Fuse
Fuse is a layer that sits infront of your data store and merges multiple incomplete data models into a single complete model.
For example, if you're making lots of requests via GraphQL, each request will return a different subset of fields on a particular model. Most architectures don't support merging multiple objects with a shared key into a single object. Fuse will do this for you automatically.
All you need to do is define your schema relations and call fuse.handle()
whenever you recieve new data. You can hook your data store into Fuse, so after the handle method is called, all mutations are sent to your store. Fuse will also only trigger a store mutation when there's a state change, keeping things efficient.
Installation
yarn add fuse-state
Introduction
Let's say your data model consists of User
and Post
. Users can have multiple posts and friends, but only one best friend. Posts can be liked by multiple users, but can only be in reply to one post.
Here is an example of what that kind of data would look like in JSON format:
{
"id": "1",
"name": "William",
"posts": [{
"id": "1",
"text": "Hello World",
"likedBy": [{
"id": "2",
"name": "Kristine"
}]
}],
"friends": [{
"id": "2",
"name": "Kristine",
"posts": [{
"id": "2",
"text": "Hello William",
"inReplyTo": {
"id": "1",
"viewCount": 10
}
}]
}, {
"id": "3",
"name": "Harry",
"bestFriend": {
"id": "4",
"name": "Bob"
}
}],
"bestFriend": {
"id": "2",
"avatarUrl": "<some_url>"
}
}
With GraphQL, you can query fields on objects without any backend changes, so in one request you might be asking for a posts viewCount
, but on another you might be asking for a posts likedBy
. It's a little all over the place.
For example, on Kristine's post at .friends[0].posts[0]
, the inReplyTo
field shows that the post from William with id:1
has 10 views, but this isn't surfaced on the object at .posts[0]
. There are countless other examples (e.g. .bestFriend
exposing avatarUrl
).
So let's try using Fuse. Creating a schema is straight forward. We just define our model on the root object, and which fields contain which models (and if they're an object or an array). Here is a Fuse schema for our data model:
const fuse = new Fuse({
schema: b => ({
user: {
posts: b.array('post'),
friends: b.array('user'),
bestFriend: b.object('user')
},
post: {
likedBy: b.array('user'),
inReplyTo: b.object('post')
}
})
})
Once this is all set up, we can use fuse.handle
to give it any model we've defined (e.g. user
or post
). user
in this case is the JSON object that was specified above:
fuse.handle({ user })
Fuse will automatically build a new state for us, which we can access with fuse.state
:
{
"users": {
"1": {
"id": "1",
"name": "William",
"posts": [
"1"
],
"friends": [
"2",
"3"
],
"bestFriend": "2"
},
"2": {
"id": "2",
"name": "Kristine",
"posts": [
"2"
],
"avatarUrl": "<some_url>"
},
"3": {
"id": "3",
"name": "Harry",
"bestFriend": "4"
},
"4": {
"id": "4",
"name": "Bob"
}
},
"posts": {
"1": {
"id": "1",
"text": "Hello World",
"likedBy": [
"2"
],
"viewCount": 10
},
"2": {
"id": "2",
"text": "Hello William",
"inReplyTo": "1"
}
}
}
This may look daunting at first, but it is fairly logical how everything is laid out.
First, we can access an object using its model name and id by selecting state.<model>.<id>
.
Second, relations on an object have been replaced with its id. So, let's find user 1's best friend:
const user = state.users['1']
const bestFriendId = user.bestFriend
const bestFriend = state.users[bestFriendId]
This is beneficial as we're accessing a single source of truth for a model, it can't exist in two places at once with potentially different fields. Fuse merges all the instances of a model it can find using the schema. So, our posts.1
now has both our likedBy
and viewCount
fields in the same object.
Usage
import Fuse from 'fuse-state'
const fuse = new Fuse({
schema: b => ({
book: {
author: b.object('author')
},
author: {
books: b.array('book')
}
}),
handlerFns: {
book: book => {
console.log(`Book with id ${book.id} updated`)
store.dispatch(updateBook(book))
},
books: books => {
console.log(`${Object.keys(books).length} books updated`)
store.dispatch(updateBooks(books))
}
}
})
fuse.handle({
book: {
id: 1,
name: 'My Book',
author: {
id: 1,
name: 'Darn Fish',
books: [{
id: 1,
year: 2022
}, {
id: 2,
name: 'My Book: The Sequel'
}]
}
},
books: [{
id: 2,
starRating: 5
}]
})
{
"books": {
"1": {
"id": 1,
"name": "My Book",
"author": 1,
"year": 2022
},
"2": {
"id": 2,
"name": "My Book: The Sequel",
"starRating": 5
}
},
"authors": {
"1": {
"id": 1,
"name": "Darn Fish",
"books": [
1,
2
]
}
}
}
fuse.handle({
author: {
id: 1,
age: 20,
books: [{
id: 2,
year: 2023,
author: {
id: 1
}
}]
}
})
{
"books": {
"1": {
"id": 1,
"name": "My Book",
"author": 1,
"year": 2022
},
"2": {
"id": 2,
"name": "My Book: The Sequel",
"starRating": 5,
"year": 2023,
"author": 1
}
},
"authors": {
"1": {
"id": 1,
"name": "Darn Fish",
"books": [
1,
2
],
"age": 20
}
}
}
Advanced Usage
Fuse can handle very complex schemas, with deeply nested objects. For example:
const fuse = new Fuse({
schema: b => ({
bankAccount: {
balanceHistory: {
amount: {
currency: b.object('currency')
},
convertedAmounts: {
currency: b.object('currency')
}
}
},
currency: {}
})
})
fuse.handle({
bankAccount: {
id: 1,
balanceHistory: [{
amount: {
amount: 100,
currency: {
id: 'USD',
name: 'United States Dollar'
}
},
convertedAmounts: [{
amount: 100,
currency: {
id: 'GBP',
name: 'British Pound Sterling'
}
}, {
amount: 100,
currency: {
id: 'EUR',
name: 'Euro'
}
}]
}]
}
})
{
"bankAccounts": {
"1": {
"id": 1,
"balanceHistory": [
{
"amount": {
"amount": 100,
"currency": "USD"
},
"convertedAmounts": [
{
"amount": 100,
"currency": "GBP"
},
{
"amount": 100,
"currency": "EUR"
}
]
}
]
}
},
"currencies": {
"USD": {
"id": "USD",
"name": "United States Dollar"
},
"GBP": {
"id": "GBP",
"name": "British Pound Sterling"
},
"EUR": {
"id": "EUR",
"name": "Euro"
}
}
}
License
MIT