If the only manager on the model is the default manager defined
by Django (`objects = models.Manager()`), this manager will not
be added to the model state. If it is custom, it needs to be
passed to the model state.
The idea behind this change is, that AlterUniqueTogether,
AlterIndexTogether and AlterOrderWithRespectTo can always be moved after
an Add/Alter/Rename/RemoveField operation if they don't refer to the
respective field and are not empty sets / None.
Combined with the optimizations of duplicate AlterUniqueTogether,
AlterIndexTogether, and AlterOrderWithRespectTo operations from
128caa1e16, these operations are optimized
in a later round of the optimizer.
Thanks Tim Graham for the review.
This is a regression caused by introducing rendered migration states in
1aa3e09c20 and the _meta refactoring in fb48eb0581.
Thanks to Danilo Bargen for reporting the issue and Marten Kenbeek and
Tim Graham for triaging the bug and providing the initial test case.
Currently some of the migrations tests rely on the fact 'input' is aliased
because of six, instead of using mock.patch. Replace this code with proper
use of mock.patch.
Also, replace one case of excessively specific python version check with
testing six.PY3
This also prevents state modifications from corrupting previous states.
Previously, when a model defining a relation was unregistered first,
clearing the cache would cause its related models' _meta to be cleared
and would result in the old models losing track of their relations.
Made MigrationGraph forwards_plan() and backwards_plan() fall back to an
iterative approach in case the recursive approach exceeds the recursion
depth limit.
Field.rel is now deprecated. Rel objects have now also remote_field
attribute. This means that self == self.remote_field.remote_field.
In addition, made the Rel objects a bit more like Field objects. Still,
marked ManyToManyFields as null=True.
If Field.choices is provided as an iterator, consume it in __init__ instead
of using itertools.tee (which ends up holding everything in memory
anyway). Fixes a bug where deconstruct() was consuming the iterator but
bypassing the call to `tee`.
This is preferrable to writing in the current working directory because
it eliminates the risk to leak unwanted files, which can result in very
weird test failures.
Also this will help if we ever try to run these tests concurrently.
Switched from an adjancency list and uncached, iterative depth-first
search to a Node-based design with direct parent/child links and a
cached, recursive depth-first search. With this change, calculating
a migration plan for a large graph takes several seconds instead of
several hours.
Marked test `migrations.test_graph.GraphTests.test_dfs` as an expected
failure due to reaching the maximum recursion depth.
The new signature enables better support for routing RunPython and
RunSQL operations, especially w.r.t. reusable and third-party apps.
This commit also takes advantage of the deprecation cycle for the old
signature to remove the backward incompatibility introduced in #22583;
RunPython and RunSQL won't call allow_migrate() when when the router
has the old signature.
Thanks Aymeric Augustin and Tim Graham for helping shape up the patch.
Refs 22583.
Swapped out models don't have a _default_manager unless they have
explicitly defined managers. ModelState.from_model() now accounts for
this case and uses an empty list for managers if no explicit managers
are defined and a model is swapped out.
