gql-ts

Enables dynamic query features for your REST. Just like GraphQL but REST. How about using:

GET https://your.rest.api/orders?$fields=id,amount,items.length&customer_id=123&$sort=createdAt:ASC&limit=10

Instead of:

POST https://your.rest.api/orders
{
	query {
			order(customer_id: 123, orderBy: { createdAt: asc }, limit: 10) {
				id,
				amount,
				items {
					length
				}
			}
		}
}

Or even more f*ck:

POST https://your.rest.api/orders/get_by_customer_for_screen_abc
{
	"customer_id": "123"
}

GraphQL is too verbose, heavy and 100% depends on their huge libs. I just love the idea of having a dynamic endpoint that clients can customize their query.

It's the best match if you already have a REST application and want to and dynamic query feature

Installation

• This is for typescript only. So we need typescript
• reflect-metadata must be turned on (tsconfig.json)

{
	"experimentalDecorators": true,
	"emitDecoratorMetadata": true
}

• Install via npm or [yarn](https://www.npmjs.com/package/yarn):

npm install --save gql-ts

Getting started

Assume we have a user model stored in MongoDB:

interface IUser {
	_id: ObjectId;
	name: string;
	age: number;
	phoneNumber: string;
	address: string;
	email: string;
}

const UserModel = mongodb.collection<IUser>('user')

To start with, we define a corresponding GQL model:

@GQLObject("user")
export class GQLUser extends GQLModel<IUser, GQLUser> {
	@GQLField()
	_id: string;

	@GQLField()
	name: string;

	@GQLField()
	age: number;

	@GQLField()
	phoneNumber: string;

	@GQLField()
	address: string;

	@GQLField()
	email: string;
}

Add a static resolver method to this class to handle queries. For now, just returns all users:

@GQLRootResolver()
static rootResolve(query: GQLQuery) {
	return UserModel.find().toArray()
}

Next, add this model to GQLGlobal. It's the pre-defined object to manage GQL's models:

// Call this function somewhere in your code, just make sure it's triggered once:
function initGQL() {
	// This script to handle ObjecId -> string conversion
	GQLU.Parsers.unshift((gql, spec, val) => {
		if (spec.rawType == GQLBaseType.STRING && ObjectId.isValid(val)) {
			return  `${val}`;
		}
	});

	GQLGlobal.add(GQLUser);
}

Finally, let's use it in our controller, for example an express routing handler:

app.get('/users', function (req, res) {
	const query = GQLGlobal.queryFromHttpQuery(q, GQLUser);
	query.resolve().then(users => res.send(users))
})

Now we can query to the /users endpoint dynamically:

# To get all users
GET https://localhost/users?$fields=*

# To get only users' name and phone
GET https://localhost/users?$fields=name,phoneNumber

Model definition

Quick Example:

@GQLObject("user")
export class GQLUser extends GQLModel<IUser, GQLUser> {
	@GQLField()
	_id: string;

	@GQLResolver({ matches:  () => true })
	static async rootResolve(query: GQLQuery) {
		return []
	}
}

Model declaration:

A GQL model class must be decorated with GQLObject and extends from GQLModel class:

• GQLObject is the decoration for GQL model class. It has only one argument:
  • name: Name of the model, cannot be duplicated with other models
• GQLModel<DataModel, GQLModelClass> is the base class of every GQL model. It requires 2 generic arguments:
  • DataModel: class or interface of the raw data model
  • GQLModelClass: the current class itself

Model components

There are 3 main components in a model:

• GQLField: Defines data fields in a model and their properties
• GQLResolver: Functions that handle query logic. How to retrieve the data
• GQLMapper: Functions that convert (format) data object to GQL ojbect

GQLField:

Field decorator, define a field in a model, the only argument is the options of the field IGQLFieldOptions:

• type: () => GQLType (Optional): a function that returns the type of the field, can be one value of enum GQLBaseType or another GQL model. Default is the field's defined type. Note: It must be explicitly declared if it's not primitive type.
• dataName: string (Optional): Name of the associated field in data object. Default is the field name.
• tags: string[] (Optional): Tags of the field, only used in customized logic. Default is empty.
• autoSelect: boolean (Optional): Whether this field is selected in wildcard (*) selection or not. Default is true if the field type is primitive, otherwise is false.

