Graphene-Django-Crud

Inspired by prisma-nexus and graphene-django-extras, this package transforms the django orm into a graphql API with the following features:

Expose CRUD opérations
Optimized queryset
Filtering with logical operators
Authentication and permissions
Nested mutations
Subscription fields

Installation

For the use of the subscription fields you must have correctly installed graphene-subscription in your project. For that follow their installation part
The generate signals method allows to connect model signals to graphene-subcription signals.

For the support of Multipart Request Spec, install graphene-file-upload according to the documentation.

Install with pip

To install graphene-django-crud, simply run this simple command in your terminal of choice:

$ pip install graphene-django-crud

Install with source code

graphene-django-crud is developed on GitHub, You can either clone the public repository:

$ git clone https://github.com/djipidi/graphene_django_crud.git

Once you have a copy of the source, you can embed it in your own Python package, or install it into your site-packages easily:

$ cd graphene_django_crud
$ python setup.py install

Usage

The DjangoCRUDObjectType class project a django model into a graphene type. The type has fields to exposes the CRUD operations.

Example

In this example, you will be able to project the auth django models on your GraphQL API and expose the CRUD operations.

# schema.py
import graphene
from graphql import GraphQLError
from django.contrib.auth.models import User, Group
from graphene_django_crud.types import DjangoCRUDObjectType, resolver_hints

class UserType(DjangoCRUDObjectType):
    class Meta:
        model = User
        exclude_fields = ("password",)
        input_exclude_fields = ("last_login", "date_joined")

    full_name = graphene.String()


    @resolver_hints(
      only=["first_name", "last_name"]
    )
    @staticmethod
    def resolve_full_name(parent, info, **kwargs):
        return parent.get_full_name()

    @classmethod
    def get_queryset(cls, parent, info, **kwargs):
        if info.context.user.is_authenticated:
            return User.objects.all()
        else:
            return User.objects.none() 

    @classmethod
    def mutate(cls, parent, info, instance, data, *args, **kwargs):
        if not info.context.user.is_staff:
            raise GraphQLError('not permited, only staff user')

        if "password" in data.keys():
            instance.set_password(data.pop("password"))
        return super().mutate(parent, info, instance, data, *args, **kwargs)

class GroupType(DjangoCRUDObjectType):
    class Meta:
        model = Group

class Query(graphene.ObjectType):

    me = graphene.Field(UserType)
    user = UserType.ReadField()
    users = UserType.BatchReadField()

    group = GroupType.ReadField()
    groups = GroupType.BatchReadField()

    def resolve_me(parent, info, **kwargs):
        if not info.context.user.is_authenticated:
            return None
        else:
            return info.context.user

class Mutation(graphene.ObjectType):

    user_create = UserType.CreateField()
    user_update = UserType.UpdateField()
    user_delete = UserType.DeleteField()

    group_create = GroupType.CreateField()
    group_update = GroupType.UpdateField()
    group_delete = GroupType.DeleteField()

class Subscription(graphene.ObjectType):

    user_created = UserType.CreatedField()
    user_updated = UserType.UpdatedField()
    user_deleted = UserType.DeletedField()

    group_created = GroupType.CreatedField()
    group_updated = GroupType.UpdatedField()
    group_deleted = GroupType.DeletedField()

#signals.py
from .schema import UserType, GroupType

# Necessary for subscription fields
UserType.generate_signals()
GroupType.generate_signals()

And get the resulting GraphQL API:

toggle me

schema {
  query: Query
  mutation: Mutation
  subscription: Subscription
}

scalar DateTime

input DatetimeFilter {
  equals: DateTime
  exact: DateTime
  in: [DateTime]
  isnull: Boolean
  gt: DateTime
  gte: DateTime
  lt: DateTime
  lte: DateTime
  year: IntFilter
  month: IntFilter
  day: IntFilter
  weekDay: IntFilter
  hour: IntFilter
  minute: IntFilter
  second: IntFilter
}

type ErrorType {
  field: String!
  messages: [String!]!
}

input GroupCreateInput {
  name: String!
  userSet: UserCreateNestedManyInput
}

input GroupCreateNestedManyInput {
  create: [GroupCreateInput]
  connect: [GroupWhereInput]
}

input GroupOrderByInput {
  id: OrderEnum
  name: OrderStringEnum
}

type GroupType {
  id: ID
  name: String
  userSet(where: UserWhereInput, orderBy: [UserOrderByInput], limit: Int, offset: Int): UserTypeConnection
}

