- π Blazing fast bundle times - multicore compilation, and a filesystem cache for fast rebuilds even after a restart.
- π¦ Out of the box support for JS, CSS, HTML, file assets, and more - no plugins to install.
- π Automatically transforms modules using Babel, PostCSS, and PostHTML when needed - even
node_modules
. - βοΈ Zero configuration code splitting using dynamic
import()
statements. - π₯ Built in support for hot module replacement
- π¨ Friendly error logging experience - syntax highlighted code frames help pinpoint the problem.
Below is the design document that was created before work on the implementation of Parcel 2 started and some sections are outdated. The actual (somewhat complete) documentation for Parcel 2 is available here: https://v2.parceljs.org/.
Before we get started, you'll need to install Node and Yarn (or npm) and create
a package.json
for your project if you haven't already.
yarn init
Then with Yarn you can install parcel
into your app:
yarn add --dev parcel@next
From there you just need to point Parcel at some of your entry files. Like if
you're building a website, an index.html
file:
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8" />
<title>My First Parcel App</title>
</head>
<body>
<h1>Hello, World!</h1>
</body>
</html>
Now if you just run:
yarn parcel index.html
You should get a URL that looks something like: http://localhost:1234/
.
Next you can start adding dependencies by specifying them in your code (however
your language specifies other assets). So for HTML we could create a
styles.css
file next to our index.html
file and include it with a <link>
tag.
h1 {
color: hotpink;
font-family: cursive;
}
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8" />
<title>My First Parcel App</title>
<link rel="stylesheet" href="./styles.css" />
</head>
<body>
<h1>Hello, World!</h1>
</body>
</html>
If we want parcel to update our changes in the browser without refreshing the page,
we need to add at least a dummy javascript file e.g. app.js
next to our index.html
.
This file allows parcel to inject all the necessary code to show your changes.
This file will later contain your javascript application.
console.log("Hello World");
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8" />
<title>My First Parcel App</title>
<link rel="stylesheet" href="./styles.css" />
<script src="./app.js"></script>
</head>
<body>
<h1>Hello, World!</h1>
</body>
</html>
- Introduction
- Parcel CLI
- CLI Args & Flags
parcel serve
parcel watch
parcel build
[...entries]
--target [name]
--open, -o [browser]
--host <host>
--port <port>, -p
--https
--dist-dir <dir>
--cache-dir <dir>
,--no-cache
--config <path>
--hot
,--no-hot
--[no-]source-maps
--autoinstall [npm/yarn], --no-autoinstall
--log-level <level>
--version, -v, -V
--help, -h
- CLI Args & Flags
- Parcel Config
- Parcel Architecture
- Asset Resolution
- Plugins
- Creating Plugins
- Plugin APIs
Parcel is a compiler for all your code, regardless of the language or toolchain.
Parcel takes all of your files and dependencies, transforms them, and merges them together into a smaller set of output files that can be used to run your code.
Parcel supports many different languages and file types out of the box, from web technologies like HTML, CSS, and JavaScript, to lower level languages like Rust, and anything that compiles to WebAssembly (WASM), to assets like images, fonts, videos, and more.
Parcel makes your code portable. You can build your code for different environments, for the web for your server, or for an app. You can even build multiple targets at once and have them live update as you make changes.
Parcel is fast and predictable. It compiles all of your files in isolation in parallel inside workers, caching all of them as it goes along. Caches are stable across machines and are only affected by the files and configs within your project (unless you want to pass specific environment variables).
The Parcel CLI is built into the main parcel
package. While you can install
it globally and run it, it is much better to install it locally into your
project as a dev dependency.
yarn add --dev parcel@next
You should also add some "scripts" to your package.json
to run it easier.
{
"name": "my-project",
"scripts": {
"build": "parcel build index.html",
"start": "parcel serve index.html"
},
"devDependencies": {
"parcel": "latest"
}
}
Now you can run yarn build
to bundle your project for production and
yarn start
to dev on your project.
