Django REST Framework缓存bug

Question

TL; DR

我正在寻找一种方法来在请求后清除缓存,或者在运行测试时完全禁用它.Django REST Framework似乎缓存了结果,我需要解决这个问题.

长版和代码

好吧,当我继续测试它时,这表现得非常奇怪.最后,我得到了它的工作,但我真的不喜欢我的解决方法,并以知识的名义,我必须找出为什么会发生这种情况以及如何正确解决这个问题.

所以,我有一个像这样声明的APITestCase类:

class UserTests(APITestCase):

在这个类中,我有一个测试函数用于我的user-list视图,因为我有一个自定义查询集取决于权限.澄清事情:

• 超级用户可以获取整个用户列表(返回4个实例),
• 工作人员看不到超级用户(返回3个实例),
• 普通用户只能获得1个结果,他们自己的用户(返回1个实例)

有效的测试功能版本:

def test_user_querysets(self):

    url = reverse('user-list')

    # Creating a user
    user = User(username='user', password=self.password)
    user.set_password(self.password)
    user.save()

    # Creating a second user
    user2 = User(username='user2', password=self.password)
    user2.set_password(self.password)
    user2.save()

    # Creating a staff user
    staff_user = User(username='staff_user', password=self.password, is_staff=True)
    staff_user.set_password(self.password)
    staff_user.save()

    # Creating a superuser
    superuser = User(username='superuser', password=self.password, is_staff=True, is_superuser=True)
    superuser.set_password(self.password)
    superuser.save()



    # SUPERUSER

    self.client.logout()
    self.client.login(username=superuser.username, password=self.password)

    response = self.client.get(url)

    # HTTP_200_OK
    self.assertEqual(response.status_code, status.HTTP_200_OK)

    # All users contained in list
    self.assertEqual(response.data['extras']['total_results'], 4)



    # STAFF USER

    self.client.logout()
    self.client.login(username=staff_user.username, password=self.password)

    response = self.client.get(url)

    # HTTP_200_OK
    self.assertEqual(response.status_code, status.HTTP_200_OK)

    # Superuser cannot be contained in list
    self.assertEqual(response.data['extras']['total_results'], 3)



    # REGULAR USER

    self.client.logout()
    self.client.login(username=user2.username, password=self.password)

    response = self.client.get(url)

    # HTTP_200_OK
    self.assertEqual(response.status_code, status.HTTP_200_OK)

    # Only 1 user can be returned
    self.assertEqual(response.data['extras']['total_results'], 1)

    # User returned is current user
    self.assertEqual(response.data['users'][0]['username'], user2.username)

如您所见,我按此顺序测试用户权限:超级用户,员工,普通用户.而这很有效,所以......

有趣的事:

如果我更改测试的顺序,并从普通用户,员工,超级用户开始,测试将失败.来自第一个请求的响应被缓存,然后当我以staff用户身份登录时,我得到相同的响应,因此结果数量再次为1.

不起作用的版本:

它与以前完全相同,只是以相反的顺序进行测试

def test_user_querysets(self):

    url = reverse('user-list')

    # Creating a user
    user = User(username='user', password=self.password)
    user.set_password(self.password)
    user.save()

    # Creating a second user
    user2 = User(username='user2', password=self.password)
    user2.set_password(self.password)
    user2.save()

    # Creating a staff user
    staff_user = User(username='staff_user', password=self.password, is_staff=True)
    staff_user.set_password(self.password)
    staff_user.save()

    # Creating a superuser
    superuser = User(username='superuser', password=self.password, is_staff=True, is_superuser=True)
    superuser.set_password(self.password)
    superuser.save()



    # REGULAR USER

    self.client.logout()
    self.client.login(username=user2.username, password=self.password)

    response = self.client.get(url)

    # HTTP_200_OK
    self.assertEqual(response.status_code, status.HTTP_200_OK)

    # Only 1 user can be returned
    self.assertEqual(response.data['extras']['total_results'], 1)

    # User returned is current user
    self.assertEqual(response.data['users'][0]['username'], user2.username)



    # STAFF USER

    self.client.logout()
    self.client.login(username=staff_user.username, password=self.password)

    response = self.client.get(url)

    # HTTP_200_OK
    self.assertEqual(response.status_code, status.HTTP_200_OK)

    # Superuser cannot be contained in list
    self.assertEqual(response.data['extras']['total_results'], 3)



    # SUPERUSER

    self.client.logout()
    self.client.login(username=superuser.username, password=self.password)

    response = self.client.get(url)

    # HTTP_200_OK
    self.assertEqual(response.status_code, status.HTTP_200_OK)

    # All users contained in list
    self.assertEqual(response.data['extras']['total_results'], 4)

我在python 2.7中使用以下软件包版本:

Django==1.8.6
djangorestframework==3.3.1
Markdown==2.6.4
MySQL-python==1.2.5
wheel==0.24.0

UPDATE

我正在使用默认的django缓存,这意味着我没有在django设置中添加任何缓存.

