Normalize GraphQL responses for persisting in the client cache/state.
Not related, in any way, to normalizr, just shamelessly piggybacking on its popularity. Also, "normalizEr" is taken...
TL;DR: Transforms:
{
"data": {
"findUser": [
{
"__typename": "User",
"id": "5a6efb94b0e8c36f99fba013",
"email": "Lloyd.Nikolaus@yahoo.com",
"posts": [
{
"__typename": "BlogPost",
"id": "5a6efb94b0e8c36f99fba016",
"title": "Dolorem voluptatem molestiae",
"comments": [
{
"__typename": "Comment",
"id": "5a6efb94b0e8c36f99fba019",
"message": "Alias quod est voluptatibus aut quis sunt aut numquam."
},
{
"__typename": "Comment",
"id": "5a6efb94b0e8c36f99fba01b",
"message": "Harum quia asperiores nemo."
},
{
"__typename": "Comment",
"id": "5a6efb94b0e8c36f99fba01c",
"message": "Vel veniam consectetur laborum."
},
{
"__typename": "Comment",
"id": "5a6efb94b0e8c36f99fba01e",
"message": "Possimus beatae vero recusandae beatae quas ut commodi laboriosam."
}
]
}
]
}
]
}
}
into:
{
"comments": {
"5a6efb94b0e8c36f99fba019": {
"id": "5a6efb94b0e8c36f99fba019",
"message": "Alias quod est voluptatibus aut quis sunt aut numquam."
},
"5a6efb94b0e8c36f99fba01b": {
"id": "5a6efb94b0e8c36f99fba01b",
"message": "Harum quia asperiores nemo."
},
"5a6efb94b0e8c36f99fba01c": {
"id": "5a6efb94b0e8c36f99fba01c",
"message": "Vel veniam consectetur laborum."
},
"5a6efb94b0e8c36f99fba01e": {
"id": "5a6efb94b0e8c36f99fba01e",
"message": "Possimus beatae vero recusandae beatae quas ut commodi laboriosam."
}
},
"blogPosts": {
"5a6efb94b0e8c36f99fba016": {
"id": "5a6efb94b0e8c36f99fba016",
"title": "Dolorem voluptatem molestiae",
"comments": [
"5a6efb94b0e8c36f99fba019",
"5a6efb94b0e8c36f99fba01b",
"5a6efb94b0e8c36f99fba01c",
"5a6efb94b0e8c36f99fba01e"
]
}
},
"users": {
"5a6efb94b0e8c36f99fba013": {
"id": "5a6efb94b0e8c36f99fba013",
"email": "Lloyd.Nikolaus@yahoo.com",
"posts": ["5a6efb94b0e8c36f99fba016"]
}
}
}
We all love GraphQL and we want to use it. There are tons of libraries and clients out there that help us do that with ease, but there is still one problem... How do you persist that data?
Yes, everything is all great when the response mirrors the exact structure we asked for, but we don't want to cache it that way, do we? We probably want a normalized version of that data which we can persist to our store and read/modify it efficiently. Flux or Redux stores work best with normalized data and there are also GraphQL clients you can use to execute queries on the local cache/state (blips or apollo-link-state), in which case, we definitely need to persist normalized data.
GraphQLNormalizr is simple, fast, light-weight and it provides all the tools needed to do just that, the only requirement is that you include the id
and __typename
fields for all the nodes (but it can do that for you if you're too lazy or you want to keep your sources thin).
npm install graphql-normalizr
The GraphQLNormalizr constructor function returns an object containing 3 methods:
Depending on how you write your queries, you may or may not use parse
or addRequiredFields
, but normalize
is the method that you will transform the GraphQL response. As you've probably seen from the TL;DR, all response nodes must contain the __typename
and id
fields. __typename
is a GraphQL meta field and the id
key may be customized when creating the GraphQLNormalizr client.
If your queries already ask for id
and __typename
there's no need to use parse or addRequiredFields. Otherwise, parse will take care of transforming your GraphQL source
into a Document
and add the __typename
and id
fields where needed. In case you already use a different parser, or only have access to the Document
you may use addRequiredFields on the Document
to add the __typename
and id
fields
import { GraphQLNormalizr } from 'graphql-normalizr'
// const config = ...
const normalizer = new GraphQLNormalizr(config)
config: optional - the configuration object containing information for instantiating the client. it takes the following props:
String
Default is "id". Configures a custom id
key for the client. Use this if your resource identifiers are found under a different key name ('_id', 'key', 'uid' etc).
Consider the following GraphQL response:
const response = {
data: {
findUser: {
__typename: 'User',
uid: '5a6efb94b0e8c36f99fba013',
email: 'Lloyd.Nikolaus@yahoo.com',
},
},
}
Normalize the data with our custom id
key:
// using destructuring to get the `normalize` method of the client
const { normalize } = new GraphQLNormalizr({ idKey: 'uid' })
normalize(response)
// =>
// {
// users: {
// '5a6efb94b0e8c36f99fba013' : {
// uid: '5a6efb94b0e8c36f99fba013',
// email: 'Lloyd.Nikolaus@yahoo.com'
// }
// }
// }
Boolean
Default is false
. If you are using GraphQL connections with edges
and nodes
, set this flag to true
otherwise you'll get a warning and the normalization won't work.