type GroupTypeConnection {
  data: [GroupType]!
  count: Int
}

type GroupTypeMutation {
  ok: Boolean
  errors: [ErrorType]
  result: GroupType
}

input GroupUpdateInput {
  name: String
  userSet: UserUpdateNestedManyInput
}

input GroupUpdateNestedManyInput {
  create: [GroupCreateInput]
  delete: [GroupWhereInput]
  connect: [GroupWhereInput]
  disconnect: [GroupWhereInput]
}

input GroupWhereInput {
  id: IdFilter
  name: StringFilter
  user: UserWhereInput
  OR: [GroupWhereInput]
  AND: [GroupWhereInput]
  NOT: GroupWhereInput
}

input IdFilter {
  equals: ID
  exact: ID
  in: [ID]
  isnull: Boolean
}

input IntFilter {
  equals: Int
  exact: Int
  in: [Int]
  isnull: Boolean
  gt: Int
  gte: Int
  lt: Int
  lte: Int
  contains: Int
  startswith: Int
  endswith: Int
  regex: String
}

type Mutation {
  userCreate(input: UserCreateInput!): UserTypeMutation
  userUpdate(input: UserUpdateInput!, where: UserWhereInput!): UserTypeMutation
  userDelete(where: UserWhereInput!): UserTypeMutation
  groupCreate(input: GroupCreateInput!): GroupTypeMutation
  groupUpdate(input: GroupUpdateInput!, where: GroupWhereInput!): GroupTypeMutation
  groupDelete(where: GroupWhereInput!): GroupTypeMutation
}

enum OrderEnum {
  ASC
  DESC
}

enum OrderStringEnum {
  ASC
  DESC
}

type Query {
  me: UserType
  user(where: UserWhereInput!): UserType
  users(where: UserWhereInput, orderBy: [UserOrderByInput], limit: Int, offset: Int): UserTypeConnection
  group(where: GroupWhereInput!): GroupType
  groups(where: GroupWhereInput, orderBy: [GroupOrderByInput], limit: Int, offset: Int): GroupTypeConnection
}

input StringFilter {
  equals: String
  exact: String
  in: [String]
  isnull: Boolean
  contains: String
  startswith: String
  endswith: String
  regex: String
  iexact: String
  icontains: String
  istartswith: String
  iendswith: String
}

type Subscription {
  userCreated(where: UserWhereInput): UserType
  userUpdated(where: UserWhereInput): UserType
  userDeleted(where: UserWhereInput): UserType
  groupCreated(where: GroupWhereInput): GroupType
  groupUpdated(where: GroupWhereInput): GroupType
  groupDeleted(where: GroupWhereInput): GroupType
}

input UserCreateInput {
  email: String
  firstName: String
  groups: GroupCreateNestedManyInput
  isActive: Boolean
  isStaff: Boolean
  isSuperuser: Boolean
  lastName: String
  password: String!
  username: String!
}

input UserCreateNestedManyInput {
  create: [UserCreateInput]
  connect: [UserWhereInput]
}

input UserOrderByInput {
  dateJoined: OrderEnum
  email: OrderStringEnum
  firstName: OrderStringEnum
  id: OrderEnum
  isActive: OrderEnum
  isStaff: OrderEnum
  isSuperuser: OrderEnum
  lastLogin: OrderEnum
  lastName: OrderStringEnum
  username: OrderStringEnum
}

type UserType {
  id: ID
  dateJoined: DateTime
  email: String
  firstName: String
  groups(where: GroupWhereInput, orderBy: [GroupOrderByInput], limit: Int, offset: Int): GroupTypeConnection
  isActive: Boolean
  isStaff: Boolean
  isSuperuser: Boolean
  lastLogin: DateTime
  lastName: String
  username: String
  fullName: String
}

type UserTypeConnection {
  data: [UserType]!
  count: Int
}

type UserTypeMutation {
  ok: Boolean
  errors: [ErrorType]
  result: UserType
}

input UserUpdateInput {
  email: String
  firstName: String
  groups: GroupUpdateNestedManyInput
  isActive: Boolean
  isStaff: Boolean
  isSuperuser: Boolean
  lastName: String
  password: String
  username: String
}

input UserUpdateNestedManyInput {
  create: [UserCreateInput]
  delete: [UserWhereInput]
  connect: [UserWhereInput]
  disconnect: [UserWhereInput]
}

input UserWhereInput {
  id: IdFilter
  dateJoined: DatetimeFilter
  email: StringFilter
  firstName: StringFilter
  groups: GroupWhereInput
  isActive: Boolean
  isStaff: Boolean
  isSuperuser: Boolean
  lastLogin: DatetimeFilter
  lastName: StringFilter
  username: StringFilter
  OR: [UserWhereInput]
  AND: [UserWhereInput]
  NOT: UserWhereInput
}

