mobx-task
Takes the suck out of managing state for async functions in MobX.
Table of Contents
- Installation
- What is it?
- Full example with classes and decorators
- Full example with plain observables
- How does it work?
- Task Groups
- API documentation
- Gotchas
- Author
Installation
npm install --save mobx-task
What is it?
mobx-task
removes the boilerplate of maintaining loading and error state of async functions in MobX.
Your code before:
class TodoStore {
@observable fetchTodosRunning = true
@observable fetchTodosError
async fetchTodos () {
try {
runInAction(() => {
this.fetchTodosRunning = true
})
// ...
await fetch('/todos')
} catch (err) {
runInAction(() => {
this.fetchTodosError = err
})
throw err
} finally {
runInAction(() => {
this.fetchTodosRunning = false
})
}
}
}
Your code with mobx-task
import { task } from 'mobx-task'
class TodoStore {
@task async fetchTodos () {
await fetch('/todos')
}
}
Full example with classes and decorators
import { observable, action } from 'mobx'
import { task } from 'mobx-task'
import React from 'react'
import { observer } from 'mobx-react'
class TodoStore {
@observable todos = []
@task async fetchTodos () {
await fetch('/todos')
.then(r => r.json())
.then(action(todos => this.todos.replace(todos)))
}
}
const store = new TodoStore()
// Start the task.
store.fetchTodos()
// and reload every 3 seconds, just cause..
setInterval(() => {
store.fetchTodos()
}, 3000)
const App = observer(() => {
return (
<div>
{store.fetchTodos.match({
pending: () => <div>Loading, please wait..</div>,
rejected: (err) => <div>Error: {err.message}</div>,
resolved: () => (
<ul>
{store.todos.map(todo =>
<div>{todo.text}</div>
)}
</ul>
)
})}
</div>
)
})
Full example with plain observables
import { observable, action } from 'mobx'
import { task } from 'mobx-task'
import React from 'react'
import { observer } from 'mobx-react'
const store = observable({
todos: [],
fetchTodos: task(async () => {
await fetch('/todos')
.then(r => r.json())
.then(action(todos => store.todos.replace(todos)))
})
})
// Start the task.
store.fetchTodos()
// and reload every 3 seconds, just cause..
setInterval(() => {
store.fetchTodos()
}, 3000)
const App = observer(() => {
return (
<div>
{store.fetchTodos.match({
pending: () => <div>Loading, please wait..</div>,
rejected: (err) => <div>Error: {err.message}</div>,
resolved: () => (
<ul>
{store.todos.map(todo =>
<div>{todo.text}</div>
)}
</ul>
)
})}
</div>
)
})
How does it work?
mobx-task
wraps the given function in another function which
does the state maintenance for you using MobX observables and computeds.
It also exposes the state on the function.
const func = task(() => 42)
// The default state is `pending`.
console.log(func.state) // pending
console.log(func.pending) // true
// Tasks are always async.
func().then((result) => {
console.log(func.state) // resolved
console.log(func.resolved) // true
console.log(func.pending) // false
console.log(result) // 42
// The latest result is also stored.
console.log(func.result) // 42
})
It also maintains error state.
const func = task(() => {
throw new Error('Nope')
})
func().catch(err => {
console.log(func.state) // rejected
console.log(func.rejected) // true
console.log(err) // Error('Nope')
console.log(func.error) // Error('Nope')
})
And it's fully reactive.
import { autorun } from 'mobx'
const func = task(async () => {
return await fetch('/api/todos').then(r => r.json())
})
autorun(() => {
// Very useful for functional goodness (like React components)
const message = func.match({
pending: () => 'Loading todos...',
rejected: (err) => `Error: ${err.message}`,
resolved: (todos) => `Got ${todos.length} todos`
})
console.log(message)
})
// start!
func().then(todos => { /*...*/ })
Task Groups
since mobx-task v2.0.0
A TaskGroup
is useful when you want to track pending, resolved and rejected state for multiple tasks but treat them as one.
Under the hood, a TaskGroup
reacts to the start of any of the tasks (when pending
flips to true
), tracks the latest started task, and proxies all getters
to it. The first pending
task (or the first task in the input array, if none are pending
) is used as the initial task to proxy to.
