Skip to content
Mongoose model converter to GraphQL types with resolvers for graphql-compose
JavaScript TypeScript
Branch: master
Clone or download
Latest commit dfeb5a8 Nov 19, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
.vscode fix: upgrade till flow 0.110.0 Nov 1, 2019
flow-typed/npm fix: graphql-compose peer dependency & upgrade other dependencies Jun 4, 2019
src chore: update dependencies Nov 19, 2019
.babelrc build(Babel): set node 8 as min build version Apr 28, 2019
.eslintignore feat: add mjs build (es6 modules) Jun 26, 2018
.eslintrc chore: update dependencies Apr 28, 2019
.flowconfig fix: definitions for Flowtype@0.85 Nov 5, 2018
.gitignore feat: add mjs build (es6 modules) Jun 26, 2018
.markdownlint.json refactor: due changes in graphql-compose@6.0.1 Mar 15, 2019
.npmignore Initial commit Jun 7, 2016
.prettierrc chore: update dependencies Apr 28, 2019
.travis.yml chore: update dependencies Sep 5, 2018
AUTHORS Initial commit Jun 7, 2016 docs(Changelog): See in the GitHub releases tab Jan 25, 2017 Initial commit Jun 7, 2016 docs: remove stale badge Oct 18, 2019
package.json chore: update dependencies Nov 19, 2019
tsconfig.json feat(TypeScript): add typings for `DiscriminatorTypeComposer`, add he… Sep 22, 2018
tslint.json test: fix after mongoose upgrade Mar 9, 2019
yarn.lock chore: update dependencies Nov 19, 2019


travis build codecov coverage npm trends Join the chat at Commitizen friendly

This is a plugin for graphql-compose, which derives GraphQLType from your mongoose model. Also derives bunch of internal GraphQL Types. Provide all CRUD resolvers, including graphql connection, also provided basic search via operators ($lt, $gt and so on).


npm install graphql graphql-compose mongoose graphql-compose-mongoose --save

Modules graphql, graphql-compose, mongoose are in peerDependencies, so should be installed explicitly in your app. They have global objects and should not have ability to be installed as submodule.

If you want to add additional resolvers connection and/or pagination - just install following packages and graphql-compose-mongoose will add them automatically.

npm install graphql-compose-connection graphql-compose-pagination --save

Different builds

This library contains different builds for any purposes:

// Default import for using under node v6 and above
import { composeWithMongoose } from 'graphql-compose-mongoose';

// For using node 8 and above (native async/await)
import { composeWithMongoose } from 'graphql-compose-mongoose/node8';

// Source code without Flowtype declarations
import { composeWithMongoose } from 'graphql-compose-mongoose/es';


Live demo:

Source code:

Small explanation for variables naming:

  • UserSchema - this is a mongoose schema
  • User - this is a mongoose model
  • UserTC - this is a ObjectTypeComposer instance for User. ObjectTypeComposer has GraphQLObjectType inside, avaliable via method UserTC.getType().
  • Here and in all other places of code variables suffix ...TC means that this is ObjectTypeComposer instance, ...ITC - InputTypeComposer, ...ETC - EnumTypeComposer.
import mongoose from 'mongoose';
import { composeWithMongoose } from 'graphql-compose-mongoose';
import { schemaComposer } from 'graphql-compose';

const LanguagesSchema = new mongoose.Schema({
  language: String,
  skill: {
    type: String,
    enum: [ 'basic', 'fluent', 'native' ],

const UserSchema = new mongoose.Schema({
  name: String, // standard types
  age: {
    type: Number,
    index: true,
  languages: {
    type: [LanguagesSchema], // you may include other schemas (here included as array of embedded documents)
    default: [],
  contacts: { // another mongoose way for providing embedded documents
    email: String,
    phones: [String], // array of strings
  gender: { // enum field with values
    type: String,
    enum: ['male', 'female', 'ladyboy'],
  someMixed: {
    type: mongoose.Schema.Types.Mixed,
    description: 'Can be any mixed type, that will be treated as JSON GraphQL Scalar Type',
const User = mongoose.model('User', UserSchema);

const customizationOptions = {}; // left it empty for simplicity, described below
const UserTC = composeWithMongoose(User, customizationOptions);