NOTE: The connections implementation needs to be according to the specification
const response = {
data: {
findUser: {
__typename: 'User',
id: '5a6efb94b0e8c36f99fba013',
email: 'Lloyd.Nikolaus@yahoo.com',
friends: {
__typename: 'FriendsConnection',
totalCount: 3,
edges: [
{
node: {
__typename: 'User',
id: '5a6cf127c2b20834f6551481',
email: 'Madisen_Braun@hotmail.com',
},
cursor: 'Y3Vyc29yMg==',
},
{
node: {
__typename: 'User',
id: '5a6cf127c2b20834f6551482',
email: 'Robel.Ansel@yahoo.com',
},
cursor: 'Y3Vyc29yMw==',
},
],
pageInfo: {
endCursor: 'Y3Vyc29yMw==',
hasNextPage: false,
},
},
},
},
}
const { normalize } = new GraphQLNormalizr({
useConnections: true,
})
normalize(response)
// =>
// {
// users: {
// '5a6efb94b0e8c36f99fba013': {
// id: '5a6efb94b0e8c36f99fba013',
// email: 'Lloyd.Nikolaus@yahoo.com',
// friends: ['5a6cf127c2b20834f6551481', '5a6cf127c2b20834f6551482'],
// },
// '5a6cf127c2b20834f6551481': {
// id: '5a6cf127c2b20834f6551481',
// email: 'Madisen_Braun@hotmail.com',
// },
// '5a6cf127c2b20834f6551482': {
// id: '5a6cf127c2b20834f6551482',
// email: 'Robel.Ansel@yahoo.com',
// },
// },
// }
Object
By default the entity name will be the plural form of the type name, converted to camel case, (PrimaryAddress
type will be stored under the primaryAddresses
key). Use this option to provide specific entity names for some/all Types, or try the plural and casing options to derive the entity names.
const response = {
data: {
findUser: {
__typename: 'User',
id: '5a6efb94b0e8c36f99fba013',
email: 'Lloyd.Nikolaus@yahoo.com',
},
},
}
const { normalize } = new GraphQLNormalizr({
typeMap: { User: 'accounts' },
})
normalize(response)
// =>
// {
// accounts: {
// '5a6efb94b0e8c36f99fba013' : {
// id: '5a6efb94b0e8c36f99fba013',
// email: 'Lloyd.Nikolaus@yahoo.com'
// }
// }
// }
Object
Prevent normalization of specified fields
const response = {
data: {
allUsers: [
{
__typename: 'User',
id: '5a6efb94b0e8c36f99fba013',
email: 'Lloyd.Nikolaus@yahoo.com',
preferences: null
posts: [
{
__typename: 'BlogPost',
id: '5a6cf127c2b20834f6551484',
likes: 10,
title: 'Sunt ut aut',
tags: {},
}
]
},
{
__typename: 'User',
id: '5a6efb94b0e8c36f99fba013',
email: 'Anna.Klaus@gmail.com',
preferences: { foo: 'apple', bar: 1, baz: { a: 'b' }, quux: null, }
posts: [
{
__typename: 'BlogPost',
id: '5a6cf127c2b20834f6551485',
likes: 23,
title: 'Nesciunt esse',
tags: [],
}
]
},
],
},
}
Normalize the data excluding the preferences
field on users
and the tags
field on blogPosts
:
// using destructuring to get the `normalize` method of the client
const { normalize } = new GraphQLNormalizr({ ignore: { users: [ 'preferences' ], blogPosts: [ 'tags' ] } })
normalize(response)
// =>
// {
// users: {
// '5a6efb94b0e8c36f99fba013': {,
// id: '5a6efb94b0e8c36f99fba013',
// email: 'Lloyd.Nikolaus@yahoo.com',
// preferences: null
// },
// '5a6efb94b0e8c36f99fba013': {
// id: '5a6efb94b0e8c36f99fba013',
// email: 'Anna.Klaus@gmail.com',
// preferences: { foo: 'apple', bar: 1, baz: { a: 'b' }, quux: null, }
// },
// },
// blogPosts: {
// '5a6cf127c2b20834f6551484': {
// id: '5a6cf127c2b20834f6551484',
// likes: 10,
// title: 'Sunt ut aut',
// tags: {},
// },
// '5a6cf127c2b20834f6551485': {
// id: '5a6cf127c2b20834f6551485',
// likes: 23,
// title: 'Nesciunt esse',
// tags: [],
// },
// }
// }
Boolean
Default is true
. Set this to false
if you don't want to pluralize entity names. Considering the previous response example:
const { normalize } = new GraphQLNormalizr({
plural: false,
})
normalize(response)
// =>
// {
// user: {
// '5a6efb94b0e8c36f99fba013' : {
// id: '5a6efb94b0e8c36f99fba013',
// email: 'Lloyd.Nikolaus@yahoo.com'
// }
// }
// }
'lower'|'upper'|'camel'|'pascal'|'snake'|'kebab'
You can also specify the preferred casing for entity names. Again, consider the above response example.
