GraphQL-Connections 💠

GraphQL-Connections 💠

Install

Install by referencing the github location and the release number:

npm install --save graphql-connections#v2.2.0

About

GraphQL-Connections helps handle the traversal of edges between nodes.

In a graph, nodes connect to other nodes via edges. In the relay graphql spec, multiple edges can be represented as a single Connection type, which has the signature:

type Connection {
  pageInfo: {
    hasNextPage: string
    hasPreviousPage: string
    startCursor: string
    endCursor: string
  },
  edges:  Array<{cursor: string; node: Node}>
}

A connection object is returned to a user when a query request asks for multiple child nodes connected to a parent node. For example, a music artist has multiple songs. In order to get all the songs for an artist you would write the graphql query request:

query {
    artist(id: 1) {
        songs {
            songName
            songLength
        }
    }
}

However, sometimes you may only want a portion of the songs returned to you. To allow for this scenario, a connection is used to represent the response type of a song.

query {
    artist(id: 1) {
        songs {
            pageInfo {
                hasNextPage
                hasPreviousPage
                startCursor
                endCursor
            }
            edges {
                cursor
                node {
                    songName
                    songLength
                }
            }
        }
    }
}

You can use the cursors (startCursor, endCursor, or cursor) to get the next set of edges.

query {
  artist(id: 1) {
    songs(next: 10, after: <endCursor>) {
      pageInfo {
        hasNextPage
        hasPreviousPage
        startCursor
        endCursor
      }
      edges {
        cursor
        node {
          songName
          songLength
        }
      }
    }
  }
}

The above logic is controlled by the connectionManager. It can be added to a resolver to:

Create a cursor for paging through a node's edges
Handle movement through a node's edges using an existing cursor.
Support multiple input types that can sort, group, limit, and filter the edges in a connection

Run locally

Run the migrations
- NODE_ENV=development npm run migrate:sqlite:latest
- NODE_ENV=development npm run migrate:mysql:latest
Seed the database
- NODE_ENV=development npm run seed:sqlite:run
- NODE_ENV=development npm run seed:mysql:run
Run the dev server
- npm run dev:sqlite (search is not supported)
- npm run dev:mysql (search IS supported :))
Visit the GraphQL playground http://localhost:4000/graphql
Run some queries!

query {
    users(
        first: 100
        orderBy: "haircolor"
        filter: {
            and: [
                {field: "id", operator: ">", value: "19990"}
                {field: "age", operator: "<", value: "90"}
            ]
        }
        search: "random search term"
    ) {
        pageInfo {
            hasNextPage
            hasPreviousPage
            startCursor
            endCursor
        }
        edges {
            cursor
            node {
                username
                lastname
                id
                haircolor
                bio
            }
        }
    }
}

query {
    users(first: 10, after: "eyJmaXJzdFJlc3VsdElkIjoxOTk5MiwibGFzdFJlc3VsdE") {
        pageInfo {
            hasNextPage
            hasPreviousPage
        }
        edges {
            cursor
            node {
                username
                lastname
                id
                haircolor
                bio
            }
        }
    }
}

How to use - Schema

1. Add input scalars to schema

Add the input scalars (First, Last, OrderBy, OrderDir, Before, After, Filter, Search) to your GQL schema.

At the very least you should add Before, After and First, Last because they allow you to move along the connection with a cursor.

type Query {
    users(
        first: First
        last: Last
        orderBy: OrderBy
        orderDir: OrderDir
        before: Before
        after: After
        filter: Filter
        search: Search
    ): QueryUsersConnection
}

2. Add typeDefs to your schema

Manually

scalar: First
scalar: Last
scalar: OrderBy
scalar: OrderDir
scalar: Before
scalar: After
scalar: Filter
scalar: Search

type Query {
  users(
      first: First
      last: Last
      orderBy: OrderBy
      orderDir: OrderDir
      before: Before
      after: After
      filter: Filter
      search: Search
  ): QueryUsersConnection
}

String interpolation

import {typeDefs} from 'graphql-connections'

const schema = `
  ${...typeDefs}
  type Query {
    users(
        first: First
        last: Last
        orderBy: OrderBy
        orderDir: OrderDir
        before: Before
        after: After
        filter: Filter
        search: Search
    ): QueryUsersConnection
  }
`

