Serializer `resource_name` for nested objects is not used

[mkohram]

Nested serializers don't use resource_name for the type. In the example below the type for the child would show something like nestedmodels.

class NestedSerializer(serializers.ModelSerializer):
    resource_name = "pretty_nested"
    class Meta:
        model = NestedModel

class ParentSerializer(serializers.ModelSerializer):
    nested = NestedSerializer(source='some_field', many=True)
    class Meta:
        model = ParentModel

Looking at the code it looks like in utils.py the only function that performs the correct checks to get the type is get_resource_name(). All the other functions simply do something like relation_type = inflection.pluralize(relation_model.__name__).lower() without checking for a resource_name on the serializer as the docs suggest. I could submit a PR if it makes sense to put these checks in place everywhere. Ideally a function like get_related_resource_type() should act as a getter for type so that we don't see these inconsistencies.

[topaxi]

We just stumbled upon the relation_type too, we use Ember.js for our frontend and create models based on the type property.

It would be very beneficial to us, if we could configure how the relation_type is formatted, very similar to JSON_API_FORMAT_KEYS. Just lowercasing the model name seems to remove information on the actual model name, a dasherized, underscored or camelcased type would be much better.

Otherwise we're forced to use "single worded" models.

[jsenecal]

Thanks for the suggestion @topaxi - I'm quite swamped however lately so unless @jerel has some time to spare, we are open to PRs ;)

[jerel]

This should be fixed by #76 when it is merged.

[mkohram]

#76 does centralize the formatting but still relation_model.__name__ is being used everywhere disregarding resource_name on serializers. The docs say

You may manually set the resource_name property on views or serializers to specify the type key in the json output.

For example:
https://github.com/django-json-api/django-rest-framework-json-api/blob/3498201af79dad06c497ffbc3f128985c21cade7/rest_framework_json_api/utils.py#l298

The serializer resource_name is not looked for when creating the type string.

[jerel]

Thanks for the comment on relation_model.__name__. I'll close this and we'll add resource_name capabilities on the serializer as part of fixing #77

[mkohram]

Sweet. Let me know if I can help!

[jsenecal]

Since this issue is not fixed by 64f8a6b I'll reopen this as a reminder that the reported problem still exists.

[jsenecal]

@mkohram could you please test code from branch feature/compound_documents to see if it fixes this issue?

[mkohram]

I'm afraid not. Looking through the code I believe extract_relationships is one of the culprits. For example here:

django-rest-framework-json-api/rest_framework_json_api/utils.py

Line 381 in 66c1eef

('type', nested_resource_instance_type),

...
        if isinstance(field, ListSerializer):
            relation_data = list()

            serializer_data = resource.get(field_name)
            resource_instance_queryset = list(relation_instance_or_manager.all())
            if isinstance(serializer_data, list):
                for position in range(len(serializer_data)):
                    nested_resource_instance = resource_instance_queryset[position]
                    # something along the lines of 
                    # nested_resource_instance_type = field.child.Meta.resource_name
                    nested_resource_instance_type = get_resource_type_from_instance(nested_resource_instance)
                    relation_data.append(OrderedDict([
                        ('type', nested_resource_instance_type),
                        ('id', encoding.force_text(nested_resource_instance.pk))
                    ]))

                data.update({field_name: {'data': relation_data}})
                continue

A check on field.child.Meta.resource_name is needed before resorting to getting the type from the queryset.

Other isinstance() statements in that sequence need similar resource_name checks.

Also my relationships don't make it into the included on feature/compound_documents for some reason.

[mkohram]

actually looking closer, looks like the extract_included() function already does this!

            serializer = field.child
            relation_type = get_resource_type_from_serializer(serializer)

We just need that up in extract_relationships() as well!