The get_(local|foreign)_related_value methods of GenericRelation must be
reversed because it defines (from|to)_fields and associated related_fields
in the reversed order as it's effectively a reverse GenericForeignKey
itself.
The related value methods must also account for the fact that referenced
primary key values might be stored as a string on the model defining the
GenericForeignKey but as integer on the model defining the GenericRelation.
This is achieved by calling the to_python method of the involved content type
in get_foreign_related_value just like GenericRelatedObjectManager does.
Lastly reverse many-to-one manager's prefetch_related_querysets should use
set_cached_value instead of direct attribute assignment as direct assignment
might are disallowed on ReverseManyToOneDescriptor descriptors. This is likely
something that was missed in f5233dc (refs #32511) when the is_cached guard
was added.
Thanks 1xinghuan for the report.
Backport of 198b30168d4e94af42e0dc7967bd3259b5c5790b from main.
Over the years we've had multiple instances of hit and misses when
emitting warnings: either setting the wrong stacklevel or not setting
it at all.
This work adds assertions for the existing warnings that were declaring
the correct stacklevel, but were lacking tests for it.
This also caused un-ordered sliced prefetches to crash as they rely on Window.
Regression in e16d0c176e9b89628cdec5e58c418378c4a2436a that made OrderByList
piggy-back ExpressionList without porting the empty handling that the latter
provided.
Supporting explicit empty ordering on Window functions and slicing is arguably
a foot-gun design due to how backends will return undeterministic results but
this is a problem that requires a larger discussion.
Refs #35064.
Thanks Andrew Backer for the report and Mariusz for the review.
Changed the cache name used for singly related objects to be the
to_attr parameter passed to a Prefetch object. This fixes issues with
checking if values have already been fetched in cases where the Field
already has some prefetched value, but not for the same model attr.
This reverts 4f8c7fd9d91b35e2c2922de4bb50c8c8066cbbc6 and adds
two regression tests:
- test_related_manager_refresh(), and
- test_create_copy_with_m2m().
Thanks joeli for the report.
In these cases Black produces unexpected results, e.g.
def make_random_password(
self,
length=10,
allowed_chars='abcdefghjkmnpqrstuvwxyz' 'ABCDEFGHJKLMNPQRSTUVWXYZ' '23456789',
):
or
cursor.execute("""
SELECT ...
""",
[table name],
)
When prefetching a set of child objects related to a set of parent
objects, we usually want to populate the relationship back from the
child to the parent to avoid a query when accessing that relationship
attribute. However, there's an edge case where the child queryset
itself specifies a prefetch back to the parent. In that case, we want
to use the prefetched relationship rather than populating the reverse
relationship from the parent.
This also replaces assertQuerysetEqual() to
assertSequenceEqual()/assertCountEqual() where appropriate.
Co-authored-by: Peter Inglesby <peter.inglesby@gmail.com>
Co-authored-by: Mariusz Felisiak <felisiak.mariusz@gmail.com>
Added internal interface to QuerySet that allows to defer next filter
call till .query is accessed. Used it to optimize prefetch_related().
Thanks Simon Charette for the review.
Changed __eq__ to return NotImplemented instead of False if compared to
an object of the same type, as is recommended by the Python data model
reference. Now these models can be compared to ANY (or other objects
with __eq__ overwritten) without returning False automatically.