GQLResolver:

Function decorator, indicate functions that handle queries:

• priority: number: A model can have multiple resolvers, use this option to indicate their order
• matches: (GQLFilter) => boolean: A function to verify if this resolver is able to handle the query or not. Usually, it use GQLU.byFields helper function to apply required & optional fields logic.

A resolver function receives only one arguments, it's the GQLQuery.

GQLMapper:

Usually, all fields will be mapped automatically from data object to GQL object. GQLMapper can be used in case of customization or association.

Example: Customize mapping for total field:

@GQLObject("transaction")
export  class  GQLTransaction  extends  GQLModel<ITransaction, GQLTransaction> {

	@GQLField()
	_id: string;

	@GQLField({})
	price: number;

	@GQLField()
	quantity: number;

	@GQLField()
	total: number;

	@GQLMapper({ fields: ['total'], addRawFields: ['price', 'quantity'] })
	static async totalMapping(query: GQLQuery, transactions: GQLTransaction[]) {
		transactions.forEach(tr => tr.total = tr.raw.price * tr.raw.quantity)
		return transactions
	}
}

Code samples

Some code samples for common use cases:

Resolve query logic

// In this example, we handle query for field `_id`
@GQLResolver({ matches:  GQLU.byFields(['_id'], []) })
static async rootResolve(query: GQLQuery) {
	// Get value of the query fields if there is
	const  ids = query.filter.get('_id').batch()
		.filter(id => ObjectID.isValid(id))
		.map(id => new ObjectID(id));

	// construct db query object, GQLU.notEmpty will omit all empty fields
	const q = GQLU.notEmpty({
		_id:  ids.length > 0 ? {$in: ids} : null
	});

	return await UserModel.find(q, { projection: query.QueryFields })
}

Resolve with pagination & sort

// Resolver for 2 fields (optional) `_id`, `phone`
@GQLResolver({ matches:  GQLU.byFields([], ['_id', 'phoneNumber']) })
static rootResolve(query: GQLQuery) {
	const  ids = query.filter.get('_id').batch().filter((id: string) =>  ObjectID.isValid(id)).map((id: string) =>  new  ObjectID(id));
	const phones = query.filter.get('phoneNumber').batch()

	// construct db query object
	const q = GQLU.notEmpty({
		_id:  ids.length > 0 ? {$in: ids} : null,
		phoneNumber:  phones.length > 0 ? {$in: phones} : null
	});

	// handle pagination
	if (query.pagination.from?._id) {
		_.set(q, '_id.$gt', new ObjectId(query.pagination.from?._id))
	}

	if (query.pagination.to?._id) {
		_.set(q, '_id.$lt', new ObjectId(query.pagination.to?._id))
	}

	const cursor = UserModel.find(q, { projection: query.QueryFields });

	// handle order by
	const sort = query.sort;
	if  (!_.isEmpty(sort)) {
		cursor.sort(sort.fields.map(f =>  ([f.field, f.OrderNumber])));
	}

	// handle limit
	if (_.isNumber(query.pagination?.limit)) {
		cursor.limit(query.pagination?.limit)
	}

	return cursor.toArray()
}

Too complicated ? Usually I write a helper function and use it for every resolver:

@GQLResolver({ matches:  GQLU.byFields([], ['_id', 'phoneNumber']) })
static async rootResolve(query: GQLQuery) {
	const ids = query.filter.get('_id').batch()
		.filter(id =>  ObjectID.isValid(id))
		.map(id =>  new  ObjectID(id));

	const phones = query.filter.get('phoneNumber').batch()

