gql-generator-node
v2.1.16
Published
Generate queries as simple function of schema.
Downloads
1,607
Readme
gql-generator-node
Generate queries based on GraphQL schema.
npm install gql-generator-node --save-dev
ToC
Functionality
Given any schema:
type Query {
user(id: Int!): User!
}
type User {
id: Int!
username: String!
email: String!
createdAt: String!
}
this library automatically creates queries like:
query user($id: Int!) {
user(id: $id){
id
username
email
createdAt
}
}
Features
It supports all query types:
- Query
- Mutation
- Subscription
as well as all fields descriptors, including unions, interfaces and fragments.
Last but not least it addresses corner cases - like circular reference.
Usage
The most basic usage is to generate all queries at once by passing schema to generateAll function:
import {generateAll} from 'gql-generator-node';
const {queries, mutations, subscriptions} = generateAll(schema);
console.log(mutations.signup);
/*
mutation signup($username: String!, email: String!, password: String!){
signup(username: $username, email: $email, password: $password){
token
user {
id
username
email
createdAt
}
}
}
*/
Advanced usage
Generate single query
import {generateQuery} from "gql-generator-node";
const query = generateQuery({
field: schema
.getQueryType()
.getFields().user
})
console.log(query);
/*
Query user($user_context_user_details_region_language: String, $user_details_region_language: String, $id: Int!){
user(id: $id){
id
username
email
createdAt
context{
user{
id
username
email
createdAt
details{
... on Guest {
region(language: $user_context_user_details_region_language)
}
... on Member {
address
}
}
}
domain
}
details{
... on Guest {
region(language: $user_details_region_language)
}
... on Member {
address
}
}
}
}
*/
Limit query fields
By default query is generated with all nested fields (skipping only circular references), however this behavior can be customised by passing skeleton of object we are interested in. For instance:
const query = generateQuery({
field: schema
.getQueryType()
.getFields().user,
skeleton: {
'email':
true
}
})
console.log(query);
/*
Query user($id: Int!){
user(id: $id){
email
}
}
*/
Custom dedupe
As default top variables names correspond to schema while nested ones can be addressed by the path - so all of them can be addressed independently in a declarative way. Ex:
mutation signup($signup_user_context_user_details_region_language: String, $signup_user_details_region_language: String, $email: String!, $username: String!, $password: String!){
signup(email: $email, username: $username, password: $password){
token
user{
id
username
email
createdAt
context{
user{
id
username
email
createdAt
details{
... on Guest {
region(language: $signup_user_context_user_details_region_language)
}
... on Member {
address
}
}
}
domain
}
details{
... on Guest {
region(language: $signup_user_details_region_language)
}
... on Member {
address
}
}
}
}
}
Yet some applications might take advantage of custom dedupe functions to for instance to send same argument to all sub fields using same name:
gqlGenerator(schema,depth,({args})=>{
const o = {};
(args || []).forEach(arg=>{
o[arg.name] = arg;
});
return o;
})
=>
mutation signup($language: String, $email: String!, $username: String!, $password: String!){
signup(email: $email, username: $username, password: $password){
token
user{
id
username
email
createdAt
context{
user{
id
username
email
createdAt
details{
... on Guest {
region(language: $language)
}
... on Member {
address
}
}
}
domain
}
details{
... on Guest {
region(language: $language)
}
... on Member {
address
}
}
}
}
}
Example use case
I personally use it to write graphql endpoints tests.
Assuming GraphQL schema:
type Mutation {
signup(
email: String!
username: String!
password: String!
): UserToken!
}
type UserToken {
token: String!
user: User!
}
type User {
id: Int!
username: String!
email: String!
createdAt: String!
}
Before this tool, one needed to write GraphQL API test like this:
test('signup', async () => {
const query = `mutation signup($username: String!, email: String!, password: String!){
signup(username: $username, email: $email, password: $password){
token
user {
id
username
email
createdAt
}
}
}`;
return graphql(query);
});
As gqlGenerator
can generate queries, above test becomes:
const {queries} = generateAll(schema.getMutationType().signup);
const variables = { username: "I", email: "[email protected]", password: '1234' };
test.each(Object.entries(queries))('%s', async ([name,query]) =>
graphql(query,{variables})
);
It not only greatly simplifies testing which might be now automated and batched but also ensures that you would never miss the field to test. Last but not least there is no code duplication between schema and test so most schema updates does not force tests update.
Notes
- Variable names are derived from argument names, so variables generated from multiple occurrences of the same argument name must be deduped. By default, subtree arguments are given path prefix (ex. can be found in dedupe description).
Credits
Code has has its origins at modelo/gql-generator, however it greatly diverged from this implementation.
Contribution
Please feel free open the issues! Although the current stage satisfies my application usage, I would be happy to provide help and improvements if there will be a need for it. Also you can gratify it with star, if you find it useful.