GraphQL Query Mapper

GraphQL Query Mapper transform resolvers info argument into structure that can be used to filter data based on fields queried by client.

Example use case

When building public GraphQL API external developers can often missuse the queries. For example getProfile query that does expensive fetch from different servers can be used only to fetch username for home page. To prevent from overfetching we can extract information about required fields from the info object and avoid expensive queries.

Query Mapping

Query mapping will prevent from server side database overfetching data by providing list of the fields that were requested in the client side query. Developers can use them to perform targeted queries against their database and rest endpoints.

Library exposes following methods:

getQueryObject: provides list of fields that user queried with additional helpers for database access

Example:

import { getQueryObject} from '@aerogear/graphql-query-mapper'

const resolvers = {
  Query: {
    models (_, params, context, info) {
      const queryData = getQueryObject(info)
      console.log(`${queryData.getRootFields()}`)
    }
  }
}

API

getQueryObject returns following type

    /**
     * Query fields specified in client side query
     */
    fields: string[];

    /**
     * All relations that are part of the query
     */
    relations: {
        [relationName: string]: any;
    };

    /**
     * Check if object has relations
     */
    hasRelations(): boolean;

    /**
     * Checks if object has specified relation
     */
    hasRelation(name: string): boolean;

    /**
     * Returns root fields in format acceptable for most of the sql queries
     * @param separator - separates variables (default ,)
     */
    getRootFields(separator?: string): string;

    /**
     * Returns relation fields in format acceptable for most of the sql queries.
     * Method works with PostgresDB, MySQL and any other database that supports
     * this syntax.
     *
     * @param mapper - argument that maps composite field to single one.
     * By default `as` for PostgreSQL. Use `on`for mysql.
     * @param separator - separates variables (default ,)
     */
    getRelationFields(relation: string, mapper?: string, separator?: string): any;

    /**
     * Expands single key structure returned from database to graph that can
     * be returned by resolver. Method picks all fields that starts with relation name.
     * For example 'relation__field' and puts them into nested relation structure.
     */
    expandToGraph(data: any): any;

Limitations

Derived fields will still require additional checks in the resolver. For example fullname that consist of the firstName+secondName from database:

if(fields.fullname){
    fields.push("firstName")
    fields.push("secondName")
}

Performance consideration

When using query mapper we can opt out from default GraphQL query execution logic and only use only top level (root) query resolvers. Root resolvers can fetch all data required from relationships and deliver it much faster than in classical execution plan that needs to traverse thru entire graph.

Additionally developers can use graphql compiler to provide V8 optimializations for Node.js queries. See https://github.com/zalando-incubator/graphql-jit for more information.

Applying this patterns will help to archieve up to ~15 times better performance comparimng to using graphql reference implementation. This aproach will not require Facebook Data Loader cache layer sice all queries and data will be controlled from the root.

Roadmap

• [x] Parsing info object
• [x] Relational data query filtering
• [ ] NoSQL data query filtering

Contributing

• Star repository
• Check CONTRIBUTING.md

License

Apache-2.0

Notes

Project maintained by AeroGear GraphQL Team: https://github.com/aerogear/graphql-home