IMPORTANT: Running the tasks concurrently will lead to wonky results. The intended use is for
tracking pending
, resolved
and rejected
states of the last run task. You should prevent your users from
concurrently running tasks in the group.
import { task, TaskGroup } from 'mobx-task'
const toggleTodo = task.resolved((id) => api.toggleTodo(id))
const deleteTodo = task.resolved((id) => { throw new Error('deleting todos is for quitters') })
const actions = TaskGroup([
toggleTodo,
deleteTodo
])
autorun(() => {
const whatsGoingOn = actions.match({
pending: () => 'Todo is being worked on',
resolved: () => 'Todo is ready to be worked on',
rejected: (err) => `Something failed on the todo: ${err.message}`
})
console.log(whatsGoingOn)
})
// initial log from autorun setup
// <- Todo is ready to be worked on
await toggleTodo('some-id')
// <- Todo is being worked on
// ...
// <- Todo is ready to be worked on
await deleteTodo('some-id')
// <- Todo is being worked on
// ...
// <- Something failed on the todo: deleting todos is for quitters
API documentation
There's only a single exported member; task
.
ES6:
import { task } from 'mobx-task'
CommonJS:
const { task } = require('mobx-task')
task
factory
The The top-level task
creates a new task function and initializes it's state.
const myAwesomeFunc = task(async () => {
return await doAsyncWork()
})
// Initial state is `pending`
console.log(myAwesomeFunc.state) // "pending"
Let's try to run it
const promise = myAwesomeFunc()
console.log(myAwesomeFunc.state) // "pending"
promise.then((result) => {
console.log('nice', result)
console.log(myAwesomeFunc.state) // "resolved"
})
Parameters:
fn
- the function to wrap in a task.opts
- options object. All options are optional.opts.state
- the initial state, default is'pending'
.opts.error
- initial error object to set.opts.result
- initial result to set.opts.swallow
- iftrue
, does not throw errors after catching them.
Additionally, the top-level task
export has shortcuts for the opts.state
option (except pending, since its the default).
task.resolved(func, opts)
task.rejected(func, opts)
For example:
const func = task.resolved(() => 42)
console.log(func.state) // resolved
Is the same as doing:
const func = task(() => 42, { state: 'resolved' })
console.log(func.state) // resolved
As a decorator
The task
function also works as a decorator.
Note: you need to add
babel-plugin-transform-decorators-legacy
to your babel config for this to work.
Example:
class Test {
@task async load () {
}
// shortcuts, too
@task.resolved async save () {
}
// with options
@task({ swallow: true }) async dontCareIfIThrow() {
}
// options for shortcuts
@task.rejected({ error: 'too dangerous lol' }) async whyEvenBother () {
}
}
task
itself
The The thing that task()
returns is the wrapped function including all that extra goodness.
state
An observable string maintained while running the task.
Possible values:
"pending"
- waiting to complete or didn't start yet (default)"resolved"
- done"rejected"
- failed
pending
, resolved
, rejected
Computed shorthands for state
. E.g. pending = state === 'pending'
result
Set after the task completes. If the task fails, it is set to undefined
.
args
An array of arguments that were used when the task function was invoked last.
error
Set if the task fails. If the task succeeds, it is set to undefined
.
match()
Utility for pattern matching on the state.
Example:
const func = task((arg1, arg2, arg3, ..asManyAsYouWant) => 42)
const result = func.match({
pending: (arg1, arg2, arg3, ...asManyAsYouWant) => 'working on it',
rejected: (err) => 'failed: ' + err.message,
resolved: (answer) => `The answer to the universe and everything: ${answer}`
})
wrap()
Used to wrap the task in another function while preserving access to the state - aka. Higher Order Functions.
Returns the new function, does not modify the original function.
// Some higher-order-function...
const addLogging = function (inner) {
return function wrapped () {
console.log('Started')
return inner.apply(this, arguments).then(result => {
console.log('Done!')
return result
})
}
}
const func = task(() => 42)
const funcWithLogging = func.wrap(addLogging)
setState()
Lets you set the internal state at any time for whatever reason you may have. Used internally as well.
