional callcount drops on this side
Mike Bayer [Wed, 7 Apr 2010 18:28:33 +0000 (14:28 -0400)] 
zoomark has additional callcount drops on this side

15 years agomerge default
Mike Bayer [Wed, 7 Apr 2010 18:26:47 +0000 (14:26 -0400)] 
merge default

15 years agoholy callcount batman
Mike Bayer [Wed, 7 Apr 2010 18:23:06 +0000 (14:23 -0400)] 
holy callcount batman

15 years agomerge default
Mike Bayer [Wed, 7 Apr 2010 18:00:23 +0000 (14:00 -0400)] 
merge default

15 years ago- Added new 'compiled_cache' execution option. A dictionary
Mike Bayer [Wed, 7 Apr 2010 17:59:18 +0000 (13:59 -0400)] 
- Added new 'compiled_cache' execution option.  A dictionary
where Compiled objects will be cached when the Connection
compiles a clause expression into a dialect- and parameter-
specific Compiled object.  It is the user's responsibility to
manage the size of this dictionary, which will have keys
corresponding to the dialect, clause element, the column
names within the VALUES or SET clause of an INSERT or UPDATE,
as well as the "batch" mode for an INSERT or UPDATE statement.

15 years ago- Fixed bug in execution_options() feature whereby the existing
Mike Bayer [Wed, 7 Apr 2010 17:42:31 +0000 (13:42 -0400)] 
- Fixed bug in execution_options() feature whereby the existing
Transaction and other state information from the parent
connection would not be propagated to the sub-connection.

15 years agough, didn't mean to commit that :-S
Chris Withers [Wed, 7 Apr 2010 17:14:06 +0000 (18:14 +0100)] 
ugh, didn't mean to commit that :-S

15 years agomerge in tip?
Chris Withers [Wed, 7 Apr 2010 17:10:32 +0000 (18:10 +0100)] 
merge in tip?

15 years agoAdd instructions to install nose adn NB that it won't happen by magic
Chris Withers [Wed, 7 Apr 2010 17:09:21 +0000 (18:09 +0100)] 
Add instructions to install nose adn NB that it won't happen by magic

15 years agomerge from default
Mike Bayer [Wed, 7 Apr 2010 16:31:15 +0000 (12:31 -0400)] 
merge from default

15 years ago- dict_ becomes explicit on _get_state_attr_by_column, _set_state_attr_by_column...
Mike Bayer [Wed, 7 Apr 2010 16:30:02 +0000 (12:30 -0400)] 
- dict_ becomes explicit on _get_state_attr_by_column, _set_state_attr_by_column, others,
to reduce on expensive state.dict calls.
- internal getattr(), setattr(), getcommitted() methods
on ColumnProperty, CompositeProperty, RelationshipProperty
have been underscored, signature has changed.

15 years ago- Postgresql now reflects sequence names associated with
Mike Bayer [Tue, 6 Apr 2010 22:53:51 +0000 (18:53 -0400)] 
- Postgresql now reflects sequence names associated with
SERIAL columns correctly, after the name of of the sequence
has been changed.  Thanks to Kumar McMillan for the patch.
[ticket:1071]

15 years ago- added missing coverage for self-referential many-to-many flushes
Mike Bayer [Tue, 6 Apr 2010 22:28:40 +0000 (18:28 -0400)] 
- added missing coverage for self-referential many-to-many flushes
- some other areas where per-state deps are called and an empty result returned
are still lacking coverage.

15 years agoremove mostly unneeded calls to conditional_post_update
Mike Bayer [Tue, 6 Apr 2010 18:16:07 +0000 (14:16 -0400)] 
remove mostly unneeded calls to conditional_post_update

15 years ago- EdgeCollection can now go away
Mike Bayer [Tue, 6 Apr 2010 17:20:31 +0000 (13:20 -0400)] 
- EdgeCollection can now go away
- fix reflection test

15 years agoa RandomSet implementation useful for swapping into topological
Mike Bayer [Tue, 6 Apr 2010 17:07:17 +0000 (13:07 -0400)] 
a RandomSet implementation useful for swapping into topological

