GraphQL API Layer for Frappe Framework
MIT
Generate the sdls first
$ bench --site test_site graphql generate_sdl
and start making your graphql requests against:
/api/method/graphql
You can get a single document by its name using <doctype>
query.
For sort by fields, only those fields that are search_indexed / unique can be used. NAME, CREATION & MODIFIED can also be used
Example
Query
{
User(name: "Administrator") {
name,
email
}
}
You can get a list of documents by querying <doctype-plural>
. You can also pass in filters and sorting details as arguments:
{
Users(filter: [["name", "like", "%a%"]], sortBy: { field: NAME, direction: ASC }) {
totalCount,
pageInfo {
hasNextPage,
hasPreviousPage,
startCursor,
endCursor
},
edges {
cursor,
node {
name,
first_name
}
}
}
}
}
All Link fields return respective doc. Add __name
suffix to the link field name to get the link name.
Example
Query
{
ToDo (limit_page_length: 1) {
name,
priority,
description,
assigned_by__name,
assigned_by {
full_name,
roles {
role__name,
role {
name,
creation
}
}
}
}
}
Result
{
"data": {
"ToDo": [
{
"name": "ae6f39845b",
"priority": "Low",
"description": "<div class=\"ql-editor read-mode\"><p>Do this</p></div>",
"assigned_by__name": "Administrator",
"assigned_by": {
"full_name": "Administrator",
"roles": [
{
"role__name": "System Manager",
"role": {
"name": "System Manager",
"creation": "2021-02-02 08:34:42.170306",
}
}
]
}
}
...
]
}
}
Get notified instantly of the updates via existing frappe's SocketIO. Please read more on the implementation details here
File uploads can be done following the GraphQL multipart request specification. uploadFile
mutation is included implementing the same
Example
Query
POST /api/method/graphql HTTP/1.1
Host: test_site:8000
Accept: application/json
Cookie: full_name=Administrator; sid=<sid>; system_user=yes; user_id=Administrator; user_image=
Content-Length: 553
Content-Type: multipart/form-data; boundary=----WebKitFormBoundary7MA4YWxkTrZu0gW
----WebKitFormBoundary7MA4YWxkTrZu0gW
Content-Disposition: form-data; name="operations"
{
"query": "mutation uploadFile($file: Upload!) { uploadFile(file: $file) { name, file_url } }",
"variables": {
"file": null
}
}
----WebKitFormBoundary7MA4YWxkTrZu0gW
Content-Disposition: form-data; name="map"
{ "0": ["variables.file"] }
----WebKitFormBoundary7MA4YWxkTrZu0gW
Content-Disposition: form-data; name="0"; filename="/D:/faztp12/Pictures/BingImageOfTheDay_20190715.jpg"
Content-Type: image/jpeg
(data)
----WebKitFormBoundary7MA4YWxkTrZu0gW
Response
{
"data": {
"uploadFile": {
"name": "ce36b2e222",
"file_url": "/files/BingImageOfTheDay_20190715.jpg"
}
}
}
Data is returned based on Read access to the resource
Cursor based pagination is implemented. You can read more about it here: Cursor Based Pagination
You can extend the SDLs with additional query / mutations / subscriptions. Examples are provided for a specific set of Scenarios. Please read GraphQL Spec regarding Extending types. There are mainly two hooks introduced:
graphql_sdl_dir
Specify a list of directories containing.graphql
files relative to the app's root directory. eg:
# hooks.py
graphql_sdl_dir = [
"./your-app/your-app/generated/sdl/dir1",
"./your-app/your-app/generated/sdl/dir2",
]
The above will look for graphql files in your-bench/apps/your-app/your-app/generated/sdl/dir1
& ./dir2
folders.
graphql_schema_processors
You can pass in a list of cmd that gets executed on schema creation. You are givenGraphQLSchema
object (please refer graphql-core) as the only parameter. You can modify it or extend it as per your requirements. This is a good place to attach the resolvers for the custom SDLs defined viagraphql_sdl_dir
We can add graphql middlewares by adding the path through hooks.
Please note the return type and arguments being passed to your custom middleware.