// STEP 3: Add needed CRUD User operations to the GraphQL Schema
// via graphql-compose it will be much much easier, with less typing
  userById: UserTC.getResolver('findById'),
  userByIds: UserTC.getResolver('findByIds'),
  userOne: UserTC.getResolver('findOne'),
  userMany: UserTC.getResolver('findMany'),
  userCount: UserTC.getResolver('count'),
  userConnection: UserTC.getResolver('connection'),
  userPagination: UserTC.getResolver('pagination'),

  userCreateOne: UserTC.getResolver('createOne'),
  userCreateMany: UserTC.getResolver('createMany'),
  userUpdateById: UserTC.getResolver('updateById'),
  userUpdateOne: UserTC.getResolver('updateOne'),
  userUpdateMany: UserTC.getResolver('updateMany'),
  userRemoveById: UserTC.getResolver('removeById'),
  userRemoveOne: UserTC.getResolver('removeOne'),
  userRemoveMany: UserTC.getResolver('removeMany'),

const graphqlSchema = schemaComposer.buildSchema();
export default graphqlSchema;

That's all! You think that is to much code? I don't think so, because by default internally was created about 55 graphql types (for input, sorting, filtering). So you will need much much more lines of code to implement all these CRUD operations by hands.

Working with Mongoose Collection Level Discriminators

Variable Namings

  • ...DTC - Suffix for a DiscriminatorTypeComposer instance, which is also an instance of ObjectTypeComposer. All fields and Relations manipulations on this instance affects all registered discriminators and the Discriminator Interface.
  import mongoose from 'mongoose';
  import { schemaComposer } from 'graphql-compose';
  import { composeWithMongooseDiscriminators } from 'graphql-compose-mongoose';

  // pick a discriminatorKey
  const DKey = 'type';

  const enumCharacterType = {
    PERSON: 'Person',
    DROID: 'Droid',

  const CharacterSchema = new mongoose.Schema({
    // _id: field...
    type: {
      type: String,
      require: true,
      enum: (Object.keys(enumCharacterType): Array<string>),
      description: 'Character type Droid or Person',

    name: String,
    height: Number,
    mass: Number,
    films: [String],

  const DroidSchema = new mongoose.Schema({
    makeDate: String,
    primaryFunction: [String],

  const PersonSchema = new mongoose.Schema({
    gender: String,
    hairColor: String,
    starships: [String],

  // set discriminator Key
  CharacterSchema.set('discriminatorKey', DKey);

  // create base Model
  const CharacterModel = mongoose.model('Character', CharacterSchema);

  // create mongoose discriminator models
  const DroidModel = CharacterModel.discriminator(enumCharacterType.DROID, DroidSchema);
  const PersonModel = CharacterModel.discriminator(enumCharacterType.PERSON, PersonSchema);

  // create DiscriminatorTypeComposer
  const baseOptions = { // regular TypeConverterOptions, passed to composeWithMongoose
    fields: {
      remove: ['friends'],
  const CharacterDTC = composeWithMongooseDiscriminators(CharacterModel, baseOptions);

  // create Discriminator Types
  const droidTypeConverterOptions = {  // this options will be merged with baseOptions -> customisationsOptions
    fields: {
      remove: ['makeDate'],
  const DroidTC = CharacterDTC.discriminator(DroidModel, droidTypeConverterOptions);
  const PersonTC = CharacterDTC.discriminator(PersonModel);  // baseOptions -> customisationsOptions applied