15 years agoadd batch execute to ProcessState
Mike Bayer [Tue, 6 Apr 2010 16:54:29 +0000 (12:54 -0400)] 
add batch execute to ProcessState

15 years agomerge default
Mike Bayer [Tue, 6 Apr 2010 16:44:35 +0000 (12:44 -0400)] 
merge default

15 years agotest another version of the mixin here
Mike Bayer [Tue, 6 Apr 2010 16:32:51 +0000 (12:32 -0400)] 
test another version of the mixin here

15 years ago- Further reworked the "mixin" logic in declarative to
Mike Bayer [Tue, 6 Apr 2010 16:27:01 +0000 (12:27 -0400)] 
- Further reworked the "mixin" logic in declarative to
additionally allow __mapper_args__ as a @classproperty
on a mixin, such as to dynamically assign polymorphic_identity.

15 years ago- Fixed an error in expression typing which caused an endless
Mike Bayer [Tue, 6 Apr 2010 15:39:09 +0000 (11:39 -0400)] 
- Fixed an error in expression typing which caused an endless
loop for expressions with two NULL types.

15 years agoremove prints
Mike Bayer [Tue, 6 Apr 2010 05:42:58 +0000 (01:42 -0400)] 
remove prints

15 years ago- cleanup, factoring, had some heisenbugs. more test coverage
Mike Bayer [Tue, 6 Apr 2010 05:23:54 +0000 (01:23 -0400)] 
- cleanup, factoring, had some heisenbugs.   more test coverage
 will be needed overall as missing dependency rules lead
to subtle bugs pretty easily

15 years agoall tests pass with this version
Mike Bayer [Tue, 6 Apr 2010 03:21:02 +0000 (23:21 -0400)] 
all tests pass with this version

15 years agobeginning to get post_update working, will need more tests
Mike Bayer [Tue, 6 Apr 2010 01:29:51 +0000 (21:29 -0400)] 
beginning to get post_update working, will need more tests

15 years agobreakthrough, wow
Mike Bayer [Mon, 5 Apr 2010 22:48:57 +0000 (18:48 -0400)] 
breakthrough, wow

15 years agoso here is kind of the idea. but it doesn't work like it used to.
Mike Bayer [Mon, 5 Apr 2010 21:49:58 +0000 (17:49 -0400)] 
so here is kind of the idea.   but it doesn't work like it used to.
so I think I want to try to build a smarter "find everything without a dependency"
system that is more inline with how this is running now anyway - i.e.
go through the whole list, find nodes with no dependencies.  maybe the
original topological.sort() can do that, not sure.

15 years agolooks like most of the issues are because we're losing insert ordering
Mike Bayer [Mon, 5 Apr 2010 21:24:00 +0000 (17:24 -0400)] 
looks like most of the issues are because we're losing insert ordering
on cycles.  so lets reintroduce the organize as tree component, which
works here.   still need to make it meaningful by teaching the save/delete state
actions to receive a set of items to match up

15 years ago- many-to-many is up.
Mike Bayer [Mon, 5 Apr 2010 20:52:56 +0000 (16:52 -0400)] 
- many-to-many is up.
- now running on the full suite of tests.   unsurprisingly, it appears
there are subtle self-referential issues causing many tests to fail.

15 years agobetter assertions, the concern is that an object will change state and not have been
Mike Bayer [Mon, 5 Apr 2010 20:13:51 +0000 (16:13 -0400)] 
better assertions, the concern is that an object will change state and not have been
appropriately preprocessed

15 years agothese are some assertions that would make things easier, if they can be maintained.
Mike Bayer [Mon, 5 Apr 2010 18:53:19 +0000 (14:53 -0400)] 
these are some assertions that would make things easier, if they can be maintained.
otherwise, if listonly can go from True to False, then there's the whole cascades issue
to deal with, same for delete moving from False to True and vice versa.

15 years ago- enabled the DetectKeySwitch, and additionally added that it need
Mike Bayer [Mon, 5 Apr 2010 18:49:35 +0000 (14:49 -0400)] 
- enabled the DetectKeySwitch, and additionally added that it need
not execute at all when a one-to-many is present on the reverse side.
- OneToMany can establish a state as "listonly" when passive_updates are enabled
and the change is due to key switch.