// casing: 'lower'
// User => user
// casing: 'upper'
// User => USER
// casing: 'camel'
// PrimaryAddress => primaryAddress
// casing: 'pascal'
// PrimaryAddress => PrimaryAddress
// casing: 'snake'
// PrimaryAddress => primary_address
// casing: 'kebab'
// PrimaryAddress => primary-address
Combine plural
and casing
options to get the desired entity names
Boolean
Default is false
. All the data is stored in key/value pairs, for easy access. If you want to use arrays, for whatever reason, set this to true
For the same response object in our previous example:
const { normalize } = new GraphQLNormalizr({
lists: true,
})
normalize(response)
// =>
// {
// users: [
// {
// id: '5a6efb94b0e8c36f99fba013',
// email: 'Lloyd.Nikolaus@yahoo.com'
// }
// ]
// }
Boolean
Default is false
. The normalized data will not contain the __typename
field. Set this to true
if you need to persist them.
const { normalize } = new GraphQLNormalizr({
typenames: true,
})
normalize(response)
// =>
// {
// users: {
// '5a6efb94b0e8c36f99fba013' : {
// __typename: 'User',
// id: '5a6efb94b0e8c36f99fba013',
// email: 'Lloyd.Nikolaus@yahoo.com'
// }
// }
// }
Boolean
Default is false
. Enables explicit type pointers - instead of an array of only identifiers and having to figure out which collection they point to, it will return objects containing the identifier as well as the collection name. Works especially well with Union types and Interfaces.
const { normalize } = new GraphQLNormalizr({
typePointers: true,
})
// ['5a6cf127c2b20834f655148a', '5a6cf127c2b20834f655148b', '5a6cf127c2b20834f655148c']
users: [
{
_id: '5a6cf127c2b20834f655148a', // '_id' or the specified key
collection: 'members', // type Member
},
{
_id: '5a6cf127c2b20834f655148b', // '_id' or the specified key
collection: 'authors', // type Author
},
{
_id: '5a6cf127c2b20834f655148c', // '_id' or the specified key
collection: 'members', // type Member
},
],
Boolean
Default is false
. The normalize method is pretty fast by itself, it does a single iteration and associates the values only for each response node and not for all the fields. Enable this if you think you'd be normalizing the same response multiple times, like when you're polling for data and it may not have changed.
const { normalize } = new GraphQLNormalizr({
caching: true,
})
const normalized = normalize(response)
const cached = normalize(response)
cached === normalized // => true
Turns a GraphQL source into a Document and adds the required fields where necessary.
// ...
import { GraphQLNormalizr } from 'graphql-normalizr'
const source = `{
allUsers {
email
posts {
title
comments {
message
}
}
}
}`
const { parse } = new GraphQLNormalizr()
const query = parse(source) // will add `id` and `__typename` fields to all the nodes
// We can use the print method from `graphql` to see/use the updated source
const { print } = require('graphql')
print(query)
// =>
// `{
// allUsers {
// __typename
// id
// email
// posts {
// __typename
// id
// comments {
// __typename
// id
// message
// }
// }
// }
// }`
// ...
If you only have access to the Document, you can use the print method from graphql
to get the source and parse it. But that may be expensive and you shouldn't have to print a document just to parse it again. addRequiredFields
will add the id
and __typename
fields to that document, without the need of extracting its source.
// ...
import { GraphQLNormalizr } from 'graphql-normalizr'
import { allUsersQuery } from './queries'
const { addRequiredFields } = new GraphQLNormalizr()
const query = addRequiredFields(allUsersQuery)
// ...
The following is a full example where we use apollo-fetch to execute a query and then normalize it with GraphQLNormalizr
const { GraphQLNormalizr } = require('graphql-normalizr')
const { createApolloFetch } = require('apollo-fetch')
const uri = 'http://localhost:8080/graphql'
const fetch = createApolloFetch({ uri })
const source = `
query {
allUsers {
...userFields
}
}
fragment userFields on User {
email
posts {
title
comments {
message
}
}
}
`
const { normalize, parse } = new GraphQLNormalizr()
const query = parse(source)
fetch({ query }).then(response => {
const normalized = normalize(response)
// persist the normalized data to our app state.
}).catch(...)
from v1.x to v2.x
There aren't many breaking changes between v1.x and v2.x. In fact, there's only one and it's about how Type names get converted into entity names.
With the default configuration, v1.x will transfrom a PrimaryAddress
type name into primaryaddresses
entity name. With 2.x, the default configuiration will transform PrimaryAddress
to primaryAddresses
. The only difference is that now, it changes to camelcase instead of lowercase
If you don't want to change your code everywhere you are accessing entities, you can configure the way GraphQLNormalizr makes the transformation with the plural and casing options:
const { normalize } = new GraphQLNormalizr({
plural: true, // true is the default value, so you can omit this
casing: 'lower',
})
// this above configuration will change `PrimaryAddress` to `primaryaddresses`