During configuration

import {typeDefs as connectionTypeDefs} from 'graphql-connections'

    const schema = makeExecutableSchema({
        ...
        typeDefs: `${typeDefs} ${connectionTypeDefs}`
    });

3. Add resolvers

In the resolver object

import {resolvers as connectionResolvers} from 'graphql-connections'

const resolvers = {
  ...connectionResolvers,

  Query: {
    users: {
      ....
    }
  }
}

During configuration

import {resolvers as connectionResolvers} from 'graphql-connections'

    const schema = makeExecutableSchema({
        ...
        resolvers: {...resolvers, ...connectionResolvers}
    });

How to use - Resolvers

Short Tutorial

In short, this is what a resolver using the connectionManager will look like:

// import the manager and relevant types
import {ConnectionManager, INode} from 'graphql-connections';

const resolver = async (obj, inputArgs) => {
    // create a new node connection instance
    const nodeConnection = new ConnectionManager<INode>(inputArgs, attributeMap);

    // apply the connection to the queryBuilder
    const appliedQuery = nodeConnection.createQuery(queryBuilder.clone());

    // run the query
    const result = await appliedQuery.select();

    // add the result to the nodeConnection
    nodeConnection.addResult(result);

    // return the relevant connection information from the resolver
    return {
        pageInfo: nodeConnection.pageInfo,
        edges: nodeConnection.edges
    };
};

types:

// the type of each returned node
interface INode {
  [nodeField: string]: any
}

// input types to control the edges returned
interface IInputArgs {
  before?: string;
  after?: string;
  first?: number;
  last?: number;
  orderBy?: string;
  orderDir?: keyof typeof ORDER_DIRECTION;
  filter?: IInputFilter;
}

// map of node field to sql column name
interface IInAttributeMap {
  [nodeField: string]: string;
}

// the nodeConnection class type
interface INodeConnection {
  createQuery: (KnexQueryBuilder) => KnexQueryBuilder
  addResult: (KnexQueryResult) => void
  pageInfo?: IPageInfo
  edges?: IEdge[]

interface IPageInfo: {
  hasNextPage: string
  hasPreviousPage: string
  startCursor: string
  endCursor: string
}

interface IEdge {
  cursor: string;
  node: INode
}

All types can be found in src/types.ts

Detailed Tutorial

A nodeConnection is used to handle connections.

To use a nodeConnection you will have to:

initialize the nodeConnection
build the connection query
build the connection from the executed query

1. Initialize the `nodeConnection`

To correctly initialize, you will need to supply a Node type, the inputArgs args, and an attributeMap map:

A. set the `Node` type

The nodes that are part of a connection need a type. The returned edges will contain nodes of this type.

For example, in this case we create an IUserNode

interface IUserNode extends INode {
    id: number;
    createdAt: string;
}

B. add inputArgs

InputArgs supports before, after, first, last, orderBy, orderDir, and filter:

interface IInputArgs {
    before?: string; // cursor
    after?: string; // cursor
    first?: number; // page size
    last?: number; // page size
    orderBy?: string; // order by a node field
    orderDir: 'asc' | 'desc';
    filter?: IOperationFilter;
}

interface IFilter {
    value: string;
    operator: string;
    field: string; // node field
}

interface IOperationFilter {
    and?: Array<IOperationFilter & IFilter>;
    or?: Array<IOperationFilter & IFilter>;
    not?: Array<IOperationFilter & IFilter>;
}

Note: The default filter operators are the normal SQL comparison operators: >, <, =, >=, <=, and <>

An example query with a filter could look like:

query {
    users(
        filter: {
            or: [
                {field: "age", operator: "=", value: "40"}
                {field: "age", operator: "<", value: "30"}
                {
                    and: [
                        {field: "haircolor", operator: "=", value: "blue"}
                        {field: "age", operator: "=", value: "70"}
                        {
                            or: [
                                {field: "username", operator: "=", value: "Ellie86"}
                                {field: "username", operator: "=", value: "Euna_Oberbrunner"}
                            ]
                        }
                    ]
                }
            ]
        }
    ) {
        pageInfo {
            hasNextPage
            hasPreviousPage
        }
        edges {
            cursor
            node {
                id
                age
                haircolor
                lastname
                username
            }
        }
    }
}

