next
>=12.0.0
effector
>=22.2.0
(effector Halley with many syntax changes)effector-react
>=22.3.3
(has important bugfix for Next.js)
Due to some Next.js bundling issues, it's currently recommended to install nextjs-effector
locally:
-
Copy
library
folder contents into your project, for example intosrc/nextjs-effector
-
Create the alias using tsconfig.json:
{ "compilerOptions": { "baseUrl": "./", "paths": { "nextjs-effector": ["./src/nextjs-effector"], "nextjs-effector/*": ["./src/nextjs-effector/*"] } } }
At first, add effector/babel-plugin
to your .babelrc
:
{
"presets": ["next/babel"],
"plugins": [["effector/babel-plugin", { "reactSsr": true }]]
}
By doing that, all our Effector units will be created with unique sid
constant, so we can safely serialize them for sending to the client.
The reactSsr
option is used to replace all effector-react
imports with effector-react/scope
version to ensure that useStore
, useEvent
, and the other hooks use the scope that is passed using Provider
.
Also, check your effector
and effector-react
versions: they should match the requirements.
Finally, enhance your App
:
/* pages/_app.tsx */
import App from 'next/app'
import { withEffector } from 'nextjs-effector'
// Passing effector-react exports is required to access the Scope
export default withEffector(App)
After that, the App
will be wrapped in Effector's Scope Provider. withEffector
function uses the smart Scope management logic under the hood, so you can focus on the writing a business logic without thinking about problems of integrating Effector into your Next.js application.
Basically, there are 2 types of data in any application:
- Shared - used in almost every page (translations, logged user info)
- Specific - used only in the relevant pages (post content on post page, group info on group page)
Usually, we want these conditions to be met:
- The shared data is loaded once in the application lifecycle (expect for manual update)
- The specific data is loaded on each navigation between the pages
- No shared events boilerplate on every page
Assume we have the following Effector events:
/* Needed everywhere */
export const loadAuthenticatedUser = createEvent()
export const loadTranslations = createEvent()
/* Needed only on the post page */
export const loadPostCategories = createEvent()
export const loadPostContent = createEvent()
Let's group them into shared and specific:
export const appStarted = createEvent()
export const postPageStarted = createEvent()
sample({
clock: appStarted,
target: [loadAuthenticatedUser, loadTranslations]
})
sample({
clock: postPageStarted,
target: [loadPostCategories, loadPostContent]
})
We want the appStarted
to be called once in application lifecycle, and the postPageStarted
to be called on requesting / navigating to Post page. nextjs-effector
library provides a 2-level GIP factory to cover this case:
export const createGIP = createGIPFactory({
// Will be called once
sharedEvents: [appStarted]
})
PostPage.getInitialProps = createGIP({
// Will be called on visiting PostPage
pageEvent: postPageStarted
})
Also, the library provides createGSSPFactory
for getServerSideProps
and createGSPFactory
for getStaticProps
. They have almost the same API, but both of them run sharedEvents
on each request / static page generation.
Although getServerSideProps
is the most modern approach, we strongly recommend to use getInitialProps
in most cases with Effector. It's easier to work with, and doesn't require executing the shared logic on each request, including navigation between pages.
/*
* 1. Create events
* GIP accepts the events with "PageContext" or "void" payload types
*/
export const appStarted = createEvent()
export const appStarted = createEvent<PageContext>()
export const pageStarted = createEvent()
export const pageStarted = createEvent<PageContext>()
export const pageStarted = createEvent<PageContext<Props, Params, Query>>()
/*
* 2. Create GIP factory
* The place depends on your architecture
*/
export const createGIP = createGIPFactory({
// Will be called once:
// - Server side on initial load
// - Client side on navigation (only if not called yet)
sharedEvents: [appStarted],
// Allows to specify shared events behavior
// When "false", the shared events run like pageEvent
runSharedOnce: true,
// Allows to customize server-side Scope creation process
// By default, the library just uses fork(), like below
// But you can fill the stores in scope with your values (cookies, for example)
createServerScope: () => fork()
})
/*
* 3. Create GIP
* Usually, it's done inside "pages" directory
*/
Page.getInitialProps = createGIP({
// Will be called on each page visit:
// - Server side on initial load
// - Client side on navigation (even if already called)
pageEvent: pageStarted,
// You can define your custom logic using "customize" function
// It's run after all events are settled, but before Scope serialization
// So, here you can safely call allSettled
async customize({ scope, context }) {
return { /* Props */ }
}
})
For every-day cases we recommend using getInitialProps
instead. But getServerSideProps
may be useful in some edge-cases like executing logic with heavy computations, or accessing the data available only on server-side.