15 years agobranch merge
Mike Bayer [Mon, 5 Apr 2010 17:48:52 +0000 (13:48 -0400)] 
branch merge

15 years agobetter yet
Mike Bayer [Mon, 5 Apr 2010 17:16:29 +0000 (13:16 -0400)] 
better yet

15 years agoclarify intent and reduce lookups here
Mike Bayer [Mon, 5 Apr 2010 17:09:24 +0000 (13:09 -0400)] 
clarify intent and reduce lookups here

15 years ago- further reduce what topological has to do, expects full list of nodes
Mike Bayer [Sun, 4 Apr 2010 16:24:01 +0000 (12:24 -0400)] 
- further reduce what topological has to do, expects full list of nodes
- fix some side-effect-dependent behaviors in uow.  we can now
unconditionally remove "disabled" actions without rewriting

15 years agothis fully moves the per_state methods to work with all states at once
Mike Bayer [Sun, 4 Apr 2010 15:29:41 +0000 (11:29 -0400)] 
this fully moves the per_state methods to work with all states at once

15 years agomoving the per-states operations to be aggreagted on the endpoints to reduce
Mike Bayer [Sun, 4 Apr 2010 15:08:00 +0000 (11:08 -0400)] 
moving the per-states operations to be aggreagted on the endpoints to reduce
procesing overhead.

15 years agomerge default branch
Mike Bayer [Sun, 4 Apr 2010 14:38:29 +0000 (10:38 -0400)] 
merge default branch

15 years ago- apparently [ticket:1761] was covered in tests already.
Mike Bayer [Sun, 4 Apr 2010 14:18:16 +0000 (10:18 -0400)] 
- apparently [ticket:1761] was covered in tests already.
- Usage of version_id_col on a backend that supports
cursor.rowcount for execute() but not executemany() now works
when a delete is issued (already worked for saves, since those
don't use executemany()). For a backend that doesn't support
cursor.rowcount at all, a warning is emitted the same
as with saves.  [ticket:1761]

15 years agocleanup and callcount reduction in mapper._save_obj, _delete_obj.
Mike Bayer [Sun, 4 Apr 2010 01:42:41 +0000 (21:42 -0400)] 
cleanup and callcount reduction in mapper._save_obj, _delete_obj.
includes an untested fix for [ticket:1761]

15 years agorow switch works for post-cycle sorts too...just needed more data in the deps. ...
Mike Bayer [Sat, 3 Apr 2010 23:41:56 +0000 (19:41 -0400)] 
row switch works for post-cycle sorts too...just needed more data in the deps.   the pattern here
is more data needed for each tweak.

15 years agogot row switch more or less up
Mike Bayer [Sat, 3 Apr 2010 22:24:04 +0000 (18:24 -0400)] 
got row switch more or less up

15 years agotweak how we indicate child deleted here
Mike Bayer [Sat, 3 Apr 2010 22:08:48 +0000 (18:08 -0400)] 
tweak how we indicate child deleted here

15 years agoo2m/m2o pretty much there, minus post update.
Mike Bayer [Sat, 3 Apr 2010 22:05:33 +0000 (18:05 -0400)] 
o2m/m2o pretty much there, minus post update.

15 years ago- basic inheritance behavior
Mike Bayer [Sat, 3 Apr 2010 21:03:12 +0000 (17:03 -0400)] 
- basic inheritance behavior
- we do need dependencies between an object and its dep when the other object
has no save or delete pending. the other object
like before isn't needed, but right now we make the dependency just 'None',
and it gets thrown away.

15 years ago- Now using cx_oracle output converters so that the
Mike Bayer [Sat, 3 Apr 2010 19:33:55 +0000 (15:33 -0400)] 
- Now using cx_oracle output converters so that the
DBAPI returns natively the kinds of values we prefer:
- NUMBER values with positive precision + scale convert
to cx_oracle.STRING and then to Decimal.   This
allows perfect precision for the Numeric type when
using cx_oracle.  [ticket:1759]
- STRING/FIXED_CHAR now convert to unicode natively.
SQLAlchemy's String types then don't need to
apply any kind of conversions.

