The `QuerySet._clone()` method has no arguments in either Django 3.2
or 4.1, so I'm assuming the arguments were necessary for a version
of Django that is no longer supported by `django-model-utils`.
There are small typos in:
- docs/managers.rst
- model_utils/managers.py
- tests/test_models/test_timestamped_model.py
Fixes:
- Should read `subclasses` rather than `subclasess`.
- Should read `queryset` rather than `querset`.
- Should read `permissible` rather than `permissable`.
- Should read `modified` rather than `modifed`.
- Should read `heterogeneous` rather than `heterogenous`.
Signed-off-by: Tim Gates <tim.gates@iress.com>
When calling `select_related()` with an empty list of arguments [1], Django will
try to prefetch some data by doing some first level joints with the related
classes.
This can lead to obvious negative performance impact, but this also breaks some
workarounds for having inheritance for foreign keys [2], as those solutions rely
on lazy evaluation of the related object.
[1]: a4e6030904/django/db/models/query.py (L1051)
Only passing an explicit `None` to `select_related` will disable the magic.
[2]: https://github.com/jazzband/django-model-utils/issues/11
As examples, here are the generated SQL requests in
InheritanceManagerRelatedTests.test_get_method_with_select_subclasses_check_for_useless_join:
* without this fix, without adding `.select_related(None)`
```sql
SELECT
"tests_inheritancemanagertestparent"."id",
"tests_inheritancemanagertestparent"."non_related_field_using_descriptor",
"tests_inheritancemanagertestparent"."related_id",
"tests_inheritancemanagertestparent"."normal_field",
"tests_inheritancemanagertestparent"."related_self_id",
"tests_inheritancemanagertestchild4"."other_onetoone_id",
"tests_inheritancemanagertestchild4"."parent_ptr_id", T3."id",
T3."non_related_field_using_descriptor", T3."related_id", T3."normal_field",
T3."related_self_id"
FROM
"tests_inheritancemanagertestchild4"
INNER JOIN
"tests_inheritancemanagertestparent" ON
("tests_inheritancemanagertestchild4"."parent_ptr_id" = "tests_inheritancemanagertestparent"."id")
INNER JOIN
"tests_inheritancemanagertestparent" T3 ON
("tests_inheritancemanagertestchild4"."other_onetoone_id" = T3."id")
WHERE
"tests_inheritancemanagertestchild4"."parent_ptr_id" = 191
```
* with either the fix, or by adding `.select_related(None)` after `.select_subclasses()`
```sql
SELECT
"tests_inheritancemanagertestparent"."id",
"tests_inheritancemanagertestparent"."non_related_field_using_descriptor",
"tests_inheritancemanagertestparent"."related_id",
"tests_inheritancemanagertestparent"."normal_field",
"tests_inheritancemanagertestparent"."related_self_id",
"tests_inheritancemanagertestchild4"."other_onetoone_id",
"tests_inheritancemanagertestchild4"."parent_ptr_id"
FROM
"tests_inheritancemanagertestchild4"
INNER JOIN
"tests_inheritancemanagertestparent" ON
("tests_inheritancemanagertestchild4"."parent_ptr_id" = "tests_inheritancemanagertestparent"."id")
WHERE
"tests_inheritancemanagertestchild4"."parent_ptr_id" = 191
```
* Overriding TimeStampedModel's save() method
to update modified field automatically when update_fields argument is used but modified was not given as a parameter.
* fixed a small error in logic
* added tests for new functionality
* add test for reproducing FieldTracker update_fields issue
FieldTracker does not see update_fields changed in tracked model.save() method
* Patch Model.save_base instead of save
This change allows proper capturing update_fields kwarg if is is changed in overridden save() method.
* Add some details about FieldTracker implementation
* Mention FieldTracker behavior change