  // You may now use CharacterDTC to add fields to all Discriminators
  // Use DroidTC, `PersonTC as any other ObjectTypeComposer.
    droidCreate: DroidTC.getResolver('createOne'),
    personCreate: PersonTC.getResolver('createOne'),

  const schema = schemaComposer.buildSchema();

  describe('createOne', () => {
    it('should create child document without specifying DKey', async () => {
      const res = await graphql.graphql(
        `mutation CreateCharacters {
          droidCreate(record: {name: "Queue XL", modelNumber: 360 }) {
            record {

          personCreate(record: {name: "mernxl", dob: 57275272}) {
            record {

        data: {
          droidCreate: {
            record: { __typename: 'Droid', type: 'Droid', name: 'Queue XL', modelNumber: 360 },
          personCreate: {
            record: { __typename: 'Person', type: 'Person', name: 'mernxl', dob: 57275272 },


Can I get generated vanilla GraphQL types?

const UserTC = composeWithMongoose(User);
UserTC.getType(); // returns GraphQLObjectType
UserTC.getInputType(); // returns GraphQLInputObjectType, eg. for args
UserTC.get('languages').getType(); // get GraphQLObjectType for nested field
UserTC.get('fieldWithNesting.subNesting').getType(); // get GraphQL type of deep nested field

How to add custom fields?

  lonLat: ObjectTypeComposer.create('type LonLat { lon: Float, lat: Float }'),
  notice: 'String', // shorthand definition
  noticeList: { // extended
    type: '[String]', // String, Int, Float, Boolean, ID, Json
    description: 'Array of notices',
    resolve: (source, args, context, info) => 'some value',
  bio: {
    type: GraphQLString,
    description: 'Providing vanilla GraphQL type'

How to build nesting/relations?

Suppose you User model has friendsIds field with array of user ids. So let build some relations:

    resolver: () => UserTC.getResolver('findByIds'),
    prepareArgs: { // resolver `findByIds` has `_ids` arg, let provide value to it
      _ids: (source) => source.friendsIds,
    projection: { friendsIds: 1 }, // point fields in source object, which should be fetched from DB
    resolver: () => UserTC.get('$findMany'), // shorthand for `UserTC.getResolver('findMany')`
    prepareArgs: { // resolver `findMany` has `filter` arg, we may provide mongoose query to it
      filter: (source) => ({
        _operators : { // Applying criteria on fields which have
                       // operators enabled for them (by default, indexed fields only)
          _id : { in: source.friendsIds },
          age: { gt: 21 }
        gender: source.gender,
      limit: 10,
    projection: { friendsIds: 1, gender: 1 }, // required fields from source object

Reusing the same mongoose Schema in embedded object fields

Suppose you have a common structure you use as embedded object in multiple Schemas. Also suppose you want the structure to have the same GraphQL type across all parent types. (For instance, to allow reuse of fragments for this type) Here are Schemas to demonstrate:

import { Schema } from 'mongoose';

const ImageDataStructure = Schema({
  url: String,
  dimensions : {
    width: Number,
    height: Number
}, { _id: false });

const UserProfile = Schema({
  fullName: String,
  personalImage: ImageDataStructure

const Article = Schema({
  title: String,
  heroImage: ImageDataStructure

If you want the ImageDataStructure to use the same GraphQL type in both Article and UserProfile you will need create it as a mongoose schema (not a standard javascript object) and to explicitly tell graphql-compose-mongoose the name you want it to have. Otherwise, without the name, it would generate the name according to the first parent this type was embedded in.

Do the following:

import { schemaComposer } from 'graphql-compose'; // get the default schemaComposer or your created schemaComposer
import { convertSchemaToGraphQL } from 'graphql-compose-mongoose';

convertSchemaToGraphQL(ImageDataStructure, 'EmbeddedImage', schemaComposer); // Force this type on this mongoose schema

Before continuing to convert your models to TypeComposers:

import mongoose from 'mongoose';
import { composeWithMongoose } from 'graphql-compose-mongoose';

const UserProfile = mongoose.model('UserProfile', UserProfile);
const Article = mongoose.model('Article', Article);

const UserProfileTC = composeWithMongoose(UserProfile);
const ArticleTC = composeWithMongoose(Article);

Then, you can use queries like this:

query {
  topUser {
    personalImage {
  topArticle {
    heroImage {
fragment fullImageData on EmbeddedImage {
  dimensions {
    width height