Queries example:

query{
  user(where: {id: {equals:1}}){
    id
    username
    firstName
    lastName
  }
}


query{
  users(
    where: {
      OR: [
        {isStaff: true},
        {isSuperuser: true},
        {groups: {name: {equals: "admin"}}},
      ]
    }
    orderBy: [{username: ASC}],
    limit: 100,
    offset: 0
  ){
    count
    data{
      id
      username
      firstName
      lastName
      groups{
        count
        data{
          id
          name
        }
      }
    }
  }
}

mutation{
  groupCreate(
    input: {
      name: "admin",
      userSet: {
        create: [
          {username: "woody", password: "raC4RjDU"},
        ],
        connect: [
          {id: {equals: 1}}
        ]
      },
    }
  ){
    ok
    result{
      id
      name
      userSet{
        count
        data{
          id
          username
        }
      }
    }
  }
}

Computed Field

You can add computed fields using the standard Graphene API. However to optimize the SQL query you must specify "only", "select_related" necessary for the resolver using the resolver_hints decorator

class UserType(DjangoCRUDObjectType):
    class Meta:
        model = User

    full_name = graphene.String()

    @resolver_hints(
      only=["first_name", "last_name"]
    )
    @staticmethod
    def resolve_full_name(parent, info, **kwargs):
        return parent.get_full_name()

User permissions

The methods mutate, create, update, delete are called for each change of model instances during mutations and nested mutations. They can be used to check permissions.

class UserType(DjangoCRUDObjectType):
    class Meta:
        model = User

    @classmethod
    def mutate(cls, parent, info, instance, data, *args, **kwargs):
        if not info.context.user.is_authenticated:
            raise GraphQLError('not authorized, you must be logged in')
        return super().mutate(parent, info, instance, data, *args, **kwargs)

    @classmethod
    def create(cls, parent, info, instance, data, *args, **kwargs):
        if not info.context.user.has_perm("add_user"):
            raise GraphQLError('not authorized, you must have add_user permission')
        return super().create(parent, info, instance, data, *args, **kwargs)

    @classmethod
    def update(cls, parent, info, instance, data, *args, **kwargs):
        if not info.context.user.has_perm("change_user"):
            raise GraphQLError('not authorized, you must have change_user permission')
        return super().update(parent, info, instance, data, *args, **kwargs)

    @classmethod
    def delete(cls, parent, info, instance, data, *args, **kwargs):
        if not info.context.user.has_perm("delete_user"):
            raise GraphQLError('not authorized, you must have delete_user permission')
        return super().delete(parent, info, instance, data, *args, **kwargs)

Filtering by user

To filter based on the authenticated user, overload the get_queryset method as the example

class UserType(DjangoCRUDObjectType):
    class Meta:
        model = User

    @classmethod
    def get_queryset(cls, parent, info, **kwargs):
        if info.context.user.is_authenticated:
            return User.objects.all()
        else:
            return User.objects.none()

Use with relay

The configuration is the same as graphene-django, just add the "relay.Node" interface.

class CategoryType(DjangoCRUDObjectType):
    class Meta:
        model = Category
        interfaces = (relay.Node, )


class IngredientType(DjangoCRUDObjectType):
    class Meta:
        model = Ingredient
        interfaces = (relay.Node, )


class Query(graphene.ObjectType):

    node = relay.Node.Field()

    category = CategoryType.ReadField()
    all_categories = CategoryType.BatchReadField()

    ingredient = IngredientType.ReadField()
    all_ingredients = IngredientType.BatchReadField()

Relay.global_id as well as model id are supported to write the query using the "id" field of whereInputType.

Extend ConnectionType

By default, graphene_django_crud creates a connection type for the bachread request and the many_to_many/many_to_one relationships.

the default connection has a "count" field returning the count() value of the queryset and a data field returning the results of the queryset.

