JSONRenderer does not work when no pk is provided

[mdettelson]

Hi there!

I have a particular use case here: I have a denormalized table that contains one data type (let's call it 'pieces') that gets aggregated up to another model (let's call them 'pies').

Since I can't do anything with pies without having all of the pieces (technical requirement), I've decided to keep a table representing only the pieces and aggregate up to the pie level when pies are asked for.

I noticed this line in the renderer code:

('id', encoding.force_text(resource_instance.pk) if resource_instance else None),
It's in the method build_json_resource_obj

Since my dataset has many repeats of the pie IDs, I can't use it as a pk, BUT once I aggregate the table the Pie Viewset, it IS unique. Since it is not the actual pk on the model, however, I get the error 'dict' object has no attribute 'pk'.

The ideal situation would be that I can set the pie ID as the id in the JSON API representiation.

Is there a built-in way to handle this sort of use case?

[sliverc]

I am not 100% sure about your use case. But when using a ModelSerializer this is currently not possible: look at discussion #155 But you could also use a python model and overwrite pk property see discussion at #704

Closing this as duplicate. If you think it is not a duplicate please comment or otherwise continue the discussion in the other issue if needed. Thanks.

[mdettelson]

Ah ok, good to know! Thanks for the references.