Thanks to Jeremy Dunck for pointing out the problem with this change. If in a
single transaction, the master deletes a record and then get_or_creates a
similar record, under the new behavior the get_or_create would find the record
in the slave db and fail to re-create it, leaving the record nonexistent, which
violates the contract of get_or_create that the record should always exist
afterwards. We need to do everything against the master here in order to ensure
correctness.
This reverts commit 901af86550.
When doing deeper than one level select_related() + only queries(), the
code introduced in b6c356b7bb errored
incorrectly.
Thanks to mrmachine for report & test case.
In an ideal world, nothing except django.db.models.query should have to
import stuff from django.models.sql.*. A few things were needing to get
hold of sql.constants.LOOKUP_SEP, so this commit moves it up to
django.db.models.constants.LOOKUP_SEP.
There are still a couple of places (admin) poking into sql.* to get
QUERY_TERMS, which is unfortunate, but a slightly different issue and
harder to adjust.
Previous this used Python's builtin hash() function, which has never been guarnteed to be stable across implementations (CPython/Jython/etc.) or 32/64 bitness. However, this in practice it was stable. However, with the impending release of Python 3.3 hash randomizations is enabled by default, which would mean the index name changed between program invocations.
Unlike other model fields, the newly introduced (1.4)
GenericIPAddressField did not accept verbose_name and name as the
first positional arguments. This commit fixes it.
Thanks Dan McGee for the report and the patch.
The joins for nested nullable foreign keys were often created as INNER
when they should have been OUTER joins. The reason was that only the
first join in the chain was promoted correctly. There were also issues
with select_related etc.
The basic structure for this problem was:
A -[nullable]-> B -[nonnull]-> C
And the basic problem was that the A->B join was correctly LOUTER,
the B->C join not.
The major change taken in this patch is that now if we promote a join
A->B, we will automatically promote joins B->X for all X in the query.
Also, we now make sure there aren't ever join chains like:
a LOUTER b INNER c
If the a -> b needs to be LOUTER, then the INNER at the end of the
chain will cancel the LOUTER join and we have a broken query.
Sebastian reported this problem and did also major portions of the
patch.
The ORM generated a query with INNER JOIN instead of LEFT OUTER JOIN
in a somewhat complicated case. The main issue was that there was a
chain of nullable FK -> non-nullble FK, and the join promotion logic
didn't see the need to promote the non-nullable FK even if the
previous nullable FK was already promoted to LOUTER JOIN. This resulted
in a query like a LOUTER b INNER c, which incorrectly prunes results.
Deferred models now automatically update only the fields which are
loaded from the db (with .only() or .defer()). In addition, any field
set manually after the load is updated on save.
This patch removes an unconditional float(value) conversion from db
backend default convert_values() method. This can cause problems when
aggregating over character fields for example. In addition, Oracle
and SQLite already return the bare value from their convert_values().
In the long term the converting should be done by fields, and the
fields should then call database backend specific converters when
needed. The current setup is inflexible for 3rd party fields.
Thanks to Merlijn van Deen for the original patch.
* Renamed the __unicode__ methods
* Applied the python_2_unicode_compatible decorator
* Removed the StrAndUnicode mix-in that is superseded by
python_2_unicode_compatible
* Kept the __unicode__ methods in classes that specifically
test it under Python 2
* Renamed smart_unicode to smart_text (but kept the old name under
Python 2 for backwards compatibility).
* Renamed smart_str to smart_bytes.
* Re-introduced smart_str as an alias for smart_text under Python 3
and smart_bytes under Python 2 (which is backwards compatible).
Thus smart_str always returns a str objects.
* Used the new smart_str in a few places where both Python 2 and 3
want a str.
Previously, the flush was done before the test case execution and now
it is performed after it.
Other changes to the testing infrastructure include:
* TransactionTestCase now doesn't reset autoincrement sequences either
(previous behavior can achieved by using `reset_sequences`.)
With this, no implicit such reset is performed by any of the provided
TestCase classes.
* New ordering of test cases: All unittest tes cases are run first and
doctests are run at the end.
THse changes could be backward-incompatible with test cases that relied
on some kind of state being preserved between tests. Please read the
relevant sections of the release notes and testing documentation for
further details.
Thanks Andreas Pelme for the initial patch. Karen Tracey and Anssi
Kääriäinen for the feedback and Anssi for reviewing.
This also fixes#12408.
The qs.bulk_create() method did not work with large batches together
with SQLite3. This commit adds a way to split the bulk into smaller
batches. The default batch size is unlimited except for SQLite3 where
the batch size is limited to 999 SQL parameters per batch.
Thanks to everybody who participated in the discussions at Trac.
Cleared aggregations on add_date_select method so only distinct dates
are returned when dealing with a QuerySet that contained aggregations.