Use list field

from .models import Product
import graphene
from graphene_django_crud import DjangoCRUDObjectType

class ProductType(DjangoCRUDObjectType):
    class Meta:
        model = Product
        use_connection = False

Extend ConnectionType without Relay

from .models import Product
from django.db.models import Avg
import graphene
from graphene_django_crud import DefaultConnection, DjangoCRUDObjectType

class ConnectionWithPriceAVG(DefaultConnection):
    class Meta:
        abstract = True

    price_avg = graphene.Float()

    def resolve_price_avg(self, info):
        return self.iterable.aggregate(Avg('price'))["price__avg"]

class ProductType(DjangoCRUDObjectType):
    class Meta:
        model = Product
        connection_class = ConnectionWithPriceAVG

Extend ConnectionType with Relay

from .models import Product
import graphene
from graphene_django_crud import DjangoCRUDObjectType

class ConnectionWithTotalCount(graphene.Connection):
    class Meta:
        abstract = True
    total_count = graphene.Int()

    def resolve_total_count(self, info):
        return self.iterable.count()

class ProductType(DjangoCRUDObjectType):
    class Meta:
        model = Product
        interfaces = (relay.Node, )
        connection_class = ConnectionWithTotalCount

DjangoCRUDObjectType Class

From the version v1.3.0, DjangoGrapheneCRUD class has been renamed to DjangoCRUDObjectType, so the name "DjangoGrapheneCRUD" is deprecated.

Meta parameters

model

Required parameter
The model used for the definition type

max_limit

default : None
To avoid too large transfers, the max_limit parameter imposes a maximum number of return items for batchreadField and nodeField. it imposes to use pagination. If the value is None there is no limit.

only_fields / exclude_fields

Tuple of model fields to include/exclude in graphql type. Only one of the two parameters can be declared.

input_only_fields / input_exclude_fields

Tuple of model fields to include/exclude in graphql create and update inputs type. Only one of the two parameters can be declared.

input_extend_fields

Field list to extend the create and update inputs. value must be a list of tuple (name: string, type: graphene.ObjectType). The parameters can be processed with methods mutate, create, update, delete

example:

class UserType(DjangoCRUDObjectType):
    class Meta:
        model = User
        input_extend_fields = (
            ("fullName", graphene.String()),
        )

    @classmethod
    def mutate(cls, parent, info, instance, data, *args, **kwargs):
        if "fullName" in data.keys():
            instance.first_name = data["fullName"].split(" ")[0]
            instance.last_name = data["fullName"].split(" ")[1]
        return super().mutate(parent, info, instance, data, *args, **kwargs)

where_only_fields / where_exclude_fields

Tuple of model fields to include/exclude in graphql where input type. Only one of the two parameters can be declared.

order_by_only_fields / order_by_exclude_fields

Tuple of model fields to include/exclude in graphql order_by input type. Only one of the two parameters can be declared.

validator

default: True
Activate/deactivate the validation of the model. if the value is True, full_clean() method of model will be called before save().

validator_exclude

default: None
The exclude argument of full_clean() method.

validator_validate_unique

default: True
The validate_unique argument of full_clean() method.

Graphene Fields

The DjangoCRUDObjectType class contains configurable operation publishers that you use for exposing create, read, update, and delete mutations against your projected models

for mutating, relation fields may be connected with an existing record or a sub-create may be inlined (generally referred to as nested mutations). If the relation is a List then multiple connections or sub-creates are permitted.

Inlined mutations are very similar to top-level ones but have the important difference that the sub-create has excluded the field where supplying its relation to the type of parent Object being created would normally be. This is because a sub-create forces its record to relate to the parent one.

Warning: By default, mutations are not atomic, specify ATOMIC_REQUESTS or ATOMIC_MUTATIONS on True in your setting.py
See: Transaction with graphene-django

ReadField

Query field to allow clients to find one particular record at time of the respective model.

BatchReadField

Query field to allow clients to fetch multiple records at once of the respective model.

CreateField

Mutation field to allow clients to create one record at time of the respective model.

UpdateField

Mutation field to allow clients to update one particular record at time of the respective model.

DeleteField

Mutation field to allow clients to delete one particular record at time of the respective model.

CreatedField

Subscription field to allow customers to subscribe to the creatied of instances of the respective model.

UpdatedField

Subscription field to allow customers to subscribe to the updated of instances of the respective model.

DeletedField

Subscription field to allow customers to subscribe to the deleted of instances of the respective model.

Input Types

WhereInputType

Input type composed of the scalar filters of each readable fields of the model. The logical operators "OR", "AND", "NO" are also included. the returned arg can be used in queryset with function where_input_to_Q

OrderByInputType

Input type composed of the orderByEnum of each readable fields of the model.

CreateInputType

Input type composed of model fields without the id. If the field is not nullable, the graphene field is required.

UpdateInputType

Input type composed of each fields of the model. No fields are required.