	// construct db query object
	const q = GQLU.notEmpty({
		_id:  ids.length > 0 ? {$in: ids} : null,
		phoneNumber:  phones.length > 0 ? {$in: phones} : null
	});

	return await gqlMongoQuery(GQLUser, query, UserModel, q);
}

The helper function, this is for mongodb but you can modify it with to db driver easily:

function gqlMongoQuery<T>(gqlModel, gqlQuery: GQLQuery, mgCollection: mongodb.Collection<T>, mgQuery: any) {
	const  gql = gqlQuery.gql
	const  spec = GQLGlobal.get(gqlModel);

	_.keys(gqlQuery.pagination.from)
		.map(k => spec.getKey(k))
		.filter(ks => ks != null && !_.isEmpty(gqlQuery.pagination.from[ks.key]))
		.forEach(ks => {
			const val = gqlQuery.pagination.from[ks.key]
			const qVal = mongodb.ObjectID.isValid(val) ? new mongodb.ObjectID(val) : GQLU.gqlParse(gql, ks, val)
			_.set(mgQuery, `${ks.key}.$gt`, qVal)
		})

	_.keys(gqlQuery.pagination.to)
	.map(k  =>  spec.getKey(k))
	.filter(ks  => ks != null && !_.isEmpty(gqlQuery.pagination.to[ks.key]))
	.forEach(ks  => {
		const val = gqlQuery.pagination.to[ks.key]
		const qVal = mongodb.ObjectID.isValid(val) ? new mongodb.ObjectID(val) : GQLU.gqlParse(gql, ks, val)
		_.set(mgQuery, `${ks.key}.$lt`, qVal)
	})

	const  cursor = mgCollection.find(mgQuery).project(GQLU.arrToObj(gqlQuery.QueryFields, f => f, f => 1));

	const  sort = gqlQuery.sort;
	if (!hera.isEmpty(sort)) {
		cursor.sort(sort.fields.map(f  => ([f.field, f.OrderNumber])));
	}

	if (_.isNumber(gqlQuery?.pagination?.limit)) {
		cursor.limit(gqlQuery?.pagination?.limit)
	}

	return cursor.toArray()
}

Some queries with pagination and sorting:

# To get all users
GET https://localhost/users?$fields=*

# To get user by id
GET https://localhost/users?$fields=*&_id=609ffe100000000000000000

# To get only users with pagination
GET https://localhost/users?$fields=*&$from=_id:609ffe100000000000000000&$limit=20&$sort=_id:ASC

Association mapping

There might be associations in the model and we should setup custom mapper for nested query:

// Assume that we have 2 associated GQL objects User and Company:
@GQLObject('company')
class GQLCompnay extends GQLModel<ICompany, GQLCompany> {
	@GQLField()
	_id: string;
	
	@GQLField()
	name: string;

	@GQLResolver({matches: GQLU.byFields([], ['_id'])})
	static async rootResolve(query: GQLQuery) {
		const ids = query.filter.get('_id').batch()
			.filter(id =>  ObjectID.isValid(id))
			.map(id =>  new  ObjectID(id));
	
		// construct db query object
		const q = GQLU.notEmpty({
			_id:  ids.length > 0 ? {$in: ids} : null
		});
	
		return await gqlMongoQuery(GQLCompany, query, CompanyModel, q);
	}
}

@GQLObject('user')
class GQLUser extends GQLModel<IUser, GQLUser> {
	@GQLField()
	_id: string

	// a user will belong to a company, indicated by field: `company_id`
	@GQLField({type: () => GQLCompany, dataName: 'company_id'})
	company: GQLCompany

	@GQLResolver({matches: GQLU.byFields([], ['_id'])})
	static async rootResolve(query: GQLQuery) {
		// ... resolve by `_id`, just like above
	}
}

We've defined a reference for company field in GQLUser , but it's not define how exactly the data is mapped, we have to create an explicit mapper for it:

class GQLUser extends GQLModel<IUser, GQLUser> {
	// ...
	