/*
* 1. Create events
* GSSP accepts the events with "PageContext" or "void" payload types
*/
export const appStarted = createEvent()
export const appStarted = createEvent<PageContext>()
export const pageStarted = createEvent()
export const pageStarted = createEvent<PageContext>()
export const pageStarted = createEvent<PageContext<Props, Params, Query>>()
/*
* 2. Create GSSP factory
* The place depends on your architecture
*/
export const createGSSP = createGSSPFactory({
// Will be called on first request and on each page navigation (always on server side)
sharedEvents: [appStarted],
})
/*
* 3. Create GSSP
* Usually, it's done inside "pages" directory
*/
export const getServerSideProps = createGSSP({
// Will be called on each page navigation (always on server side)
// Always called after shared events
pageEvent: pageStarted,
// You can define your custom logic using "customize" function
// It's run after all events are settled, but before Scope serialization
// So, here you can safely call allSettled
customize({ scope, context }) {
return { /* GSSP Result */ }
}
})
Recommended for static pages.
/*
* 1. Create events
*/
export const appStarted = createEvent()
export const appStarted = createEvent<StaticPageContext>()
export const pageStarted = createEvent()
export const pageStarted = createEvent<StaticPageContext>()
export const pageStarted = createEvent<StaticPageContext<Props, Params>>()
/*
* 2. Create GSP factory
* The place depends on your architecture
*/
export const createGSP = createGSPFactory({
// Will be called on each page generation (always on server side)
sharedEvents: [appStarted],
})
/*
* 3. Create GSP
* Usually, it's done inside "pages" directory
*/
export const getStaticProps = createGSP({
// Will be called on each page generation (always on server side)
pageEvent: pageStarted,
// You can define your custom logic using "customize" function
// It's run after all events are settled, but before Scope serialization
// So, here you can safely call allSettled
customize({ scope, context }) {
return { /* GSP Result */ }
}
})
Executes the provided Event<void> | Event<PageContext>
on the client side.
The hook may be useful for the getStaticProps
cases - it allows to keep Next.js optimization and request some user-specific global data at the same time.
The second parameter is options to enhance the event using enhancePageEvent
.
Usage:
const Page: NextPage<Props> = () => {
usePageEvent(appStarted, { runOnce: true })
return <AboutPage />
}
export const getStaticProps: GetStaticProps<Props> = async () => { /* ... */ }
export default Page
Wraps your event and adds some logic to it.
The enhanced event can be safely used anywhere.
It doesn't cause any changes to the original event - you may use it just as before.
