effects
This package is inspired by Sagas and gives you advanced effect management solutions.
included in @reatom/framework
First of all you should know that some effects and async (reatom/async + reatom/hooks) logic uses AbortController under the hood and if some of the controller aborted all nested effects will aborted too! It is a powerful feature for managing async logic which allows you to easily write concurrent logic, like with redux-saga or rxjs, but with the simpler native API.
Before we start, you could find a lot of useful helpers to manage aborts in reatom/utils
The differences between Redux-Saga and Reatom.
- Sagas
takeis liketake+await. - Sagas
takeMaybe- is liketakeWITHOUTawait. - Sagas
takeEvery- is likeanAtom.onChange/anAction.onCall. - Sagas
takeLatest- is likeanAtom.onChange/anAction.onCall+concurrent. - Sagas
takeLeading- is likeanAtom.onChange+reatomAsync().pipe(withAbort({ strategy: 'first-in-win' })). - Sagas
callis a regular function call with a context +await. - Sagas
forkis a regular function call with a context WITHOUTawait. - Sagas
spawnisspawn - Sagas
join- is justawaitin Reatom. - Sagas
cancelis likegetTopController(ctx.cause)?.abort(). - Sagas
cancelled- is likeonCtxAbort.
API
concurrent
This is the basic, useful API for performing concurrent async logic. Wrap your function with the concurrent decorator, and all scheduled tasks of the passed ctx will throw the abort error when a new request appears.
Main use case for the concurrent API is onChange handling. Just wrap your function to always get only fresh results, no matter how often the changes occur.
Here, when someAtom changes for the first time, the hook will be called and start fetching. If someAtom changes during the fetch execution, the ctx.schedule of the previous (first) call will throw an AbortError, and the new fetching will start.
import { concurrent } from '@reatom/effects'
someAtom.onChange( concurrent(async (ctx, some) => { const other = await ctx.schedule(() => api.getOther(some)) otherAtom(ctx, other) }),)Another example is how easily you could implement the “debounce” pattern with additional logic. Here is a comparison of the classic “debounce” decorator from “lodash” or any other utility library with the concurrent API. Each of the three examples has the same behavior for the debounce and concurrent examples.
You can see that each new logic addition forces a lot of changes for code with the simple debounce decorator and takes a really small amount of changes for code with the concurrent decorator.
Base debounce.
const onChangeDebounce = debounce((ctx, event) => { inputAtom(ctx, event.currentTarget.value)}, 500)
const onChangeConcurrent = concurrent(async (ctx, event) => { await ctx.schedule(() => sleep(500)) inputAtom(ctx, event.currentTarget.value)})Debounce after some mappings
const _onChangeDebounce = debounce((ctx, value) => { inputAtom(ctx, value)}, 500)const onChangeDebounce = (ctx, event) => { _onChangeDebounce(ctx, event.currentTarget.value)}
const onChangeConcurrent = concurrent(async (ctx, event) => { const { value } = event.currentTarget await ctx.schedule(() => sleep(500)) inputAtom(ctx, value)})Debounce with a condition.
const _onChange = (ctx, value) => { inputAtom(ctx, value)}const _onChangeDebounce = debounce(_onChange, 500)const onChangeDebounce = (ctx, event) => { const { value } = event.currentTarget if (Math.random() > 0.5) _onChange(ctx, value) else handleDebounceChange(ctx, value)}
const onChangeConcurrent = concurrent(async (ctx, event) => { const { value } = event.currentTarget if (Math.random() > 0.5) await ctx.schedule(() => sleep(500)) inputAtom(ctx, value)})take
This is the simplest and most powerful API that allows you to wait for an atom update, which is useful for describing certain procedures. It is a shortcut for subscribing to the atom and unsubscribing after the first update. take respects the main Reatom abort context and will throw AbortError when the abort occurs. This allows you to describe redux-saga-like procedural logic in synchronous code style with native async/await.