That would cause the query set to return repeated dates because it
would look for distinct (date kind, aggregation) pairs.
At least Oracle needs parentheses in negated where conditions, even if
there is only single condition negated. Fixed this by reverting to old
logic in that part of as_sql() and adding a comment about this.
I did not investigate why the parentheses are needed. The original
offending commit was bd283aa844.
When the postgresql_psycopg2 backend was used with DB-level autocommit
mode enabled, after entering transaction management and then leaving
it, the isolation level was never set back to autocommit mode.
Thanks brodie for report and working on this issue.
Made sure the WhereNode.as_sql() handles various EmptyResultSet and
FullResultSet conditions correctly. Also, got rid of the FullResultSet
exception class. It is now represented by '', [] return value in the
as_sql() methods.
This commit tackles a couple of issues. First, in certain cases there
were some mixups if field.attname or field.name should be deferred.
Field.attname is now always used.
Another issue tackled is a case where field is both deferred by
.only(), and selected by select_related. This case is now an error.
A lot of thanks to koniiiik (Michal Petrucha) for the patch, and
to Andrei Antoukh for review.
In addition, removed a possibly problematic .filter() call from
backends.test_query_encoding test. It is possible the .filter could
cause collation problems on MySQL, and as it wasn't absolutely needed
for the test it seemed better to get rid of the call.
Refs #18461.
This does remove the requirement to call features.confirm() method
before checking the properties.
Thanks cdestiger and Ramiro Morales for their work on the patch.
On MySQL when checking the server version, a new connection could be
created but never closed. This could result in open connections on
server startup.
Generic cleanup and dead code removal in deferred model field loading
and model.__reduce__().
Also fixed an issue where if an inherited model with a parent field
chain parent_ptr_id -> id would be deferred loaded, then accessing
the id field caused caused a database query, even if the id field's
value is already loaded in the parent_ptr_id field.
When order_by causes new joins to be added to the query, the joins must
be LEFT OUTER joins for nullable relations, otherwise the order_by
could cause the results to be altered. This commit fixes the logic to
only promote new joins, previously all joins in the order_by lookup
path were promoted.
Thanks to Bruno Desthuilliers for spotting this corner case.
This was recently fixed for one-to-one relations; this patch adds
support for foreign keys. Thanks kaiser.yann for the report and
the initial version of the patch.
Databases with update_can_self_select = False (MySQL for example)
generated non-necessary queries when saving a multitable inherited
model, and when the save resulted in update.
after the full file name is generated by the storage class.
Thanks Refefer for the report, carsongee for the patch, and
everyone else involved in the discussion.
Added the ability to update only part of the model's fields in
model.save() by introducing a new kwarg "update_fields". Thanks
to all the numerous reviewers and commenters in the ticket
Fixed#18248 -- proxy models were added to included_inherited_models
in sql.query.Query. The variable is meant to be used for multitable
inheritance only. This mistake caused problems in situations where
proxy model's query was reused.
Fixed#15933, #18082 -- the get_indexes() method introspection was
done inconsitently depending on the backend. For example SQLite
included all the columns in the table in the returned dictionary,
while MySQL introspected also multicolumn indexes.
All backends return now consistenly only single-column indexes.
Thanks to andi for the MySQL report, and ikelly for comments on
Oracle's get_indexes() changes.
Fixed#17957 -- when using Oracle and character fields, the fields
were set null = True to ease the handling of empty strings. This
caused problems when using multiple databases from different vendors,
or when the character field happened to be also a primary key.
The handling was changed so that NOT NULL is not emitted on Oracle
even if field.null = False, and field.null is not touched otherwise.
Thanks to bhuztez for the report, ramiro for triaging & comments,
ikelly for the patch and alex for reviewing.
QuerySet had previously some complex logic for dealing with nullable
fields in negated add_filter() calls. It seems the logic is leftover
from a time where the WhereNode wasn't as intelligent in handling
field__in=[] conditions.
Thanks to aaugustin for comments on the patch.
Fixed#18218 -- previously Django's introspection table_names() and
get_table_list() methods did not sort the output consistently. This
resulted in random order of inspected models.
This commit also removed all external usages of get_table_list().
table_names() should be used instead.
Thanks to claudep for patch and report.
Carl Meyer
Anssi Kääriäinen
Claude Paroz
Malcolm Tredinnick
Alex Gaynor
Mitar
Alex Gaynor
Aymeric Augustin
Simon Meers
Dmitry Shevchenko
Andrei Antoukh
Brendan MacDonell
Ramiro Morales
Nuno Maltez
Josh Smeaton
Luke Plant
Jann Kleen
Michael Newman
Simon Charette
Adrian Holovaty
Julien Phalip
Jannis Leidel
Karen Tracey
Paul McMillan