const enhancedEvent = enhancePageEvent(appStarted, {
// Works like the "runSharedOnce" option in GIP fabric, but for the single event
// This option applies to both client and server environments:
// If the enhanced event was called on the server side, it won't be called on the client side
runOnce: true
})
/*
* PageContext has the "env" field with "client" or "server" value,
* so you can determine the environment where the code is executed
*
* The library provides useful types and type-guards for these purposes
*/
import {
isClientPageContext,
isServerPageContext,
ClientPageContext,
ServerPageContext
} from 'nextjs-effector'
const pageStartedOnClient = createEvent<ClientPageContext>()
const pageStartedOnServer = createEvent<ServerPageContext>()
sample({
source: pageStarted,
filter: isClientPageContext,
target: pageStartedOnClient
})
sample({
source: pageStarted,
filter: isServerPageContext,
target: pageStartedOnServer
})
sample({
source: pageStartedOnServer,
fn: (context) => {
// You can access "req" and "res" on server side
const { req, res } = context
return req.cookie
}
})
/*
* GSP accepts the events with "StaticPageContext | void" payload
* Unlike PageContext, the StaticPageContext doesn't include query
* and some other properties
*/
export const appStarted = createEvent<StaticPageContext>()
export const pageStarted = createEvent<StaticPageContext>()
export const pageStarted = createEvent<StaticPageContext<Props, Params>>()
/*
* Also, the library exports some utility types
*/
type PageEvent<...> = Event<PageContext<...>>
type StaticPageEvent<...> = Event<StaticPageContext<...>>
type EmptyOrPageEvent<...> = PageEvent<...> | Event<void>
type EmptyOrStaticPageEvent<...> = StaticPageEvent<...> | Event<void>
You can use createServerScope
to set cookies before executing any logic:
export const createGIP = createGIPFactory({
sharedEvents: [appStarted],
createServerScope: (context) => {
return fork({
values: [
[$cookies, context.req?.headers.cookie ?? '']
],
})
},
})
Also, you can access req
object in effector logic by using isServerContext
import { isServerPageContext } from 'nextjs-effector'
sample({
source: appStarted,
filter: isServerPageContext,
fn: (context) => context.req.cookie,
target: $cookies
})
The place depends on your architecture. But one thing is certain - creating factories on each page is a really bad idea. They are designed to simplify and encapsulate the repeated logic parts.
For example, with Feature Sliced Design
you might consider creating layouts
layer, which can be used to create reusable page layouts and factories.
getServerSideProps
, unlike the getInitialProps
, is run only on server-side. The problem is that pageEvent
logic may depend on global shared data. So, we need either to run shared events on each request (as we do now), or get this global shared data in some other way, for example by sending it back from the client in a serialized form (sounds risky and hard).
Also, to check if shared events are need to be executed, we should either to ask it from the client, or persist this data on server. The both ways sound hard to implement.
That's why getInitialProps
is more recommended way to bind your Effector models to Page lifecycle. When navigating between pages, it runs on client side, so we can easily omit the app event execution.
You can create GIP / GSSP fabric without sharedEvents
, and define the flow manually:
const createGIP = createGIPFactory()
Page.getInitialProps = createGIP({
pageEvent: pageStarted
})
sample({
source: pageStarted,
target: appStarted
})
Also, you can use enhancePageEvent
to run specific events only once in the application lifecycle.
- Fork this repo
- Use the Regular flow
Please follow Conventions
The dev branch is main
- any developer changes is merged in there.
Also, there is a release/latest
branch. It always contains the actual source code for release published with latest
tag.
All changes is made using Pull Requests - push is forbidden. PR can be merged only after successfull test-and-build
workflow checks.
When PR is merged, release-drafter
workflow creates/updates a draft release. The changelog is built from the merged branch scope (feat
, fix
, etc) and PR title. When release is ready - we publish the draft.
Then, the release
workflow handles everything:
- It runs tests, builds a package, and publishes it
- It synchronizes released tag with
release/latest
branch
- Create feature branch
- Make changes in your feature branch and commit
- Create a Pull Request from your feature branch to
main
The PR is needed to test the code before pushing to release branch - If your PR contains breaking changes, don't forget to put a
BREAKING CHANGES
label - Merge the PR in
main
- All done! Now you have a drafted release - just publish it when ready
- Assume your prerelease tag is
beta
- Create
release/beta
branch - Create feature branch
- Make changes in your feature branch and commit
- Create a Pull Request from your feature branch to
release/beta
The PR is needed to test the code before pushing to release branch - Create Github release with tag like
v1.0.0-beta
, pointing torelease/beta
branch
For nextbeta
versions use semver build syntax:v1.0.0-beta+1
- After that, the
release
workflow will publish your package with thebeta
tag - When the
beta
version is ready to becomelatest
- create a Pull Request fromrelease/beta
tomain
branch - Continue from the Regular flow's #5 step
Feature branches:
- Should start with
feat/
,fix/
,docs/
,refactor/
, and etc., depending on the changes you want to propose (see pr-labeler.yml for a full list of scopes)
Commits:
- Should follow the Conventional Commits specification
Pull requests:
- Should have human-readable name, for example: "Add a TODO list feature"
- Should describe changes
- Should have correct labels