This patch iproved two major parts in Django. First, the fields.related
was refactored. The main addition there was ForeignObject. Second, the
ORM now handles multicolumn joins in most cases, though there are still
cases that do not work correcly (split_exclude() for example).
In addition there were extesive changes to how GenericRelation works.
Before it was a fake m2m field, now it is a pure virtual fields and is
based on ForeignObject.
There is still much room for improvement. The related fields code is
still somewhat confusing, and how fields are represented in model._meta
should also be revisited.
This patch was written mostly by Jeremy Tillman with some final polish
by the committer.
Previously, depending on the database backend or the cursor type,
you'd need to double the percent signs in the query before passing
it to cursor.execute. Now cursor.execute consistently need percent
doubling whenever params argument is not None (placeholder substitution
will happen).
Thanks Thomas Güttler for the report and Walter Doekes for his work
on the patch.
This is provided as a new "validate_max" formset_factory option defaulting to
False, since the non-validating behavior of max_num is longstanding, and there
is certainly code relying on it. (In fact, even the Django admin relies on it
for the case where there are more existing inlines than the given max_num). It
may be that at some point we want to deprecate validate_max=False and
eventually remove the option, but this commit takes no steps in that direction.
This also fixes the DoS-prevention absolute_max enforcement so that it causes a
form validation error rather than an IndexError, and ensures that absolute_max
is always 1000 more than max_num, to prevent surprising changes in behavior
with max_num close to absolute_max.
Lastly, this commit fixes the previous inconsistency between a regular formset
and a model formset in the precedence of max_num and initial data. Previously
in a regular formset, if the provided initial data was longer than max_num, it
was truncated; in a model formset, all initial forms would be displayed
regardless of max_num. Now regular formsets are the same as model formsets; all
initial forms are displayed, even if more than max_num. (But if validate_max is
True, submitting these forms will result in a "too many forms" validation
error!) This combination of behaviors was chosen to keep the max_num validation
simple and consistent, and avoid silent data loss due to truncation of initial
data.
Thanks to Preston for discussion of the design choices.
The ticket dealt with a case where one query had .exclude() that
produced a subquery, the other query had a join to the same model that
was subqueried in the first query. This was already fixed in master, so
only test added.
This reverts commit 2cd0edaa47.
This commit was the cause of a memory leak. See ticket for more details.
Thanks Anssi Kääriäinen for identifying the source of the bug.
Before this change, the get_admin_log method would expect User model's
FK to be named `id`. When changing that FK name, admin/index.html
rendering would fail.
This includes:
* Changed the use of id for the use of pk property.
* Added a regression test that fails without the patch.
This commit refs #20088.
The original problem was that when filtering two levels up in
inheritance chain, Django optimized the join generation so that the
middle model was skipped. But then Django generated joins from top
to middle to bottom for SELECT clause, and thus there was one extra
join (top->middle->bottom + top -> bottom).
This case is fixed in master as the filtering optimization is gone.
This has the side effect that in some cases there is still extra join
if the SELECT clause doesn't contain anything from middle or bottom.
Model.save() will use UPDATE - if not updated - INSERT instead of
SELECT - if found UPDATE else INSERT. This should save a query when
updating, but will cost a little when inserting model with PK set.
Also fixed#17341 -- made sure .save() commits transactions only after
the whole model has been saved. This wasn't the case in model
inheritance situations.
The save_base implementation was refactored into multiple methods.
A typical chain for inherited save is:
save_base()
_save_parents(self)
for each parent:
_save_parents(parent)
_save_table(parent)
_save_table(self)
Thanks Anssi for haggling until I implemented this.
This change alleviates the need for atomic_if_autocommit. When
autocommit is disabled for a database, atomic will simply create and
release savepoints, and not commit anything. This honors the contract of
not doing any transaction management.
This change also makes the hack to allow using atomic within the legacy
transaction management redundant.
None of the above will work with SQLite, because of a flaw in the design
of the sqlite3 library. This is a known limitation that cannot be lifted
without unacceptable side effects eg. triggering arbitrary commits.
The sql/query.py add_q method did a lot of where/having tree hacking to
get complex queries to work correctly. The logic was refactored so that
it should be simpler to understand. The new logic should also produce
leaner WHERE conditions.
The changes cascade somewhat, as some other parts of Django (like
add_filter() and WhereNode) expect boolean trees in certain format or
they fail to work. So to fix the add_q() one must fix utils/tree.py,
some things in add_filter(), WhereNode and so on.
This commit also fixed add_filter to see negate clauses up the path.
A query like .exclude(Q(reversefk__in=a_list)) didn't work similarly to
.filter(~Q(reversefk__in=a_list)). The reason for this is that only
the immediate parent negate clauses were seen by add_filter, and thus a
tree like AND: (NOT AND: (AND: condition)) will not be handled
correctly, as there is one intermediary AND node in the tree. The
example tree is generated by .exclude(~Q(reversefk__in=a_list)).
Still, aggregation lost connectors in OR cases, and F() objects and
aggregates in same filter clause caused GROUP BY problems on some
databases.
Fixed#17600, fixed#13198, fixed#17025, fixed#17000, fixed#11293.
Before there was need to have both .relabel_aliases() and .clone() for
many structs. Now there is only relabeled_clone() for those structs
where alias is the only mutable attribute.
Replaced them with per-database options, for proper multi-db support.
Also toned down the recommendation to tie transactions to HTTP requests.
Thanks Jeremy for sharing his experience.
Since "unless managed" now means "if database-level autocommit",
committing or rolling back doesn't have any effect.
Restored transactional integrity in a few places that relied on
automatically-started transactions with a transitory API.
For users who didn't activate autocommit in their database options, this
is backwards-incompatible in "non-managed" aka "auto" transaction state.
This state now uses database-level autocommit instead of ORM-level
autocommit.
Also removed the uses_autocommit feature which lost its purpose.
Autocommit cannot be manipulated independently from an open connection.
This commit introduces a minor change in behavior: entering transaction
management forces opening a databasse connection. This shouldn't be
backwards incompatible in any practical use case.
enter_transaction_management() was nearly always followed by managed().
In three places it wasn't, but they will all be refactored eventually.
The "forced" keyword argument avoids introducing behavior changes until
then.
This is mostly backwards-compatible, except, of course, for managed
itself. There's a minor difference in _enter_transaction_management:
the top self.transaction_state now contains the new 'managed' state
rather than the previous one. Django doesn't access
self.transaction_state in _enter_transaction_management.
There were a couple of errors in ._dirty flag handling:
* It started as None, but was never reset to None.
* The _dirty flag was sometimes used to indicate if the connection
was inside transaction management, but this was not done
consistently. This also meant the flag had three separate values.
* The None value had a special meaning, causing for example inability
to commit() on new connection unless enter/leave tx management was
done.
* The _dirty was tracking "connection in transaction" state, but only
in managed transactions.
* Some tests never reset the transaction state of the used connection.
* And some additional less important changes.
This commit has some potential for regressions, but as the above list
shows, the current situation isn't perfect either.