15 years agodont create tables here
Mike Bayer [Sat, 3 Apr 2010 18:08:45 +0000 (14:08 -0400)] 
dont create tables here

15 years agocorrect changelog
Mike Bayer [Sat, 3 Apr 2010 15:09:58 +0000 (11:09 -0400)] 
correct changelog

15 years agobranch merge
Mike Bayer [Sat, 3 Apr 2010 15:00:19 +0000 (11:00 -0400)] 
branch merge

15 years ago- the Numeric type raises an *enormous* warning when expected
Mike Bayer [Sat, 3 Apr 2010 14:58:13 +0000 (10:58 -0400)] 
- the Numeric type raises an *enormous* warning when expected
to convert floats to Decimal from a DBAPI that returns floats.
This includes SQLite, Oracle, Sybase, MS-SQL.
[ticket:1759]

15 years agoadd FunctionElement example
Mike Bayer [Fri, 2 Apr 2010 21:45:10 +0000 (17:45 -0400)] 
add FunctionElement example

15 years ago- The ORM will set the docstring of all generated descriptors
Mike Bayer [Fri, 2 Apr 2010 21:22:16 +0000 (17:22 -0400)] 
- The ORM will set the docstring of all generated descriptors
to None by default.  This can be overridden using 'doc'
(or if using Sphinx, attribute docstrings work too).

- Added kw argument 'doc' to all mapper property callables
as well as Column().  Will assemble the string 'doc' as
the '__doc__' attribute on the descriptor.

15 years agobranch merge
Mike Bayer [Fri, 2 Apr 2010 20:27:59 +0000 (16:27 -0400)] 
branch merge

15 years agoremove erronrous link to "addresses"
Mike Bayer [Fri, 2 Apr 2010 20:15:17 +0000 (16:15 -0400)] 
remove erronrous link to "addresses"

15 years ago... also when changing the row dynamically
Gaëtan de Menten [Fri, 2 Apr 2010 18:34:02 +0000 (20:34 +0200)] 
... also when changing the row dynamically

15 years ago- made the C version of RowProxy accept any sequence for the row, instead of
Gaëtan de Menten [Fri, 2 Apr 2010 18:29:37 +0000 (20:29 +0200)] 
- made the C version of RowProxy accept any sequence for the row, instead of
  only tuples

15 years ago- Updated attribute_shard.py example to use a more robust
Mike Bayer [Fri, 2 Apr 2010 17:10:55 +0000 (13:10 -0400)] 
- Updated attribute_shard.py example to use a more robust
method of searching a Query for binary expressions which
compare columns against literal values.

15 years ago- Declarative will raise an informative error message
Mike Bayer [Fri, 2 Apr 2010 16:42:54 +0000 (12:42 -0400)] 
- Declarative will raise an informative error message
if a non-mapped class attribute is referenced in the
string-based relationship() arguments.

15 years agoworking through cycles tests...
Mike Bayer [Thu, 1 Apr 2010 23:06:35 +0000 (19:06 -0400)] 
working through cycles tests...

15 years agomany-to-one completed for self-referential
Mike Bayer [Thu, 1 Apr 2010 22:40:11 +0000 (18:40 -0400)] 
many-to-one completed for self-referential

15 years agothe delete parent o2m test
Mike Bayer [Thu, 1 Apr 2010 20:23:02 +0000 (16:23 -0400)] 
the delete parent o2m test

15 years agorefactor dependency elements
Mike Bayer [Thu, 1 Apr 2010 20:13:06 +0000 (16:13 -0400)] 
refactor dependency elements

15 years agothis version passes one to many tests so far
Mike Bayer [Thu, 1 Apr 2010 20:00:57 +0000 (16:00 -0400)] 
this version passes one to many tests so far

15 years agoself-referential working to a small degree
Mike Bayer [Thu, 1 Apr 2010 17:19:14 +0000 (13:19 -0400)] 
self-referential working to a small degree