This would yield a sql query equivalent to:

  SELECT *
    FROM `mock`
   WHERE `age` = '40' OR `age` < '30' OR (`haircolor` = 'blue' AND `age` = '70' AND (`username` = 'Ellie86' OR `username` = 'Euna_Oberbrunner'))
ORDER BY `id`
     ASC
   LIMIT 1001

C. specify an attributeMap

attributeMap is a map of GraphQL field names to SQL column names

Only fields defined in the attribute map can be orderBy or filtered on. An error will be thrown if you try to filter on fields that don't exist in the map.

ex:

const attributeMap = {
    id: 'id',
    createdAt: 'created_at'
};

2. build the query

// import the manager and relevant types
import {ConnectionManager, INode} from 'graphql-connections';

const resolver = async (obj, inputArgs) => {
    // create a new node connection instance
    const nodeConnection = new ConnectionManager<
        IUserNode,
    >(inputArgs, attributeMap);

    // apply the connection to the queryBuilder
    const appliedQuery = nodeConnection.createQuery(queryBuilder.clone());

    ....
}

3. execute the query and build the `connection`

A connection type has the signature:

type Connection {
  pageInfo: {
    hasNextPage: string
    hasPreviousPage: string
    startCursor: string
    endCursor: string
  },
  edges:  Array<{cursor: string; node: Node}>
}

This type is constructed by taking the result of executing the query and adding it to the connectionManager instance via the addResult method.

// import the manager and relevant types
import {ConnectionManager, INode} from 'graphql-connections';

const resolver = async (obj, inputArgs) => {
    ...

    // run the query
    const result = await appliedQuery

    // add the result to the nodeConnection
    nodeConnection.addResult(result);

    // return the relevant connection information from the resolver
    return {
        pageInfo: nodeConnection.pageInfo,
        edges: nodeConnection.edges
    };

Options

You can supply options to the ConnectionManager via the third parameter. Options are used to customize the QueryContext, the QueryBuilder, and the QueryResult classes.

    const options = {
      contextOptions: { ... }
      resultOptions: { ... }
      builderOptions: { ... }
    }
    const nodeConnection = new ConnectionManager(inputArgs, attributeMap, options);

Currently, the options are:

contextOptions

defaultLimit

number;

The default limit (page size) if none is specified in the page input params

cursorEncoder

interface ICursorEncoder<CursorObj> {
    encodeToCursor: (cursorObj: CursorObj) => string;
    decodeFromCursor: (cursor: string) => CursorObj;
}

builderOptions - common

filterTransformer

type filterTransformer = (filter: IFilter) => IFilter;

The filter transformer will will be called on every filter { field: string, operator: string, value: string}

It can be used to transform a filter before being applied to the query. This is useful if you want to transform say UnixTimestamps to DateTime format, etc...

See the filter transformation section for more details.

filterMap

interface IFilterMap {
    [operator: string]: string;
}

The filter operators that can be specified in the filter input params.

If no map is specified, the default one is used:

const defaultFilterMap = {
    '>': '>',
    '>=': '>=',
    '=': '=',
    '<': '<',
    '<=': '<=',
    '<>': '<>'
};

builderOptions - MySQL specific

searchColumns

Used with full text Search input. It is an array of column names that will be used in the full text search sql expression MATCH (col1,col2,...) AGAINST (expr [search_modifier])

searchColumns: string[]

searchModifier

Used with full text Search input. It is an array of column names that will be used in the full text search sql expression MATCH (col1,col2,...) AGAINST (expr [search_modifier])

searchColumns: string;

The values will likely be one of:

      'IN NATURAL LANGUAGE MODE',
    | 'IN NATURAL LANGUAGE MODE WITH QUERY EXPANSION',
    | 'IN BOOLEAN MODE',
    | 'WITH QUERY EXPANSION'

resultOptions

nodeTransformer

type NodeTransformer<Node> = (node: any) => Node;

A function that is will be called during the creation of each node. This can be used to map the query result to a Node type that matches the graphql Node for the resolver.

cursorEncoder

interface ICursorEncoder<CursorObj> {
    encodeToCursor: (cursorObj: CursorObj) => string;
    decodeFromCursor: (cursor: string) => CursorObj;
}

Extensions

To extend the connection to a new datastore or to use an adapter besides Knex, you will need to create a new QueryBuilder. See src/KnexQueryBuilder for an example of what a query builder looks like. It should have the type signature:

interface IQueryBuilder<Builder> {
    createQuery: (queryBuilder: Builder) => Builder;
}

Architecture

Internally, the ConnectionManager manages the orchestration of the QueryContext, QueryBuilder, and QueryResult.