Access and modify mongoose doc before save

This library provides some amount of ready resolvers for fetch and update data which was mentioned above. And you can create your own resolver of course. However you can find that add some actions or light modifications of mongoose document directly before save at existing resolvers appears more simple than create new resolver. Some of resolvers accepts before save hook which can be provided in resolver params as param named beforeRecordMutate. This hook allows to have access and modify mongoose document before save. The resolvers which supports this hook are:

  • createOne
  • createMany
  • removeById
  • removeOne
  • updateById
  • updateOne

The prototype of before save hook:

(doc: mixed, rp: ResolverResolveParams) => Promise<*>,

The typical implementation may be like this:

// extend resolve params with hook
rp.beforeRecordMutate = async function(doc, rp) {
  doc.userTouchedAt = new Date();

  const canMakeUpdate  = await performAsyncTask( ...provide data from doc... )
  if (!canMakeUpdate) {
    throw new Error('Forbidden!');

  return doc;

You can provide your implementation directly in type composer:

UserTC.wrapResolverResolve('updateById', next => async rp => {

  // extend resolve params with hook
  rp.beforeRecordMutate = async (doc, resolveParams) => { ... };

  return next(rp);

or you can create wrappers for example to protect access:

function adminAccess(resolvers) {
  Object.keys(resolvers).forEach((k) => {
    resolvers[k] = resolvers[k].wrapResolve(next => async rp => {

      // extend resolve params with hook
      rp.beforeRecordMutate = async function(doc, rp) { ... }

      return next(rp)
  return resolvers

// and wrap the resolvers
  createResource: ResourceTC.getResolver('createOne'),
  createResources: ResourceTC.getResolver('createMany'),
    updateResource: ResourceTC.getResolver('updateById'),
    removeResource: ResourceTC.getResolver('removeById'),

How can I push/pop or add/remove values to arrays?

The default resolvers, by design, will replace (overwrite) any supplied array object when using e.g. updateById. If you want to push or pop a value in an array you can use a custom resolver with a native MongoDB call.

For example (push):-

// Define new resolver 'pushToArray'
  name: 'pushToArray',
  type: UserTC,
  args: { userId: 'MongoID!', valueToPush: 'String' },
  resolve: async ({ source, args, context, info }) => {
    const user = await User.update({ _id: args.userId }, { $push: { arrayToPushTo: args.valueToPush } } })
    if (!user) return null // or gracefully return an error etc...
    return User.findOne({ _id: args.userId }) // return the record

// Then add 'pushToArray' as a graphql field e.g.
schemaComposer.Mutation.addFields({userPushToArray: UserTC.getResolver('pushToArray')})

User is the corresponding Mongoose model. If you do not wish to allow duplicates in the array then replace $push with $addToSet. Read the graphql-compose docs on custom resolvers for more info:

NB if you set unique: true on the array then using the update $push approach will not check for duplicates, this is due to a MongoDB bug: For more usage examples with $push and arrays see the MongoDB docs here Also note that $push will preserve order in the array (append to end of array) whereas $addToSet will not.

Is it possible to use several schemas?

By default composeWithMongoose uses global schemaComposer for generated types. If you need to create different GraphQL schemas you need create own schemaComposers and provide them to customizationOptions:

import { SchemaComposer } from 'graphql-compose';

const schema1 = new SchemaComposer();
const schema2 = new SchemaComposer();

const UserTCForSchema1 = composeWithMongoose(User, { schemaComposer: schema1 });
const UserTCForSchema2 = composeWithMongoose(User, { schemaComposer: schema2 });

Customization options

When we convert model const UserTC = composeWithMongoose(User, customizationOptions); you may tune every piece of future derived types and resolvers.

