Annotating with the count of a subquery filtering by jsonb fields in Django

Question

I have a single model with a jsonb field. There is a value inside this jsonb field that can be shared amongst other rows. I am trying to get the count of a subquery while filtering by this jsonb field.

Some pseudo code of what I have been attempting borrowing examples from this post.

alpha_qs = MyModel.objects.filter(class_id="ALPHA")
# unnest jsonb field so I can leverage it via OuterRef
alpha_qs = alpha_qs.annotate(nested_value_id=KeyTextTransform("nested_value_id", "a_jsonb_field"))
related_beta_subquery = MyModel.objects.filter(class_id="BETA", a_jsonb_field__nested_value_id=OuterRef("nested_value_id"))
related_beta_subquery_count = related_beta_subquery.annotate(count=Count("*")).values("count")
alpha_qs = alpha_qs.annotate(related_beta_count=Subquery(related_beta_subquery))

Using this example data I would expect the top instance to have a related_beta_count of 2 because there are two associated betas with the same nested_value_id.

{
  "class_id": "ALPHA",
  "a_jsonb_field": {
    "nested_value_id": 'abc'
  }
}

{
  "class_id": "BETA",
  "a_jsonb_field": {
    "nested_value_id": 'abc'
  }
}
{
  "class_id": "BETA",
  "a_jsonb_field": {
    "nested_value_id": 'abc'
  }
}
{
  "class_id": "BETA",
  "a_jsonb_field": {
    "nested_value_id": 'zyz'
  }
}

I've been getting an error below but haven't been able to resolve it.

ProgrammingError: operator does not exist: jsonb = text
LINE 1: ...d AND (U0."a_jsonb_field" -> 'nested_value_id') = ("my_model...
                                                             ^
HINT:  No operator matches the given name and argument type(s). You might need to add explicit type casts.

Have you tried this: `alpha_qs.annotate(nested_value_id=KeyTextTransform("a_jsonb_field__nested_value_id", "your_field_name"))` — NKSM, Apr 09 '21 at 22:59
@NKSM I'm not sure what "your_field_name" would be in your example. But I can confirm what I have on that line: `alpha_qs = alpha_qs.annotate(nested_value_id=KeyTextTransform("nested_value_id", "a_jsonb_field"))` does work. As I am able to do `alpha_qs[0].nested_value_id` and get the value I expect. — anthony-dandrea, Apr 10 '21 at 01:57

score 0 · Accepted Answer · answered Apr 14 '21 at 02:07

I got it working my use case. I don't consider myself well versed in Django so there may be a cleaner way to achieve this.

# Get initial queryset
alpha_qs = MyModel.objects.filter(class_id="ALPHA")
# Pull nested jsonb value out via annotation/KeyTextTransform
alpha_qs = alpha_qs.annotate(nested_value_id=KeyTextTransform("nested_value_id", "a_jsonb_field"))
related_beta_subquery = (
    MyModel.objects.filter(
        class_id="BETA"
    )
    # Pull out nested jsonb for subquery as well
    .annotate(
        nested_value_id=KeyTextTransform("nested_value_id", "a_jsonb_field")
    )
    # Filter to ensure these BETA models have a matched nested_value_id to my ALPHA models
    .filter(
        nested_value_id=OuterRef("nested_value_id")
    )
    # This was honestly the harder part that took a lot of trial and error
    .order_by()
    .values("nested_value_id")
    .annotate(count=Count("*"))
    .values("count")
)
alpha_qs = alpha_qs.annotate(
    count=Case(
        # This Case statement is mostly just bulletproofing. May be unnecessary for your use case
        When(a_jsonb_field__has_key="nested_value_id", then=Subquery(related_beta_subquery)),
        default=Value(0),
        output_field=IntegerField(),
    )
)

Annotating with the count of a subquery filtering by jsonb fields in Django

1 Answers1