	@GQLMapper({fields: ['company']})
	static async companyMapping(query: GQLQuery, users: GQLUsers[]) {
		// get the list of associated company ids
		const companyIds = _.uniq(users.map(u => u.raw.company_id))

		// retreive the query and add filter for those company ids
		const subQuery = query.select.get('company')?.subQuery || query.emptyQuery(GQLCompany)
		subQuery.filter.add(new GQLFieldFilter('_id', companyIds));
    subQuery.select.addRawField('_id');

		// get the companies:
		const companies = await subQuery.resolve();

		// mapping data
		users.forEach(u => u.company = companies.find(c => c.raw._id == u.raw.company_id)
		
		return users
	}
}

Now we can select nested company from the user query:

GET https://localhost/users?$fields=*,company._id,company.name

Still too complicated ? We have a helper decorator to handle that mapping called GQLFieldRevMapping (sorry for the bad name). The case above become much simpler:

class GQLUser extends GQLModel<IUser, GQLUser> {
	
	// ...

	@GQLField({type: () => GQLCompany, dataName: 'company_id'})
	@GQLFieldRevMapping({queryField: '_id'}) // <- here, just this line instead of a long mapper
	company: GQLCompany
}

Meta resolver

Sometimes, we need to retrieve other information beside of the data (eg: count number of matched objects, next, prev link for pagination...). We called them metadata :

@GQLObject('user')
class GQLUser extends GQLModel<IUser, GQLUser> {
	@GQLField()
	_id: string

	@GQLField()
	company_id: string;

	@GQLResolver({matches: GQLU.byFields([], ['_id', 'company'])})
	static async rootResolve(query: GQLQuery) {
		const q = this.makeDbQuery(query)
		return await gqlMongoQuery(GQLUser, query, UserModel, q)
	}

	@GQLMetaResolver({field: 'count'})
	static async metaCountResolve(query: GQLQuery) {
		const q = this.makeDbQuery(query)
		return await UserModel.count(q)
	}

	static makeDbQuery(query: GQLQuery) {
		const ids = query.filter.get('_id').batch()
			.filter(id =>  ObjectID.isValid(id))
			.map(id =>  new  ObjectID(id));

		const companyIds = query.filter.get('company_id').batch()

		return GQLU.notEmpty({
			_id:  ids.length > 0 ? {$in: ids} : null,
			company_id: companyIds.length > 0 ? {$in: companyIds} : null
		})
	}
}

In above example, we defined a metadata field called count that return the number of matched users for the query. Now we can query it from the HTTP request:

GET https://localhost/users?$fields=*&company_id=1&$meta=count

Permission filter

... To be added. Check Whitelist / blacklist filters

Other objects

GQLUtils

GQLU is an utilities class used in the lib. It has many useful functions that might be very helpful during applying this lib:

• GQLU.notEmpty(data: any, isEmpty: Function, deep: boolean) : Returns a new object that filtered out all empty fields.

  • data: the object to be filter
  • isEmpty: function that check if an object is empty or not (default: GQLU.isEmpty)
  • deep: Deep or shallow filter (default: false)

• GQLU.isEmpty(data: any) : Checks if a data is empty or not

  isEmpty(obj?: any): boolean {
      return  ((obj == null || obj === NaN || obj === false) ||
              (isString(obj) && obj.length == 0) ||
              ((obj instanceof Array) && obj.length == 0) ||
              ((obj instanceof Object) && Object.keys(obj).length == 0));
  }

• GQLU.whiteListSelect(query: GQLQuery, ...whiteList: string[]) : Throws GQLUnauthorizedQuery if the query select a field that not in the passed whitelist

• GQLU.blackListSelect(query: GQLQuery, ...blacklist: string[]) : Same as above, but for blacklist

• GQLU.whiteListFilter(query: GQLQuery, ...whitelist: string[]) : Same as above, but for filter instead of select

• GQLU.blackListFilter(query: GQLQuery, ...blacklist: string[]) : Same as above, for blacklist filter

• GQLU.requireFilter(query: GQLQuery, ...requireds: string[]) : Same as above, the query must have filter on all passed fields

• GQLU.byFields(requiredFields: string[], optionalFields?: string[]) : Returns a matching function for GQLQuery

  • requiredFields: Return false if the query filter doesn't have query on one of these fields
  • optionalFields: Return false if the query filter have at least one field that is not in this list or requiredFields list.