Usage:
$ parcel [command] [...entries] [...flags]
Serve assets on a local server.
Watch and rebuild code on file changes.
Build code once, in production mode.
Entry files to start bundling, these will be preserved as entry points in the
output. Defaults to package.json#source
, falling back to src/index.*
or
index.*
. See #Sources.
Specifies a specific target to build. If unspecified, Parcel builds all targets specified in package.json. See #Targets.
Open your local server in a browser. You can optionally pass the name of the browser you want to open, otherwise it will use your default browser.
Configure the host to serve assets on. The default is to listen on all interfaces.
Configure the port to serve assets on. Alternatively you can use the $PORT
environment variable.
This will generate a local certificate (which will be untrusted by your
browser, you'll need to approve it) and serve your assets over https://
Specify the filepath to your SSL certificate when using --https
.
Specify the filepath to your SSL key when using --https
.
Configure the directory where compiled assets are output. Default is ./dist
.
Configure the cache directory with --cache <dir>
or disable it altogether
with --no-cache
.
specify which config to use. can be a path or a package name.
--config ../.myparcelrc
or --config @parcel/config-default
Turn hot reloading on or off.
Configure the hot reloading hostname.
Configure the hot reloading port.
Turn source maps on or off. Source maps are turned on by default.
When enabled, whenever Parcel discovers a dependency that isn't installed it
will attempt to install it with either npm or Yarn (defaults to npm unless a
yarn.lock
exists).
Set the log level, either either "none", "error", "warn", "info", or "verbose". The default is "info".
Return the current version of Parcel.
Get help with the CLI.
Parcel has always and will always work out of the box for many projects with zero configuration. It should always be extremely simple to get started. But if you do want more control, we give you the tools to do so.
A huge part of what Parcel does is run other tools over your code. Instead of
pulling all that configuration into Parcel, we make use of their own
configuration systems. So if you're using Babel, you should just use .babelrc
files to configure it.
When we do need to introduce config, we create tool specific config files in order to do so.
When you do need to configure Parcel, it will be in one of 3 places.
- If you need to configure the CLI, it will be a CLI flag
- If you need to configure your package, it will be in the
package.json
- If you need to configure something with your files or the Parcel asset
pipeline, it will be in
.parcelrc
[todo]
{
"name": "foo",
"main": "dist/main/index.js",
"module": "dist/module/index.js",
"browser": "dist/browser/index.js",
"browserslist": ["> 1%", "not dead"],
"engines": {
"node": ">=4.x"
},
"source": "src/index.js",
"targets": {
"main": {
"engines": {
"node": ">=4.x"
}
},
"module": {
"engines": {
"node": ">=8.x"
}
},
"browser": {
"engines": {
"browsers": ["> 1%", "not dead"]
}
}
},
"alias": {
"react": "preact-compat",
"react-dom": "preact-compat"
}
}
(Required) The name of the package is always required in order to be
considered a valid package.json
.
{
"name": "my-package"
}
(Required) All packages inside node_modules
must have a package.json#version
.
{
"version": "1.0.0"
}
This is the "main" target's entry point for the package, by default in library mode (doesn't bundle dependencies).
{
"main": "dist/main/index.js"
}
See Targets
This is the "module" target's entry point for the package, by default in library mode (doesn't bundle dependencies).
{
"module": "dist/module/index.js"
}
See Targets
This is the "browser" target's entry point for the package, by default in library mode (doesn't bundle dependencies).
{
"browser": "dist/browser/index.js"
}
See Targets
Specify the entry points for your source code which gets mapped to your targets.
{
"source": "src/index.js",
"source": ["src/index.js", "src/index.html"]
}
See Sources
As specified by Browserslist, this field is for specifying which transformers should be applied to browser bundles.
{
"browserslist": ["> 0.2%", "not dead"]
}
See Environments
Specify what versions of what engines you want to support.
{
"engines": {
"node": ">=4.x",
"electron": ">=2.x"
}
}
See Environments
Configuration for individual targets.
