Current limitations — google-cloud-spanner-django documentation (original) (raw)

AutoField generates random IDs

Spanner doesn’t have support for auto-generating primary key values. Therefore, django-google-spanner monkey-patches AutoField to generate a random UUID4. It generates a default using Field’s default option which means AutoFields will have a value when a model instance is created. For example:

ExampleModel() ExampleModel.pk 4229421414948291880

To avoidhotspotting, these IDs are not monotonically increasing. This means that sorting models by ID isn’t guaranteed to return them in the order in which they were created.

ForeignKey constraints aren’t created (#313)

Spanner does not support ON DELETE CASCADE when creating foreign-key constraints, so this is not supported in django-google-spanner.

No native support for DecimalField

Spanner’s support for Decimaltypes is limited toNUMERICprecision. Higher-precision values can be stored as strings instead.

Meta.order_with_respect_to model option isn’t supported

This feature uses a column name that starts with an underscore (_order) which Spanner doesn’t allow.

Random QuerySet ordering isn’t supported

Spanner does not support it and will throw an exception. For example:

ExampleModel.objects.order_by('?') ... django.db.utils.ProgrammingError: 400 Function not found: RANDOM ... FROM example_model ORDER BY RANDOM() ASC

Schema migrations

There are some limitations on schema changes to consider:

DurationField arithmetic doesn’t work with DateField values (#253)

Spanner requires using different functions for arithmetic depending on the column type:

Django does not provide ways to determine which database function to use. DatabaseOperations.combine_duration_expression() arbitrarily usesTIMESTAMP_ADD and TIMESTAMP_SUB. Therefore, if you use aDateField in a DurationField expression, you’ll likely see an error such as:

"No matching signature for function TIMESTAMP_ADD for argument types: DATE, INTERVAL INT64 DATE_TIME_PART."

Computations that yield FLOAT64 values cannot be assigned to INT64 columns

Spanner does not support this (#331) and will throw an error:

ExampleModel.objects.update(integer=F('integer') / 2) ... django.db.utils.ProgrammingError: 400 Value of type FLOAT64 cannot be assigned to integer, which has type INT64 [at 1:46]\nUPDATE example_model SET integer = (example_model.integer /...

Addition with null values crash

Additions cannot include None values. For example:

Book.objects.annotate(adjusted_rating=F('rating') + None) ... google.api_core.exceptions.InvalidArgument: 400 Operands of + cannot be literal NULL ...

stddev() and variance() function call with sample population only

Spanner supports stddev() and variance() functions (link).

Django’s Variance and StdDev database functions have 2 modes. One with full population STDDEV_POP and another with sample population STDDEV_SAMP and VAR_SAMP. Currently spanner only supports these functions with samples and not the full population STDDEV_POP,

Interleaving is not supported currently

Interleaving is a feature that is supported by spanner database link. But currently django spanner does not support this feature, more details on this is discussed in this github issue.

Update object by passing primary key

In django3.1 a new feature was introduced, <instance>._state.adding, this allowed spanner to resolve this bug.

But introduced a new issue with spanner django. Calling instance.save() an object after setting it’s primary key to an existing primary key value, will cause a IntegrityError as follows: django.db.utils.IntegrityError: (1062, "Duplicate entry ....

The workaround for this is to update <instance>._state.adding to False. Example: .. code:: python

This test case passes.

def test_update_primary_with_default(self): obj = PrimaryKeyWithDefault() obj.save() obj_2 = PrimaryKeyWithDefault(uuid=obj.uuid) obj_2._state.adding = False obj_2.save()

This test case fails with IntegrityError.

def test_update_primary_with_default(self): obj = PrimaryKeyWithDefault() obj.save() obj_2 = PrimaryKeyWithDefault(uuid=obj.uuid) obj_2.save()

More details about this issue can be tracked here.

Support for query inside JSONfield is currently not there

We have also added support for JSON object storage and retrieval with Django 3.2.x support in v2.2.1b4 release, but querying inside the JSONfield is not supported in the current django-google-spanner release. This feature is being worked on and can be tracked here.