GQLGlobal

GQLGlobal is an instance of class GQL. It's a repository that manage all GQL models. It means, in order to use a GQL model, it has to be added into a GQL object. Usually, using GQLGLobal is enough. But we could have multiple GQL object in some advance situations.

GQLQuery

GQLQuery contains all information of a query. It includes:

• select: GQLSelect
  • fields: GQLFieldSelect[] : List of selected fields
    • field: string : Field name
    • type: GQLType : Type of the field
    • subQuery: GQLQuery : Sub query of that field if there is
  • rawFields: string[]: List of raw fields. Raw fields is fields should be retrieved from the raw data objects (usually selected from database in resolver)- while normal fields is fields will be responsed from the GQL
  • get(field: string) : Get select info of a field or null if there's none
  • add(...fields: string[]) : Add fields into selection
  • addRawField(...fields: string[]) : Add fields into raw selection
• filter: GQLFilter
  • filters: GQLFieldFilter[] : List of filtered fields
    • field: string : Name of the field
    • first(): string : Returns the value of the filter
    • batch(): string[] : return the list of the filter values
• pagination: GQLPagination Contains pagination information
  • from: {[field: string]: string} : Dictionary contains lower bound infos
  • to: {[field: string]: string} : Dictionary contains upper bound infos
  • limit: number : Limit size
  • offset: number : Offset / skip number
• sort: GQLSort : Sort / Orderby information
  • fields: GQLSortField[] : List of sorted fields
    • field: string : field name
    • order: 'ASC' | 'DESC'
    • OrderNumber: number 1 for ASC and -1 for DESC
• meta: GQLMetaSelect: Metadata informat
  • fields: string[]: List of selected metadata fields
• resolve(): Promise<any[]>: Resolve the query, returns matched & processed GQL objects
• resolveMeta(): Promise<any> : Resolve query's metadata. Returns an objects contains selected metadata
• hasMeta: boolean : Whether a query has metadata or not
• emptyQuery(model: GQLModel): returns an empty query for a specified models
• QueryFields: string[] List of field that must be retrieved in data objects. Usually it's fields that have to selected from database query in resolver.

Decorators

Some useful decorators that helps writing GQLModel much simpler:

• @GQLRootResolver : If your model only have 1 resolver, use this decorator instead

• @GQLIdenticalMapping : Instead of using default Parser or mapper. The field value will be copied identcally from data object

• @GQLFieldRevMapping : Use to mapping association between models

  • targetType: GQLModel : Type of the target model. Default is the type of the field.
  • queryField: string : Name of the field in target model that is associated with the current field. Default is 'id'
  • extractField: string : name of the field to get mapping value from the source objects. Default is name of the current field. Advance usecase only.
  • extractor: (obj) => any : Function that extracts the mapping values from the source objects. Default is getting value of the extractField. Advance usecase only.
  • rawField: string: name of the raw field to be added in the target object. Default is the queryField. Advance usecase only.
  • mappingFilter(sourceObject, targetObject) => boolean : Function determining should we map a targetObject into the sourceObject. Default is just == comparison between values of extractField from sourceObject and queryField of targetObject
  • mappingFunc(sourceObject, targetObjects): Function determine exactly how to map from a list targetObjets into a sourceObject. Advance usecase only.

  It might look complicated. But usually, there're only 2 things we have to think about:

  • queryField : If the mapping field is not id. Eg for MongoDB: It's _id
  • mappingFilter: When the == is not enough. Eg for MongoDB: ObjectId must be compared via equals method

Issue Reporting

If you have found a bug or have a feature request, feel free to report them at this repository issues section.

Contributing

You are welcome