# hooks.py
graphql_middlewares = ["frappe_graphql.utils.middlewares.disable_introspection_queries.disable_introspection_queries"]
def disable_introspection_queries(next_resolver, obj, info: GraphQLResolveInfo, **kwargs):
# https://github.com/jstacoder/graphene-disable-introspection-middleware
if is_introspection_disabled() and info.field_name.lower() in ['__schema', '__introspection']:
raise IntrospectionDisabled(frappe._("Introspection is disabled"))
return next_resolver(obj, info, **kwargs)
def is_introspection_disabled():
return not cint(frappe.local.conf.get("developer_mode")) and \
not cint(frappe.local.conf.get("enable_introspection_in_production"))
## Introspection in Production Introspection is disabled by default in production mode. You can enable by setting the site config `enable_introspection_in_production: 1`.
- Exception Handling in Resolvers. We provide a utility resolver wrapper function which could be used to return your expected exceptions as user errors. You can read more about it here.
- Role Permissions for Resolver. We provide another utility resolver wrapper function which could be used to verify the logged in User has the roles specified. You can read more about it here
- Generate the SDLs in your app directory
# Generate sdls for all doctypes in <your-app>
$ bench --site test_site graphql generate_sdl --output-dir <your-app/graphql> --app <your-app>
# Generate sdls for all doctype in module <m1> <m2>
$ bench --site test_site graphql generate_sdl --output-dir <your-app/graphql> --module <m1> -m <m2> -m <name>
# Generate sdls for doctype <d1> <2>
$ bench --site test_site graphql generate_sdl --output-dir <your-app/graphql> --doctype <d1> -dt <d2> -dt <name>
# Generate sdls for all doctypes in <your-app> without Enums for Select Fields
$ bench --site test_site graphql generate_sdl --output-dir <your-app/graphql> --app <your-app> --disable-enum-select-fields
- Specify this directory in
graphql_sdl_dir
hook and you are done.
- Add the
Custom Field
in frappe - Add the following to a
.graphql
file in your app and specify its directory viagraphql_sdl_dir
extend type User {
is_super: Int!
}
- Add a cmd in
graphql_schema_processors
hook - Use the following function definition for the cmd specified:
def hello_world(schema: GraphQLSchema):
def hello_resolver(obj, info: GraphQLResolveInfo, **kwargs):
return f"Hello {kwargs.get('name')}!"
schema.query_type.fields["hello"] = GraphQLField(
GraphQLString,
resolve=hello_resolver,
args={
"name": GraphQLArgument(
type_=GraphQLString,
default_value="World"
)
}
)
- Now, you can query like query like this:
# Request
query Hello($var_name: String) {
hello(name: $var_name)
}
# Variables
{
"var_name": "Mars"
}
# Response
{
"data": {
"hello": "Hello Mars!"
}
}
# Add the cmd to the following function in `graphql_schema_processors`
def docmeta_query(schema):
from graphql import GraphQLField, GraphQLObjectType, GraphQLString, GraphQLInt
schema.query_type.fields["docmeta"] = GraphQLField(
GraphQLList(GraphQLObjectType(
name="DocTypeMeta",
fields={
"name": GraphQLField(
GraphQLString,
resolve=lambda obj, info: obj
),
"number_of_docs": GraphQLField(
GraphQLInt,
resolve=lambda obj, info: frappe.db.count(obj)
),
}
)),
resolve=lambda obj, info: [x.name for x in frappe.get_all("DocType")]
)
Please refer graphql-core
for more examples
There are two ways:
-
Write SDL and Attach Resolver to the Schema
# SDL for Mutation type MY_MUTATION_OUTPUT_TYPE { success: Boolean } extend type Mutation { myNewMutation(args): MY_MUTATION_OUTPUT_TYPE }
# Attach Resolver (Specify the cmd to this function in `graphql_schema_processors` hook) def myMutationResolver(schema: GraphQLSchema): def _myMutationResolver(obj: Any, info: GraphQLResolveInfo): # frappe.set_value(..) return { "success": True } mutation_type = schema.mutation_type mutation_type.fields["myNewMutation"].resolve = _myMutationResolver
-
Make use of
graphql-core
apis# Specify the cmd to this function in `graphql_schema_processors` hook def bindMyNewMutation(schema): def _myMutationResolver(obj: Any, info: GraphQLResolveInfo): # frappe.set_value(..) return { "success": True } mutation_type = schema.mutation_type mutation_type.fields["myNewMutation"] = GraphQLField( GraphQLObjectType( name="MY_MUTATION_OUTPUT_TYPE", fields={ "success": GraphQLField( GraphQLBoolean, resolve=lambda obj, info: obj["success"] ) } ), resolve=_myMutationResolver )
def schema_processor(schema: GraphQLSchema):
schema.query_type.fields["hello"] = GraphQLField(
GraphQLString, resolve=lambda obj, info: "World!")