15 years agoone more rule
Mike Bayer [Thu, 1 Apr 2010 16:50:57 +0000 (12:50 -0400)] 
one more rule

15 years ago- id(obj) is no longer used internally within topological.py,
Mike Bayer [Thu, 1 Apr 2010 16:45:51 +0000 (12:45 -0400)] 
- id(obj) is no longer used internally within topological.py,
as the sorting functions now require hashable objects
only.  [ticket:1756]

15 years agobeginning to address cycles but its not worked out yet
Mike Bayer [Wed, 31 Mar 2010 21:31:34 +0000 (17:31 -0400)] 
beginning to address cycles but its not worked out yet

15 years agodeletes, rudimentary many-to-ones
Mike Bayer [Wed, 31 Mar 2010 19:46:48 +0000 (15:46 -0400)] 
deletes, rudimentary many-to-ones

15 years agoreally got topological going. now that we aren't putting fricking mapped objects...
Mike Bayer [Wed, 31 Mar 2010 19:12:29 +0000 (15:12 -0400)] 
really got topological going.   now that we aren't putting fricking mapped objects into
it all that id() stuff can go

15 years ago- 'cycle' is a stack here - needs to be a list.
Mike Bayer [Wed, 31 Mar 2010 17:52:57 +0000 (13:52 -0400)] 
- 'cycle' is a stack here - needs to be a list.

15 years agoits alive !
Mike Bayer [Wed, 31 Mar 2010 17:25:13 +0000 (13:25 -0400)] 
its alive !

15 years ago- Restored some bind-labeling logic from 0.5 which ensures
Mike Bayer [Wed, 31 Mar 2010 17:01:40 +0000 (13:01 -0400)] 
- Restored some bind-labeling logic from 0.5 which ensures
that tables with column names that overlap another column
of the form "<tablename>_<columnname>" won't produce
errors if column._label is used as a bind name during
an UPDATE.  Test coverage which wasn't present in 0.5
has been added.  [ticket:1755]

15 years agoThis is turning out to be a rewrite of the accounting system of unitofwork.py, but...
Mike Bayer [Tue, 30 Mar 2010 23:19:41 +0000 (19:19 -0400)] 
This is turning out to be a rewrite of the accounting system of unitofwork.py, but the overarching
method of doing things stays the same.   it should be easy to add new dependencies
between actions and to change the structure of how things are done.

15 years agomerge trunk. Re-instating topological._find_cycles for the moment
Mike Bayer [Tue, 30 Mar 2010 22:15:02 +0000 (18:15 -0400)] 
merge trunk.   Re-instating topological._find_cycles for the moment

15 years agofix pk violation error
Mike Bayer [Tue, 30 Mar 2010 14:41:07 +0000 (10:41 -0400)] 
fix pk violation error

15 years ago- the compiler extension now allows @compiles decorators
Mike Bayer [Tue, 30 Mar 2010 14:39:36 +0000 (10:39 -0400)] 
- the compiler extension now allows @compiles decorators
on base classes that extend to child classes, @compiles
decorators on child classes that aren't broken by a
@compiles decorator on the base class.

15 years ago- Session.refresh() now does an equivalent expire()
Mike Bayer [Mon, 29 Mar 2010 21:56:02 +0000 (17:56 -0400)] 
- Session.refresh() now does an equivalent expire()
on the given instance first, so that the "refresh-expire"
cascade is propagated.   Previously, refresh() was
not affected in any way by the presence of "refresh-expire"
cascade.   This is a change in behavior versus that
of 0.6beta2, where the "lockmode" flag passed to refresh()
would cause a version check to occur.  Since the instance
is first expired, refresh() always upgrades the object
to the most recent version.

- The 'refresh-expire' cascade, when reaching a pending object,
will expunge the object if the cascade also includes
"delete-orphan", or will simply detach it otherwise.
[ticket:1754]

15 years agobranch merge
Mike Bayer [Sun, 28 Mar 2010 20:41:58 +0000 (16:41 -0400)] 
branch merge

15 years agotag merge
Mike Bayer [Sun, 28 Mar 2010 20:41:30 +0000 (16:41 -0400)] 
tag merge