The orchestration follows the steps:

The QueryContext extracts the connection attributes from the input connection arguments.
The QueryBuilder (or KnexQueryBuilder in the default case) consumes the connection attributes and builds a query. The query is submitted to the database by the user and the result is sent to the QueryResult.
The QueryResult uses the result to build the edges (which contain a cursor and node) and extract the page info.

This can be visualized as such:

Search

Search inputs are provided for executing full text search query strings against a datastore. At the moment only MySQL support exists. Using filters may slow down the query.

An example resolver might look like:

const attributeMap = {
    id: 'id',
    username: 'username',
    firstname: 'firstname',
    age: 'age',
    haircolor: 'haircolor',
    lastname: 'lastname',
    bio: 'bio'
};

const builderOptions = {
    searchColumns: ['username', 'firstname', 'lastname', 'bio', 'haircolor'],
    searchModifier: 'IN NATURAL LANGUAGE MODE'
};
const nodeConnection = new ConnectionManager<IUserNode>(inputArgs, attributeMap, {
    builderOptions
});

const query = nodeConnection.createQuery(queryBuilder.clone()).select();
const result = (await query) as KnexQueryResult;

nodeConnection.addResult(result);

return {
    pageInfo: nodeConnection.pageInfo,
    edges: nodeConnection.edges
};

Filtering on computed columns

Sometimes you may compute a field that depends on some other table than the one being paged over. In this case, you can use a derived table as your from and alias it to the primary table. In the following example we create a derived alias of "segment", the table we are paging over, to allow filtering and sorting on "popularity", a field computed on the aggregation of values from another table.

import {Resolver} from 'types/resolver';
import {ISegmentNode} from 'types/graphql';
import {segment as segmentTransformer} from 'transformers/sql_to_graphql';
import {IQueryResult, IInputArgs, ConnectionManager} from 'graphql-connections';

const attributeMap = {
    createdAt: 'created_at',
    updatedAt: 'updated_at',
    name: 'name',
    explorer: 'explorer_id',
    popularity: 'popularity'
};

const segments: Resolver<Promise<IQueryResult<ISegmentNode>>, undefined, IInputArgs> = async (
    _,
    input: IInputArgs,
    ctx
) => {
    const {connection} = ctx.clients.sqlClient;

    const queryBuilder = connection.queryBuilder().from(
        connection.raw(
            `(
            select
                segment.*,
                coalesce(sum(user_segment.usage_count), 0) as popularity
            from segment
                     left join user_segment on user_segment.segment_id = segment.id
            group by
                segment.id
        ) as segment`
        )
    );

    const nodeConnection = new ConnectionManager<ISegmentNode>(input, attributeMap, {
        builderOptions: {
            filterTransformer(filter) {
                if (filter.field === 'popularity') {
                    return {
                        field: 'popularity',
                        operator: filter.operator,
                        value: Number(filter.value) as any
                    };
                }

                return filter;
            }
        },
        resultOptions: {
            nodeTransformer: segmentTransformer
        }
    });

    const query = nodeConnection.createQuery(queryBuilder).select('*');

    nodeConnection.addResult(await query);

    return {
        pageInfo: nodeConnection.pageInfo,
        edges: nodeConnection.edges
    };
};

export default segments;

Filter Transformation

Sometimes you may have a completely different data type in a filter from what is actually in your database. At Social Native, for example, our graph exposes all timestamps as Unix Seconds, but in our databases, the timestamp type is used. In order to easily manage filter value transformation from seconds to sql timestamps, we use the filterTransformer option. In the following example, we use the library-provided FilterTransformers.castUnixSecondsFiltersToMysqlTimestamps which takes a list of field names that should be transformed from unix seconds to mysql timestamps, if they are present and not falsy.

