使用Dataloader时出现Cannot return null for non-nullable字段错误

问题背景 #

已创建Episode、Character和Director模型，编写了对应的types、queries及mutations，未使用Dataloader前一切正常，返回QuerySet时运行正常。但尝试用Dataloader提升性能，在resolver中返回Promise对象后，执行指定GraphQL查询出现错误：Cannot return null for non-nullable field EpisodeTypeConnection.edges.

文件夹结构 #

models.py #

import graphene
from graphene import ObjectType, relay
from graphene_django import DjangoObjectType

from general.models import Character, Director, Episode


class CharacterType(DjangoObjectType):
    pk = graphene.Int(source="pk", required=True)

    class Meta:
        model = Character
        filter_fields = {
            "name": ["exact", "icontains", "istartswith"],
            "character_species": ["exact", "icontains"],
            "location": ["exact"],
            "status": ["exact"],
        }
        fields = "__all__"
        interfaces = (relay.Node,)


class EpisodeType(DjangoObjectType):
    pk = graphene.Int(source="pk")

    class Meta:
        model = Episode
        filter_fields = {
            "name": ["exact", "icontains", "istartswith"],
            "directed_by__name": ["exact", "icontains"],
            "aired_date": ["exact"],
        }
        fields = "__all__"
        interfaces = (relay.Node,)


class DirectorType(DjangoObjectType):
    pk = graphene.Int(source="pk", required=True)

    class Meta:
        model = Director
        filter_fields = {
            "name": ["exact", "icontains", "istartswith"],
            "first_directed_episode__name": ["exact", "icontains"],
            "last_directed_episode__name": ["exact", "icontains"],
            "age": ["exact"],
        }
        fields = "__all__"
        interfaces = (relay.Node,)

types.py #

import graphene
from graphene import ObjectType, relay
from graphene_django import DjangoObjectType

from general.models import Character, Director, Episode


class CharacterType(DjangoObjectType):
    pk = graphene.Int(source="pk", required=True)

    class Meta:
        model = Character
        filter_fields = {
            "name": ["exact", "icontains", "istartswith"],
            "character_species": ["exact", "icontains"],
            "location": ["exact"],
            "status": ["exact"],
        }
        fields = "__all__"
        interfaces = (relay.Node,)


class EpisodeType(DjangoObjectType):
    pk = graphene.Int(source="pk")

    class Meta:
        model = Episode
        filter_fields = {
            "name": ["exact", "icontains", "istartswith"],
            "directed_by__name": ["exact", "icontains"],
            "aired_date": ["exact"],
        }
        fields = "__all__"
        interfaces = (relay.Node,)


class DirectorType(DjangoObjectType):
    pk = graphene.Int(source="pk", required=True)

    class Meta:
        model = Director
        filter_fields = {
            "name": ["exact", "icontains", "istartswith"],
            "first_directed_episode__name": ["exact", "icontains"],
            "last_directed_episode__name": ["exact", "icontains"],
            "age": ["exact"],
        }
        fields = "__all__"
        interfaces = (relay.Node,)

loaders.py #

from collections import defaultdict
from promise import Promise
from promise.dataloader import DataLoader
from general.models import Episode


class EpisodesByDirectorIdLoader(DataLoader):
    def batch_load_fn(self, director_ids):
        episodes_by_director_ids = defaultdict(list)
        all_episodes_of_directors = Episode.objects.filter(
            directed_by__pk__in=director_ids
        )
        for episode in all_episodes_of_directors.iterator():
            episodes_by_director_ids[episode.directed_by.pk].append(episode)

        return Promise.resolve(
            [
                episodes_by_director_ids.get(director_id, [])
                for director_id in director_ids
            ]
        )

queries.py #

import graphene
from graphene_django.filter import DjangoFilterConnectionField

from general.models import Character, Director, Episode

from .types import EpisodeType


class Query(graphene.ObjectType):
    all_episodes = DjangoFilterConnectionField(EpisodeType)

    @staticmethod
    def resolve_all_episodes(root, info):
        # return Episode.objects.all() # RETURNS QUERYSET
        return info.context.episodes_by_director_id_loader.load(1) # RETURNS PROMISE OBJECT

执行的GraphQL查询 #

query{
  allEpisodes{
    edges{
      node{
        name
      }
    }
  }
}

错误输出 #

{
  "errors": [
    {
      "message": "Cannot return null for non-nullable field EpisodeTypeConnection.edges.",
      "locations": [
        {
          "line": 26,
          "column": 5
        }
      ],
      "path": [
        "allEpisodes",
        "edges"
      ]
    }
  ],
  "data": {
    "allEpisodes": null
  }
}

问题原因及解决方法 #

核心原因 #

你用DjangoFilterConnectionField定义的all_episodes字段，期望返回Connection类型（包含edges、pageInfo等结构），但resolver返回的是Dataloader加载的Episode对象列表，类型不匹配导致GraphQL无法解析成Connection结构，最终返回null触发非空字段错误。

另外，Dataloader的核心作用是解决N+1查询问题，比如批量加载关联数据，而非直接替换顶层QuerySet的返回。

解决步骤 #

1. 修正Dataloader的使用场景（推荐用法）

如果目标是优化关联查询（比如查询Director时加载其执导的Episodes），应在DirectorType中添加自定义字段，用Dataloader批量加载：

修改types.py中的DirectorType：

class DirectorType(DjangoObjectType):
    pk = graphene.Int(source="pk", required=True)
    # 添加自定义字段，返回该导演的所有剧集
    directed_episodes = graphene.List(EpisodeType)

    class Meta:
        model = Director
        filter_fields = {
            "name": ["exact", "icontains", "istartswith"],
            "first_directed_episode__name": ["exact", "icontains"],
            "last_directed_episode__name": ["exact", "icontains"],
            "age": ["exact"],
        }
        fields = "__all__"
        interfaces = (relay.Node,)
    
    # 为directed_episodes字段添加resolver，使用Dataloader
    def resolve_directed_episodes(self, info):
        return info.context.episodes_by_director_id_loader.load(self.pk)

恢复queries.py中的resolve_all_episodes为原QuerySet返回：

@staticmethod
def resolve_all_episodes(root, info):
    return Episode.objects.all()

这样查询Director及其执导的剧集时，Dataloader会批量加载所有关联Episodes，避免N+1查询，同时保持顶层查询的Connection结构正常工作。

2. 确认Dataloader已注册到context

确保已在GraphQL视图中将Dataloader实例添加到context，比如自定义视图：

from django.views.decorators.csrf import csrf_exempt
from graphene_django.views import GraphQLView
from .loaders import EpisodesByDirectorIdLoader

class CustomGraphQLView(GraphQLView):
    def get_context(self, request):
        context = super().get_context(request)
        # 注册Dataloader到context
        context.episodes_by_director_id_loader = EpisodesByDirectorIdLoader()
        return context

在urls.py中使用该自定义视图：

urlpatterns = [
    path("graphql/", csrf_exempt(CustomGraphQLView.as_view(graphiql=True))),
]

内容的提问来源于stack exchange，提问作者HarilalAV