{
"targets": {
"main": {
"engines": {
"node": ">=4.x",
"electron": ">=2.x"
},
},
"browser": {
"engines": {
"browsers": ["> 1%", "not dead"]
}
}
}
}
See Targets
Aliases asset names/paths to other assets.
{
"alias": {
"react": "preact-compat",
"react-dom": "preact-compat"
}
}
See Aliases
Your .parcelrc
file will likely contain just a few fields (if you have one at
all), but here's an example of a .parcelrc
file that contains every field:
{
"extends": ["@parcel/config-default"],
"resolvers": ["@parcel/resolver-default"],
"transformers": {
"*.vue": ["@parcel/transformer-vue"],
"*.scss": ["@parcel/transformer-sass"],
"*.js": ["@parcel/transformer-babel"],
"*.css": ["@parcel/transformer-postcss"],
"*.html": ["@parcel/transformer-posthtml"]
},
"bundler": "@parcel/bundler-default",
"namers": ["@parcel/namer-default"],
"runtimes": ["@parcel/runtime-js", "@parcel/runtime-browser-hmr"],
"packagers": {
"*.js": "@parcel/packager-js",
"*.css": "@parcel/packager-css",
"*.html": "@parcel/packager-html",
"*.wasm": "@parcel/packager-wasm",
"*.raw": "@parcel/packager-raw"
},
"optimizers": {
"*.js": ["@parcel/optimizer-uglify"],
"*.css": ["@parcel/optimizer-cssnano"],
"*.html": ["@parcel/optimizer-htmlnano"],
"*.{png,jpg,jpeg,svg,...}": ["@parcel/optimizer-imagemin"]
},
"reporters": ["@parcel/reporter-cli"]
}
Many config properties like transformers
or packagers
use objects as maps of
globs to package names. While objects in JSON are technically unordered, Parcel
does use the order to give globs priority when a file name is being tested
against them.
{
"transformers": {
"icons/*.svg": ["highest-priority"],
"*.svg": ["lowest-priority"]
}
}
Here if we are trying to find a transform for the file icons/home.svg
, we'll
work our way down the globs until we find a match, which would be
icons/*.svg
, we never reach *.svg
.
extends
can either be a string or an array of strings that specify base
configs to extend. That base configuration can be the path to another
.parcelrc
file or the name of a Parcel config package.
{
"extends": "@parcel/config-default",
"extends": "../.parcelrc",
"extends": ["@parcel/config-default", "@company/parcel-config"]
}
When extending a config, Parcel does a shallow merge of the two configs.
resolvers
is an array of strings that specifies the name of a Parcel resolver
package.
{
"resolvers": ["@parcel/resolver-default"]
}
See Resolvers
transformers
is an object map of globs to arrays of Parcel transform packages.
{
"transformers": {
"*.js": ["@parcel/transformer-babel"]
}
}
See Transformers
bundler
is a string that specifies the name of a Parcel bundler package.
{
"bundler": "@parcel/bundler-default"
}
See Bundlers
namers
is an array of Parcel namer packages.
{
"namers": ["@parcel/namer-default"]
}
See Namers
runtimes
is an array of Parcel runtime packages.
{
"runtimes": ["@parcel/runtime-js", "@parcel/runtime-browser-hmr"],
}
See Runtimes
packagers
is an object map of globs to Parcel packager packages.
{
"packagers": {
"*.js": ["@parcel/packager-js"]
}
}
See Packagers
optimizers
is an object map of globs to arrays of Parcel optimizer packages.
{
"optimizers": {
"*.js": ["@parcel/optimizer-uglify"]
}
}
See Optimizers
reporters
is an array of Parcel reporter packages.
{
"reporters": ["@parcel/reporter-detailed"]
}
See Reporters.
validators
is an object map of globs to arrays of Parcel validator packages.
"validators": {
"*.ts": ["@parcel/validator-typescript"]
}
}
See Validators.
Even if you aren't doing anything that complex, if you are going to use Parcel a lot it makes sense to take some time and understand how it works.