Here is flow typed definition of this options:

The top level of customization options. Here you setup name and description for the main type, remove fields or leave only desired fields.

export type customizationOptions = {
  schemaComposer?: SchemaComposer<TContext>, // will be used global schema if not provided specific instance
  name?: string,
  description?: string,
  fields?: {
    only?: string[],
    remove?: string[],
  inputType?: typeConverterInputTypeOpts,
  resolvers?: false | typeConverterResolversOpts,

This is opts.inputType level of options for default InputTypeObject which will be provided to all resolvers for filter and input args.

export type typeConverterInputTypeOpts = {
  name?: string,
  description?: string,
  fields?: {
    only?: string[],
    remove?: string[],
    required?: string[]

This is opts.resolvers level of options. If you set the option to false it will disable resolver or some of its input args. Every resolver's arg has it own options. They described below.

export type typeConverterResolversOpts = {
  findById?: false,
  findByIds?: false | {
    limit?: limitHelperArgsOpts | false,
    sort?: sortHelperArgsOpts | false,
  findOne?: false | {
    filter?: filterHelperArgsOpts | false,
    sort?: sortHelperArgsOpts | false,
    skip?: false,
  findMany?: false | {
    filter?: filterHelperArgsOpts | false,
    sort?: sortHelperArgsOpts | false,
    limit?: limitHelperArgsOpts | false,
    skip?: false,
  updateById?: false | {
    record?: recordHelperArgsOpts | false,
  updateOne?: false | {
    record?: recordHelperArgsOpts | false,
    filter?: filterHelperArgsOpts | false,
    sort?: sortHelperArgsOpts | false,
    skip?: false,
  updateMany?: false | {
    record?: recordHelperArgsOpts | false,
    filter?: filterHelperArgsOpts | false,
    sort?: sortHelperArgsOpts | false,
    limit?: limitHelperArgsOpts | false,
    skip?: false,
  removeById?: false,
  removeOne?: false | {
    filter?: filterHelperArgsOpts | false,
    sort?: sortHelperArgsOpts | false,
  removeMany?: false | {
    filter?: filterHelperArgsOpts | false,
  createOne?: false | {
    record?: recordHelperArgsOpts | false,
  createMany?: false | {
    records?: recordHelperArgsOpts | false,
  count?: false | {
    filter?: filterHelperArgsOpts | false,
  connection?: false | {
    uniqueFields: string[],
    sortValue: mixed,
    directionFilter: (<T>(filterArg: T, cursorData: CursorDataType, isBefore: boolean) => T),
  pagination?: false | {
    perPage?: number,

This is opts.resolvers.[resolverName].[filter|sort|record|limit] level of options. You may tune every resolver's args independently as you wish. Here you may setup every argument and override some fields from the default input object type, described above in opts.inputType.

export type filterHelperArgsOpts = {
  filterTypeName?: string, // type name for `filter`
  isRequired?: boolean, // set `filter` arg as required (wraps in GraphQLNonNull)
  onlyIndexed?: boolean, // leave only that fields, which is indexed in mongodb
  requiredFields?: string | string[], // provide fieldNames, that should be required
  operators?: filterOperatorsOpts | false, // provide filtering fields by operators, eg. $lt, $gt
                                           // if left empty - provides all operators on indexed fields

// supported operators names in filter `arg`
export type filterOperatorNames =  'gt' | 'gte' | 'lt' | 'lte' | 'ne' | 'in[]' | 'nin[]';
export type filterOperatorsOpts = { [fieldName: string]: filterOperatorNames[] | false };

export type sortHelperArgsOpts = {
  sortTypeName?: string, // type name for `sort`

export type recordHelperArgsOpts = {
  recordTypeName?: string, // type name for `record`
  isRequired?: boolean, // set `record` arg as required (wraps in GraphQLNonNull)
  removeFields?: string[], // provide fieldNames, that should be removed
  requiredFields?: string[], // provide fieldNames, that should be required

export type limitHelperArgsOpts = {
  defaultValue?: number, // set your default limit, if it not provided in query (default: 1000)

Used plugins


This plugin adds connection resolver. Build in mechanism allows sort by any unique indexes (not only by id). Also supported compound sorting (by several fields).

Besides standard connection arguments first, last, before and after, also added great arguments:

  • filter arg - for filtering records
  • sort arg - for sorting records

This plugin completely follows to Relay Cursor Connections Specification.


This plugin adds pagination resolver.



You can’t perform that action at this time.