import {FilterTransformers} from 'graphql-connections';

type SomeGraphQLNode = {
    createdAt: string | number;
    updatedAt: string | number;
};

const timestampFilterTransformer = FilterTransformers.castUnixSecondsFiltersToMysqlTimestamps<
    SomeGraphQLNode
>(['createdAt', 'updatedAt']);

const nodeConnection = new ConnectionManager<GqlActualDistribution | null>(input, inAttributeMap, {
    builderOptions: {
        filterTransformer: timestampFilterTransformer
    },
    resultOptions: {nodeTransformer: sqlToGraphql.actualDistribution}
});

A compose function is also exposed to combine multiple transformers together. The following example composes a transformer on 'createdAt', 'updatedAt' with one on 'startedAt', 'completedAt', creating one that will cast if any of the four were given.

import {FilterTransformers} from 'graphql-connections';

const timestampFilterTransformer = FilterTransformers.castUnixSecondsFiltersToMysqlTimestamps([
    'createdAt',
    'updatedAt'
]);

const nodeConnection = new ConnectionManager<GqlActualDistribution | null>(input, inAttributeMap, {
    builderOptions: {
        filterTransformer: FilterTransformers.compose(
            timestampFilterTransformer,
            FilterTransformers.castUnixSecondsFiltersToMysqlTimestamps(['startedAt', 'completedAt'])
        )
    },
    resultOptions: {nodeTransformer: sqlToGraphql.actualDistribution}
});

Filter Transformers provided by this library

FilterTransformers.castUnixSecondsFiltersToMysqlTimestamps

castUnixSecondsFiltersToMysqlTimestamps takes four arguments

function castUnixSecondsFiltersToMysqlTimestamps<T extends Record<string, unknown>>(
    filterFieldsToCast: Array<keyof T>,
    timezone: DateTimeOptions['zone'] = 'UTC',
    includeOffset = false,
    includeZone = false
): FilterTransformer;

filterFieldsToCast refers to the keys in a graphql node being filtered upon that will be transformed from unix seconds to MySQL timestamps. timezone is the expected timezone of the input seconds. Default: UTC includeOffset dictates whether the timestamp offset will be included in the generated timestamp Default: false includeZone dictates whether the timestamp's timezone will be included in the generated timestamp Default: false

Name		Name	Last commit message	Last commit date
Latest commit History 219 Commits
.circleci		.circleci
.github		.github
__tests__		__tests__
db		db
dev		dev
src		src
.env.dev		.env.dev
.env.test		.env.test
.gitignore		.gitignore
.prettierrc		.prettierrc
CLA.md		CLA.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
README.md		README.md
jest.config.js		jest.config.js
knexfile.mysql.ts		knexfile.mysql.ts
knexfile.sqlite.ts		knexfile.sqlite.ts
package-lock.json		package-lock.json
package.json		package.json
rollup.config.js		rollup.config.js
tsconfig.dev.json		tsconfig.dev.json
tsconfig.json		tsconfig.json
tslint.json		tslint.json

License

social-native/graphql-connections

Folders and files

Latest commit

History

Repository files navigation

GraphQL-Connections 💠

Install

About

Run locally

How to use - Schema

1. Add input scalars to schema

2. Add typeDefs to your schema

Manually

String interpolation

During configuration

3. Add resolvers

In the resolver object

During configuration

How to use - Resolvers

Short Tutorial

Detailed Tutorial

1. Initialize the nodeConnection

A. set the Node type

B. add inputArgs

C. specify an attributeMap

2. build the query

3. execute the query and build the connection

Options

contextOptions

defaultLimit

cursorEncoder

builderOptions - common

filterTransformer

filterMap

builderOptions - MySQL specific

searchColumns

searchModifier

resultOptions

nodeTransformer

cursorEncoder

Extensions

Architecture

Search

Filtering on computed columns

Filter Transformation

Filter Transformers provided by this library

FilterTransformers.castUnixSecondsFiltersToMysqlTimestamps

About

Topics

Resources

License

Stars

Watchers

Forks

Releases 29

Packages 0

Contributors 5

Languages

1. Initialize the `nodeConnection`

A. set the `Node` type

3. execute the query and build the `connection`

Packages