At a high level Parcel runs through several phases:
- Resolving
- Transforming
- Bundling
- Packaging
- Optimizing
- (Validating)
The resolving and transforming phases work together in parallel to build a graph of all your assets.
This asset graph gets translated into bundles in the bundling phase.
Then the packaging phase takes the assets in the calculated bundles and merges them together into files each containing an entire bundle.
Finally, in the optimizing phase, Parcel takes these bundles files and runs them through optimizing transforms.
During the resolving and transforming phases, Parcel discovers all the assets in your app or program. Every asset can have its own dependencies on other assets which Parcel will pull in.
The data structure that represents all of these assets and their dependencies on one another is known as "The Asset Graph".
Asset Name | Dependencies |
---|---|
index.html |
app.css , app.js |
app.css |
N/A |
app.js |
navbar.js |
navbar.js |
etc. |
Once Parcel has built the entire Asset Graph, it begins turning it into "bundles". These bundles are groupings of assets that get placed together in a single file.
Bundles will (generally) contain only assets in the same language:
Bundle Name | Assets |
---|---|
index.html |
index.html |
app.css |
app.css |
app.js |
app.js , navbar.js , etc. |
Some assets are considered "entry" points into your app, and will stay as
separate bundles. For example, if your index.html
file links to an
about.html
file, they won't be merged together.
Bundle Name | Assets | Entry URL |
---|---|---|
index.html |
index.html |
/ |
about.html |
about.html |
/about |
"Sources" are the files that contain the source code to your app before being compiled by Parcel.
Parcel discovers these sources by following their dependencies on one another starting at your "entries".
These entries will be one of:
$ parcel <...entries>
~/package.json#source
./src/index.*
./index.*
From there, everything those assets depend on will be considered a "source" in Parcel.
When Parcel runs, it can build your asset graph in multiple different ways simultaneously. These are called "targets".
For example, you could have a "modern" target that targets newer browsers and a "legacy" target for older browsers.
Sources get mapped to targets,
In the most explicit form, targets are configured via the
package.json#targets
field.
{
"app": "dist/browser/index.js",
"appModern": "dist/browserModern/index.js",
"targets": {
"app": { /* target env */ },
"appModern": { /* target env */ }
}
}
Each target has a name which corresponds to a top-level package.json
field
such as package.json#main
or package.json#browser
which specify the primary
entry point for that target.
Inside each of those targets contains the target's environment configuration:
Option | Possible values | Description |
---|---|---|
context |
'node' | 'browser' | 'web-worker' | 'electron-main' | 'electron-renderer' |
Where the bundle should run |
includeNodeModules |
boolean | [String] |
Whether to bundle all/none/some node_module dependency |
outputFormat |
'global' | 'esmodule' | 'commonjs' |
Which type of imports/exports should be emitted |
publicUrl |
string |
The public url of the bundle at runtime |
isLibrary |
boolean |
Library as in 'npm library' |
sourceMap |
boolean | {inlineSources?: boolean, sourceRoot?: string, inline?: boolean} |
Enable/disable sourcemap and set options |
engines |
Engines | Same as package.json#engines |
However, a lot of the normal configuration you might want will already have defaults provided for you:
targets = {
main: {
engines: {
node: value("package.json#engines.node"),
browsers: unless exists("package.json#browser") then value("package.json#browserlist")
},
isLibrary: true
},
module: {
engines: {
node: value("package.json#engines.node"),
browsers: unless exists("package.json#browser") then value("package.json#browserlist")
},
isLibrary: true
},
browser: {
engines: {
browsers: value("package.json#browserslist")
},
isLibrary: true
},
...value("package.json#targets"),
}
Environments tell Parcel how to transform and bundle each asset. They tell Parcel if an asset is going to be run in a browser or in NodeJS/Electron.
They also tell Parcel's transform plugins how they should run. They tell Babel or Autoprefixer what browsers your asset is targetting.
You can configure environments through your targets.