Example:
const func = task(() => 42)
func.setState({ state: 'resolved', result: 1337 })
console.log(func.state) // 'resolved'
console.log(func.resolved) // true
console.log(func.result) // 1337
bind()
The wrapped function patches bind()
so the bound function contains the task state, too.
Other than that it functions exactly like Function.prototype.bind
.
const obj = {
value: 42,
doStuff: task(() => this.value)
}
const bound = obj.doStuff.bind(obj)
bound()
console.log(bound.pending) // true
reset()
Resets the state to what it was when the task was initialized.
This means if you use const t = task.resolved(fn)
, calling t.reset()
will set the state to resolved
.
TaskGroup
Creates a TaskGroup
. Takes an array of tasks to track. Implements the readable parts of the Task
.
Uses the first task in the array as the proxy target.
import { task, TaskGroup } from 'mobx-task'
const task1 = task(() => 42)
const task2 = task(() => 1337)
const group = TaskGroup([
task1,
task2
])
console.log(group.state)
console.log(group.resolved)
console.log(group.result)
Gotchas
Wrapping the task function
It's important to remember that if you wrap the task in something else, you will loose the state.
Bad:
import once from 'lodash/once'
const func = task(() => 42)
const funcOnce = once(func)
console.log(funcOnce.pending) // undefined
This is nothing special, but it's a common gotcha when you like to compose your functions. We can make this work though,
by using .wrap(fn => once(fn))
. See the wrap()
documentation.
Good:
import once from 'lodash/once'
const func = task(() => 42)
const funcOnce = func.wrap(once)
console.log(funcOnce.pending) // true
Using the decorator on React Components
Using the @task
decorator on React components is absolutely a valid use case, but if you use React Hot Loader or
any HMR technology that patches functions on components, you will loose access to the task state.
A workaround is to not use the decorator, but a property initializer:
class Awesome extends React.Component {
fetchTodos = task(() => {
return fetch('/api/todos')
})
render () {
return (
<div>
{this.fetchTodos.match(...)}
</div>
)
}
}
autobind-decorator
Using the decorator with Because of the way the autobind-decorator
class decorator works, it won't pick up any @task
-decorated
class methods because @task
rewrites descriptor.value
to descriptor.get
which autobind-decorator
does
not look for. This is due to the fact that autobind-decorator
does not (and should not) evaluate getters.
You can either bind the tasks in the constructor, use field initializers, or apply the @autobind
method decorator before the @task
decorator. @task @autobind method() {}
is the correct order.
import autobind from 'autobind-decorator'
@autobind
class Store {
value = 42
// Using decorator
@task boo () {
return this.value
}
// Using field initializer
woo = task(() => {
return this.value
})
// Decorator with autobind applied first
@task @autobind woohoo () {
return this.value
}
}
// Nay
const store = new Store()
store.boo() // 42
const boo = store.boo
boo() // Error: cannot read property "value" of undefined
// Yay
store.woo() // 42
const woo = store.woo
woo() // 42
const woohoo = store.woohoo
woohoo() // 42
Alternatively, use this.boo = this.boo.bind(this)
in the constructor.
typescript
Using with Best way to work with typescript is to install @types/mobx-task
. Definitions covers most use cases. The tricky part is decorators because they are not able to change the type of the decorated target. You will have to do type assertion or use plain observables.
npm install --save-dev @types/mobx-task
Example:
class Test {
@task taskClassMethod(arg1: string, arg2: number) {
let result: boolean
...
return result
}
@task assertTypeHere = <Task<boolean, [string, number]>>((arg1: string, arg2: number) => {
let result: boolean
...
return result
})
@task assertTypeHereWithAs = ((arg1: string, arg2: number) => {
let result: boolean
...
return result
}) as Task<boolean, [string, number]>
}
const test = new Test()
// dont care about task methods, props and return value and type
const doSomething = async () => {
await test.taskClassMethod('a', 1)
...
}
// want to use task props and returned promise
(test.taskClassMethod as Task)("one", 2).then(...) // Task<any, any[]>
const {result} = <Task<Result>>test.taskClassMethod // Task<Result, any[]>
const {args} = test.taskClassMethod as Task<void, [string]>
Author
Jeff Hansen - @Jeffijoe