import { action } from '@reatom/core'import { take } from '@reatom/effects'
export const validateBeforeSubmit = action(async (ctx) => { let errors = validate(ctx.get(formDataAtom))
while (Object.keys(errors).length) { formDataAtom.errorsAtom(ctx, errors) // wait any field change await take(ctx, formDataAtom) // recheck validation errors = validate(ctx.get(formDataAtom)) }})You can also await actions!
import { take } from '@reatom/effects'import { onConnect } from '@reatom/hooks'import { historyAtom } from '@reatom/npm-history'import { confirmModalAtom } from '~/features/modal'
// some model logic, doesn't matterexport const formAtom = reatomForm(/* ... */)
onConnect(formAtom, (ctx) => { // "history" docs: https://github.com/remix-run/history/blob/main/docs/blocking-transitions.md const unblock = historyAtom.block(ctx, async ({ retry }) => { if (!ctx.get(formAtom).isSubmitted && !ctx.get(confirmModalAtom).opened) { confirmModalAtom.open(ctx, 'Are you sure want to leave?')
const confirmed = await take(ctx, confirmModalAtom.close)
if (confirmed) { unblock() retry() } } })})take filter
You can pass the third argument to map the update to the required format.
const input = await take(ctx, onChange, (ctx, event) => event.target.value)More than that, you can filter unneeded updates by returning the skip mark from the first argument of your callback.
const input = await take(ctx, onChange, (ctx, event, skip) => { const { value } = event.target return value.length < 6 ? skip : value})The cool feature of this skip mark is that it helps TypeScript understand the correct type of the returned value, which is hard to achieve with the extra “filter” function. If you have a union type, you could receive the needed data with the correct type easily. It just works.
const someRequest = reatomRequest<{ data: Data } | { error: string }>()// type-safe destructuringconst { data } = await take(ctx, someRequest, (ctx, payload, skip) => 'error' in payload ? skip : payload,)takeNested
Allow you to wait all dependent effects, event if they was called in the nested async effect or by spawn.
For example, we have a routing logic for SSR.
import { historyAtom } from '@reatom/npm-history'
historyAtom.locationAtom.onChange((ctx, location) => { if (location.pathname === '/some') { fetchSomeData(ctx, location.search) }})How to track fetchSomeData call? We could use takeNested for this.
// SSR prerenderawait takeNested(ctx, (trackedCtx) => { historyAtom.push(trackedCtx, req.url)})render()You could pass an arguments in the rest params of takeNested function to pass it to the effect.
await takeNested(ctx, historyAtom.push, req.url)render()onCtxAbort
Handle an abort signal from the cause stack. For example, if you want to separate a task from the body of the concurrent handler, you can do it without explicit abort management; all tasks are carried out on top of ctx.
import { action } from '@reatom/core'import { reatomAsync, withAbort } from '@reatom/async'import { onCtxAbort } from '@reatom/effects'
const doLongImportantAsyncWork = action((ctx) => ctx.schedule(() => { const timeoutId = setTimeout(() => { /* ... */ }) onCtxAbort(ctx, () => clearTimeout(timeoutId)) }),)
export const handleImportantWork = reatomAsync((ctx) => { /* ... */ doLongImportantAsyncWork(ctx) /* ... */}).pipe(withAbort())getTopController
This is a simple util to find an abort controller on top of your cause stack. For example, it is useful to stop some async operation inside a regular actions, which are probably called from a concurrent context.
import { action } from '@reatom/core'import { getTopController } from '@reatom/effects'import { throwAbort } from '@reatom/utils'import { onConnect } from '@reatom/hooks'
const doSome = action(async (ctx) => { const data = await ctx.schedule(() => fetchData())
if (!data) throwAbort('nullable data', getTopController(ctx.cause))
// ... perform data}, 'doSome')spawn
This utility allow you to start a function with context which will NOT follow an abort of the cause.
For example, you want to start a fetch when Atom gets a connection, but don’t want to abort the fetch when the connection is lost. This is because you want to persist the results.
import { spawn } from '@reatom/effects'import { onConnect } from '@reatom/hooks'
onConnect(someAtom, (ctx) => { spawn(ctx, async (spawnCtx) => { const some = await api.getSome(spawnCtx) someAtom(spawnCtx, some) })})