{
"targets": {
"main": {
"engines": {
"node": ">=4.x",
"electron": ">=2.x",
"browsers": ["> 1%", "not dead"]
}
}
}
}
When one asset depends on another, the environment is inherited from its parent. But how you depend on the asset can change some properties of that environment.
For example:
navigator.serviceWorker.register('./service-worker.js');
let childEnvironment = {...parentEnvironment, browserContext: 'service-worker'};
Parcel will create a /.parcel-cache
directory. It will be filled with
directories with two letters, which are the start of a hash which is finished
by the names of the JSON files inside.
/.parcel-cache
/00/
213debd8ddd45819b79a3a974ed487.json
40ae9b581afc53841307a4b3c2463d.json
63a9dd58fc1e8f8bb819759ea9793c.json
...
/01/
/../
/zy/
/zz/
It follows this weird structure in order to avoid too many files being created in a single directory, which degrades file system performance.
Parcel follows the Node module resolution algorithm with a few additions.
./path/to/file
./path/to/file.js
These follow the Node module resolution algorithm.
preact
lodash/cloneDeep
@sindresorhus/is
These follow the Node module resolution algorithm.
https://unpkg.com/[email protected]/dist/preact.min.js
Parcel by default will ignore URL dependencies, other resolver plugins may choose to do something with them.
~/src/file.js
Only when used outside of node_modules
directories, the ~
is replaced by an
absolute path to the closest package root:
/path/to/app #(/package.json)
To form a path that looks like:
/path/to/app/src/file.js
Then it follows the Node module resolution algorithm.
Aliases come in two forms:
- Package Aliases:
react -> preact
- File/Directory Aliases:
utils
->./src/utils
{
"name": "my-project",
"alias": {
"react": "preact-compat",
"react-dom": "preact-compat",
"utils": "./src/utils",
"components": "./src/components"
}
}
There are a couple of rules:
- Aliases will only be respected when specified outside of
node_modules
. - Aliases specified outside of
node_modules
will affect assets inside ofnode_modules
. - Aliases cannot build off of other aliases.
- Only one alias will be applied at a time.
- Aliases must be valid npm package names.
When one asset depends on another through an asset specifier, the resolver is responsible for determining what asset is being requested.
See Asset Resolution for more details.
{
"resolvers": ["@parcel/resolver-v1"]
}
Official Resolvers:
@parcel/resolver-v1
transformers transform single assets as they are discovered and added to the asset graph. They mostly call out to different compilers and preprocessors.
{
"transformers": {
"*.js": ["@parcel/transformer-babel"]
}
}
Official Transformers:
@parcel/transformer-babel
@parcel/transformer-coffeescript
@parcel/transformer-glsl
@parcel/transformer-graphql
@parcel/transformer-json
@parcel/transformer-json5
@parcel/transformer-less
@parcel/transformer-posthtml
@parcel/transformer-postcss
@parcel/transformer-pug
@parcel/transformer-raw
@parcel/transformer-reason
@parcel/transformer-rust
@parcel/transformer-stylus
@parcel/transformer-toml
@parcel/transformer-typescript
@parcel/transformer-vue
@parcel/transformer-wasm
@parcel/transformer-webmanifest
@parcel/transformer-yaml
@parcel/transformer-elm
- ...
Bundlers accept the entire asset graph and turn it into sets of bundles.
{
"bundler": "@parcel/bundler-default"
}
Official Bundlers:
@parcel/bundler-default
Namers accept a bundle and return a filename for that bundle.
{
"namers": ["@parcel/namer-default"]
}
Official Namers:
@parcel/namer-default
Runtimes get called after the bundler phase and generate an asset which gets included in the final bundle.
{
"runtimes": ["@parcel/runtime-js", "@parcel/runtime-browser-hmr"]
}
Official Runtimes:
@parcel/runtime-js
@parcel/runtime-hmr
Packagers determine how to merge different asset types into a single bundle.
{
"packagers": {
"*.css": "@parcel/packager-css"
}
}
Official Packagers:
@parcel/packager-html
@parcel/packager-js
@parcel/packager-css
@parcel/packager-wasm
@parcel/packager-raw
Optimizers are similar to transformers, but they accept a bundle instead of a single asset.