Methods to override

get_queryset(cls, parent, info, **kwargs)

@classmethod
def get_queryset(cls, parent, info, **kwargs):
    return queryset_class

Default it returns "model.objects.all()", the overload is useful for applying filtering based on user. The method is called in nested request, fetch instances for mutations and subscription filter.

mutate, create, update, delete

Methods called for each mutation and nested mutation impacting the model. Overload this method to add preprocessing and / or overprocessing. The mutate method is called before the create, update, delete methods. The "data" argument is a dict corresponding to the graphql input argument.

@classmethod
def mutate(cls, parent, info, instance, data, *args, **kwargs):
    # code before save instance
    instance = super().mutate(cls, parent, info, instance, data, *args, **kwargs)
    # code after save instance
    return instance

@classmethod
def create(cls, parent, info, instance, data, *args, **kwargs):
    # code before save instance
    instance = super().create(cls, parent, info, instance, data, *args, **kwargs)
    # code after save instance
    return instance

@classmethod
def update(cls, parent, info, instance, data, *args, **kwargs):
    # code before save instance
    instance = super().update(cls, parent, info, instance, data, *args, **kwargs)
    # code after save instance
    return instance

@classmethod
def delete(cls, parent, info, instance, data, *args, **kwargs):
    # code before save instance
    instance = super().delete(cls, parent, info, instance, data, *args, **kwargs)
    # code after save instance
    return instance

(Deprecated) Middleware methods before_XXX(cls, parent, info, instance, data) / after_XXX(cls, parent, info, instance, data)

from the version v1.3.0, these methods are deprecated, use the methods mutate, create, update, delete

@classmethod
def before_mutate(cls, parent, info, instance, data):
    pass

@classmethod
def before_create(cls, parent, info, instance, data):
    pass

@classmethod
def before_update(cls, parent, info, instance, data):
    pass

@classmethod
def before_delete(cls, parent, info, instance, data):
    pass

@classmethod
def after_mutate(cls, parent, info, instance, data):
    pass

@classmethod
def after_create(cls, parent, info, instance, data):
    pass

@classmethod
def after_update(cls, parent, info, instance, data):
    pass

@classmethod
def after_delete(cls, parent, info, instance, data):
    pass

Methods called before or after a mutation. The "instance" argument is the instance of the model that goes or has been modified retrieved from the "where" argument of the mutation, or it's been created by the model constructor. The "data" argument is a dict of the "input" argument of the mutation. The method is also called in nested mutation.

generate_signals()

Graphene-subscription needs to connect the model signals to its own signals. the "generate_signals()" method does this for the model.

# signals.py

from .schema import UserType, GroupTyp

UserType.generate_signals()
GroupType.generate_signals()

# apps.py

from django.apps import AppConfig

class CoreConfig(AppConfig):
    name = 'core'

    def ready(self):
        import core.signals

Settings

Graphene-django-crud reads your configuration from a single Django setting named GRAPHENE_DJANGO_CRUD:

GRAPHENE_DJANGO_CRUD = {
    "DEFAULT_CONNECTION_NODES_FIELD_NAME": "nodes"
}

Here’s a list of settings available in graphene-django-crud and their default values:

Customize

DEFAULT_CONNECTION_NODES_FIELD_NAME

Name of node field in connection field.
Default: 'data'

FILE_TYPE_CONTENT_FIELD_ACTIVE

Add a content field with the content of the file. The type used is Binary.
Default: False

Compatibility with old version

SCALAR_FILTERS_ADD_EQUALS_FIELD

From version 1.3.0 the "equals" field of all scalar filters has been renamed to "exact". To keep the client compatible we can add it by set the parameter to True.
Default: False

BOOLEAN_FILTER_USE_BOOLEAN_FIELD

From version 1.3.0 the filter boolean is like the other scalar filters. To keep the client compatible we can add it by set the parameter to True.
Default: False

Utils

@resolver_hints(only: list[str], select_related:list[str])

Each query uses "only", "select_related" and "prefetch_related" methods of queryset to get only the necessary attributes. To extend fields, the decorator is necessary for the queryset builder with its arguments which model attributes are needed to resolve the field.

show Computed field section for more informations

where_input_to_Q(where_input: dict) -> Q

In order to be able to reuse where input generated, the where_input_to_Q function transforms the returned argument into a Q object

example :

<model>.objects.filter(where_input_to_Q(where))

order_by_input_to_args(order_by_input: list[dict]) -> list[str]

In order to be able to reuse order_by input generated, the order_by_input_to_args function transforms the returned argument into args for order_by method of queryset.