正如所建议的那样,我尝试禁用默认的Django缓存:

CACHES = {
    'default': {
        'BACKEND': 'django.core.cache.backends.dummy.DummyCache',
    }
}

REST_FRAMEWORK = {
    'DEFAULT_AUTHENTICATION_CLASSES': (
        'rest_framework.authentication.SessionAuthentication',
    )
}

问题就在于此.

即使我认为问题不在这里,这是我的UserViewSet:

api.py(重要部分)

class UserViewSet(  
    mixins.RetrieveModelMixin, 
    mixins.UpdateModelMixin,
    mixins.ListModelMixin,
    viewsets.GenericViewSet
):
    queryset = User.objects.all()
    serializer_class = UserExpenseSerializer
    permission_classes = (IsAuthenticated, )
    allowed_methods = ('GET', 'PATCH', 'OPTIONS', 'HEAD')

    def get_serializer_class(self):
        if self.action == 'retrieve':
            return UserExpenseSerializer
        return UserSerializer

    def get_queryset(self):
        if(self.action == 'list'):
            return User.objects.all()
        if self.request.user.is_superuser:
            return User.objects.all()
        if self.request.user.is_staff:
            return User.objects.exclude(is_superuser=True)
        return User.objects.filter(pk = self.request.user.id)

    def list(self, request):
        filter_obj = UsersFilter(self.request)
        users = filter_obj.do_query()
        extras = filter_obj.get_extras()
        serializer = UserSerializer(users, context={'request' : request}, many=True)
        return Response({'users' : serializer.data, 'extras' : extras}, views.status.HTTP_200_OK)

filters.py

class UsersFilter:
    offset = 0
    limit = 50
    count = 0
    total_pages = 0
    filter_params = {}

    def __init__(self, request):

        if not request.user.is_superuser:
            self.filter_params['is_superuser'] = False

        if (not request.user.is_superuser and not request.user.is_staff):
            self.filter_params['pk'] = request.user.id

        # Read query params
        rpp = request.query_params.get('rpp') or 50
        page = request.query_params.get('page') or 1
        search_string = request.query_params.get('search')

        # Validate

        self.rpp = int(rpp) or 50
        self.page = int(page) or 1

        # Set filter
        set_if_not_none(self.filter_params, 'username__contains', search_string)

        # Count total results

        self.count = User.objects.filter(**self.filter_params).count()
        self.total_pages = int(self.count / self.rpp) + 1

        # Set limits
        self.offset = (self.page - 1) * self.rpp
        self.limit = self.page * self.rpp

    def get_filter_params(self):
        return self.filter_params

    def get_offset(self):
        return self.offset

    def get_limit(self):
        return self.limit

    def do_query(self):
        users = User.objects.filter(**self.filter_params)[self.offset:self.limit]
        return users

    def get_query_info(self):
        query_info = {
            'total_results' : self.count,
            'results_per_page' : self.rpp,
            'current_page' : self.page,
            'total_pages' : self.total_pages
        }
        return query_info

更新2

正如Linovia指出的那样,问题不是缓存或任何其他DRF问题,而是过滤器.这是固定的过滤器类:

class UsersFilter:

    def __init__(self, request):

        self.filter_params = {}
        self.offset = 0
        self.limit = 50
        self.count = 0
        self.total_pages = 0
        self.extras = {}

        if not request.user.is_superuser:
        # and so long...

Answer 1

实际上，您创建了一个新用户，该用户应该有 2 个用户，并且您将长度断言为 3。即使没有缓存，也无法工作。

编辑：所以你实际上遇到了问题，因为在类级别使用可变对象。

这是邪恶的代码：

class UsersFilter:
    filter_params = {}

    def __init__(self, request):
        if not request.user.is_superuser:
            self.filter_params['is_superuser'] = False

实际上应该是：

class UsersFilter:
    def __init__(self, request):
        filter_params = {}
        if not request.user.is_superuser:
            self.filter_params['is_superuser'] = False

否则，UsersFilter.filter_params 将从一个请求保留到另一个请求，并且永远不会重置。有关此内容的更多详细信息，请参阅http://www.toptal.com/python/python-class-attributes-an-overly-thorough-guide 。