{
"optimizers": {
"*.js": ["@parcel/optimizer-terser"],
"*.css": ["@parcel/optimizer-csso"]
}
}
Official Optimizers:
@parcel/packager-terser
@parcel/packager-csso
- [todo]
Reporters receive events as they happen and can either use the Parcel logger to output to stdout/stderr or they can return assets to be generated on the file system.
{
"reporters": ["@parcel/reporter-cli", "@parcel/reporter-dev-server"]
}
Official Reporters:
@parcel/reporter-cli
@parcel/reporter-dev-server
- [todo]
Validators emit errors for source code after a build is completed. For example, type checking and linting.
{
"validators": {
"*.ts": ["@parcel/validator-typescript"]
}
}
Official Validators:
@parcel/validator-typescript
@parcel/validator-eslint
- [todo]
All plugins must follow a naming system:
Official package | Community packages | Private company/scoped team packages | |
---|---|---|---|
Configs | @parcel/config-{name} |
parcel-config-{name} |
@scope/parcel-config[-{name}] |
Resolvers | @parcel/resolver-{name} |
parcel-resolver-{name} |
@scope/parcel-resolver[-{name}] |
Transformers | @parcel/transformer-{name} |
parcel-transformer-{name} |
@scope/parcel-transformer[-{name}] |
Bundlers | @parcel/bundler-{name} |
parcel-bundler-{name} |
@scope/parcel-bundler[-{name}] |
Namers | @parcel/namer-{name} |
parcel-namer-{name} |
@scope/parcel-namer[-{name}] |
Runtimes | @parcel/runtime-{name} |
parcel-runtime-{name} |
@scope/parcel-runtime[-{name}] |
Packagers | @parcel/packager-{name} |
parcel-packager-{name} |
@scope/parcel-packager[-{name}] |
Optimizers | @parcel/optimizer-{name} |
parcel-optimizer-{name} |
@scope/parcel-optimizer[-{name}] |
Reporters | @parcel/reporter-{name} |
parcel-reporter-{name} |
@scope/parcel-reporter[-{name}] |
Validators | @parcel/validator-{name} |
parcel-validator-{name} |
@scope/parcel-validator[-{name}] |
The {name}
must be descriptive and directly related to the purpose of the
package. Someone should be able to have an idea of what the package does simply
by reading the name in a .parcelrc
or package.json#devDependencies
.
parcel-transformer-posthtml
parcel-packager-wasm
parcel-reporter-graph-visualizer
If your plugin adds support for a specific tool, please use the name of the tool.
parcel-transformer-es6 (bad)
parcel-transformer-babel (good)
If your plugin is a reimplementation of something that exists, try naming it something that explains why it is a separate:
parcel-transformer-better-typescript (bad)
parcel-transformer-typescript-server (good)
We ask that community members work together and when forks happen to try and resolve them. If someone made a better version of your plugin, please consider giving the better package name over, have them make a major version bump, and redirect people to the new tool.
You must follow semantic versioning (to the best of your ability). No, it's not the perfect system, but it's the best one we have and people do depend on it.
If plugin authors intentionally don't follow semantic versioning, Parcel may start warning users that they should be locking down the version number for your plugin.
Warning: The plugin "parcel-transform-typescript" does not follow semantic versioning. You should lock the version range down so your code does not break when they make changes. Please upvote this issue to encourage them to follow semver: https://github.com/user/parcel-transform-typescript/issues/43
You must specify a package.json#engines.parcel
field with the version range
of Parcel that your plugin supports:
{
"name": "parcel-transform-imagemin",
"engines": {
"parcel": "2.x"
}
}
If you do not specify this field, Parcel will output a warning:
Warning: The plugin "parcel-transform-typescript" needs to specify a `package.json#engines.parcel` field with the supported Parcel version range.
