Dataloader and N+1 Issue

I have seen various snippets and various solutions to using a dataloader to limit queries to the db but am struggling to piece together a example that works as I expect.

Does anyone have a concrete example or project that implements prefetch_related, select_related, dataloaders etc? I would also like to use the FilterSet class in my resolver that implements a dataloader. How can I do this?

This is what I have implemented so far:

class YearDataLoader(DataLoader):
    def batch_load_fn(self, keys):
        return Promise.resolve(YearDataModel.objects.filter(pk__in=keys))

class CompanyNode(DjangoObjectType):

    yeardata = DjangoFilterConnectionField(YearDataNode, filterset_class=YearDataFilter)

    class Meta:
        model = CompanyModel
        interfaces = (relay.Node, )

    def resolve_yeardata(self, info, *args, **kwargs):
        return year_data_loader.load_many(YearDataModel.objects.filter(company=self).values_list('id', flat=True))

If I can figure this out then I would gladly write an optimisation guide, as I feel like this is not covered at all in the current docs. All the pieces are there but someone needs to put them all together…

[filipeximenes]

Here is how I got it working:

def get_sorter(ids):
    def order_by(items):
        items_dict = {i.id: i for i in items}
        return [items_dict[i] for i in ids]
    return order_by

class CompanyNode(DjangoObjectType):

    yeardata = DjangoConnectionField(
        YearDataNode, 
        filter1=graphene.String(),
        filter2=graphene.DateTime(),
    )

    class Meta:
        model = CompanyModel
        interfaces = (relay.Node, )

    def resolve_yeardata(self, info, *args, **kwargs):
        ids = YearDataFilter(
            kwargs, YearDataModel.objects.all()).qs.values_list('id', flat=True)
        return (year_data_loader
                .load_many(ids)
                .then(get_sorter(id)))

Not perfect nor the most optimized, but should be enough for a temporary solution

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.