DjangoFilterConnectionField with DataLoader

Consider a query like this:

{
    category(id: 1) {
        products {
            edges {
                node {
                    category {
                        products {
                            edges {
                                node {
                                    category {
                                        products {
                                            edges {
                                                node {
                                                    id
                                                }
                                            }
                                        }
                                    }
                                }
                            }
                        }
                    }
                }
            }
        }
    }
}

If someone runs a query like this we are possibly executing hundreds of queries even if the response contains a fairly small number of unique objects. We can use a DataLoader for the product.category relation but as category.products is a DjangoFilterConnectionField I see no way to have it memorize results within the same execution context.

Ideally I’d want it to only query the database once for each unique combination of category.id and input parameters (pagination, filterset data). Currently it will fetch the category, then fetch all its products, then for each result once again fetch all products in the category and then once again fetch all products in the category for each of those.

[ulgens]

I think the issue is still relevant. Opening it back.

[AndrewIngram]

I’ve hit this same requirement myself, so i’ll see if I can port over my method from Node. The “perfect” solution is variant of Dataloader that’s aware of the semantics of connections.