If you do specify the parcel engine field and the user is using an incompatible version of Parcel, they will see an error:
Error: The plugin "parcel-transform-typescript" is not compatible with the
current version of Parcel. Requires "2.x" but the current version is "3.1.4"
Parcel uses node-semver to match version ranges.
There are several different types of plugins. They all look very similar, but are kept separate so we can have strict contracts one what each one is allowed to do.
There are some rules that should be followed across every type of plugin:
- Stateless β Avoid any kind of state, it will likely be the source of bugs for your users. For example, the same transform may exist in multiple separate workers which are not allowed to communicate with one another, state will not work as expected.
- Pure β Given the same input, a plugin must produce the same output, and you must not have any observable side effects, or implicit dependencies. Otherwise Parcel's caching will break and your users will be sad. You should never have to tell users to delete their caches.
The plugin APIs all follow a common shape:
import {NameOfPluginType} from '@parcel/plugin';
export default new NameOfPluginType({
async methodName(opts: JSONObject): Promise<JSONObject> {
return result;
}
});
They are made up of modules with well-known named exports of async functions that:
- Accept a strictly validated JSON-serializable
opts
object. - Return a strictly validated JSON-serializable
vals
object.
If something you need is not being passed through opts
, please come talk to
the Parcel team about it. Avoid trying to get information yourself from other
sources, especially from the file system.
Resolvers get called with an asset request (consisting of a source file path
and the specifier of what is being requested) which it then attempts to
resolve. If the resolver isn't sure how to handle a request, it can also return
null
and pass it to the next resolver in the chain.
import {Resolver} from '@parcel/plugin';
export default new Resolver({
async resolve({dependency}) {
// ...
return {filePath} || null;
}
});
Transformers transform single assets as they are discovered and added to the asset graph. They mostly call out to different compilers and preprocessors.
import {Transform} from '@parcel/plugin';
export default new Transform({
async parse({asset}) {
// ...
return ast;
},
async transform({asset}) {
// ...
return [assets];
},
async generate({asset}) {
// ...
return {code, map};
}
});
Bundlers accept the entire asset graph and modify it to add bundle nodes that group the assets into output bundles.
import {Bundler} from '@parcel/plugin';
export default new Bundler({
async bundle({graph}) {
// ...
},
async optimize({graph}) {
// ...
}
});
Namers accept a bundle and output a filename for that bundle.
import {Namer} from '@parcel/plugin';
export default new Namer({
async name({bundle, bundleGraph}) {
// ...
return name;
}
});
Runtimes accept a bundle and return assets to be inserted into that bundle.
import {Runtime} from '@parcel/runtime';
export default new Runtime({
async apply({bundle, bundleGraph}) {
// ...
return assets;
}
});
Packagers determine how to merge different asset types into a single bundle.
import {Packager} from '@parcel/plugin';
export default new Packager({
async package({bundle}) {
// ...
return {contents, map};
},
});
Optimizers are similar to transformers, but they accept a bundle instead of a single asset.
import {Optimizer} from '@parcel/plugin';
export default new Optimizer({
async optimize({bundle, contents, map}) {
// ...
return {contents, map};
}
});
Reporters receive events as they happen and can output to stdout/stderr, or perform other actions.
import {Reporter} from '@parcel/plugin';
export default new Reporter({
async report({ event: { type, ... } }) {
// ...
}
});
Validators receive an asset, and can throw errors if that asset is invalid in some way, e.g. type errors or linting errors.
import {Validator} from '@parcel/plugin';
export default new Validator({
async validate({asset}) {
// ...
throw error;
}
});
Some validators (such as @parcel/validator-typescript
) may wish to maintain a project-wide cache for efficiency. For these cases, it is appropriate to use a different interface where parcel hands all changed files to the validator at the same time:
import {Validator} from '@parcel/plugin';
export default new Validator({
async validateAll({assets}) {
// ...
throw error;
}
});
If your plugin implements validateAll
, Parcel will make sure to always invoke this method on the same thread (so that your cache state is accessible).
This project is licensed under the MIT License - see the LICENSE.md file for details