ESM -> WebAssembly Component creator,
via a SpiderMonkey JS engine embedding
A Bytecode Alliance project
Provides a Mozilla SpiderMonkey embedding that takes as input a JavaScript source file and a WebAssembly Component WIT World, and outputs a WebAssembly Component binary with the same interface.
Note: This is an experimental project, no guarantees are provided for stability or support and breaking changes may be made in future.
See the end-to-end example workflow for creating a JS component and running it in Wasmtime or Node.js.
For background on the concepts involved, see https://bytecodealliance.org/articles/making-javascript-run-fast-on-webassembly.
The goal of this project specifically is to provide a comprehensive dynamic bindings system for creating arbitrary WebAssembly Components from JavaScript. That is, to provide full flexibility over the resultant JS environment and WIT World.
Adaption follows the standard Wizer technique in pre-initializing a snapshot of the runtime against the source and bindings.
The snapshotting process executes the JS engine initialization, globals and parsing and compiling of the source code. Currently we also evaluate the top-level of the source so that the executed exports of the top-level ES module are provided already initialized.
As a result, at runtime - only the bytecode is being executed, without any initialization costs. This makes on-demand Wasm execution of JS incredibly fast.
As a dynamic language with quirks, JavaScript cannot be compiled directly into bytecode without including a comprehensive ECMA-262 spec-compliant runtime engine. Componentization of JavaScript thus involves embedding the JS runtime engine into the component itself.
SpiderMonkey is chosen here as a JS engine with first-class WASI build support, using an embedding of the StarlingMonkey Wasm engine. The total embedding size is around 8MB.
One of the security benefits of the component model is complete code isolation apart from the shared-nothing code boundaries between components. By fully encapsulating the engine embedding for each individual component, this maintains comprehensive per-component isolation.
As more components are written in JavaScript, and there exist scenarios where multiple JS components are communicating in the same application, the plan for optimization here is to share the SpiderMonkey engine embedding between them. This can be done without breaking the shared-nothing semantics by having the engine itself loaded as a shared library of the components. Sharing functions via same SpiderMonkey build, not memory.
Establishing this initial prototype as a singular flexible engine foundation that can be turned into a shared library is therefore the focus for this project.
The following APIs are available:
- Legacy Encoding:
atob
,btoa
,decodeURI
,encodeURI
,decodeURIComponent
,encodeURIComponent
- Streams:
ReadableStream
,ReadableStreamBYOBReader
,ReadableStreamBYOBRequest
,ReadableStreamDefaultReader
,ReadableStreamDefaultController
,ReadableByteStreamController
,WritableStream
ByteLengthQueuingStrategy
CountQueuingStrategy
,TransformStream
- URL:
URL
URLSearchParams
- Console:
console
- Performance:
Performance
- Task:
queueMicrotask
,setInterval
setTimeout
clearInterval
clearTimeout
- Location:
WorkerLocation
,location
- Encoding:
TextEncoder
,TextDecoder
,CompressionStream
,DecompressionStream
- Structured Clone:
structuredClone
- Fetch:
fetch
Request
Response
Headers
- Crypto:
SubtleCrypto
Crypto
crypto
CryptoKey
Install and run as a Node.js library:
npm install @bytecodealliance/componentize-js
import { componentize } from '@bytecodealliance/componentize-js';
import { writeFile } from 'node:fs/promises';
const { component } = await componentize(`
import { log } from 'local:hello/logger';
export function sayHello (name) {
log(`Hello ${name}`);
}
`, `
package local:hello;
interface logger {
log: func(msg: string);
}
world hello {
import logger;
export say-hello: func(name: string);
}
`);
await writeFile('test.component.wasm', component);
See types.d.ts for the full interface options.
The component iself can be executed in any component runtime, see the example for an end to end workflow in Wasmtime.
To enable AOT compilation, set the enableAot: true
option to run Weval ahead-of-time compilation.
To use a custom (pre-downloaded) weval
binary, set the wevalBin
option to the path to your desired weval binary.
To support asynchronous operations, all functions may optionally be written as sync or async functions, even though they will always be turned into sync component functions.
For example, to use fetch
which requires async calls, we can write the same example component using an async function:
export async function sayHello (name) {
const text = await (await fetch(`http://localhost:8080/${name}`)).text();
console.log(text);
}
ComponentizeJS will automatically resolve promises returned by functions to syncify their return values, running the event loop within the JS component to resolution.
This asynchrony is only supported for exported functions - imported functions can only be synchronous pending component-model-level async support.
ComponentizeJS can be used as a CLI from jco
:
npm install -g @bytecodealliance/jco @bytecodealliance/componentize-js
For example:
jco componentize source.js --wit wit -o component.wasm
See jco componentize --help
for more details.
The set of enabled features in the engine can be customized depending on the target world and expected capabilities.
The default set of features includes:
'stdio'
: Output to stderr and stdout for errors and console logging, depends onwasi:cli
andwasi:io
.'random'
: Support for cryptographic random, depends onwasi:random
. When disabled, random numbers will still be generated but will not be random and instead fully deterministic.'clocks'
: Support for clocks and duration polls, depends onwasi:clocks
andwasi:io
. When disabled, using any timer functions like setTimeout or setInterval will panic.'http'
: Support for outbound HTTP via thefetch
global in JS.
Setting disableFeatures: ['random', 'stdio', 'clocks', 'http']
will disable all features creating a minimal "pure component", that does not depend on any WASI APIs at all and just the target world.
Note that pure components will not report errors and will instead trap, so that this should only be enabled after very careful testing.
Note that features explicitly imported by the target world cannot be disabled - if you target a component to a world that imports wasi:clocks
, then disableFeatures: ['clocks']
will not be supported.
The StarlingMonkey engine provides the ability to use fetchEvent
to handle calls to wasi:http/incoming-handler@0.2.0#handle
. When targeting worlds that export wasi:http/incoming-handler@0.2.0
the fetch event will automatically be attached. Alternatively, to override the fetch event with a custom handler, export an explict incomingHandler
or 'wasi:http/incoming-handler@0.2.0'
object. Using the fetchEvent
requires enabling the http
feature.
export function componentize(jsSource: string, opts: {
witPath: string,
worldName: string,
debug?: bool,
sourceName?: string,
engine?: string,
preview2Adapter?: string,
disableFeatures?: ('stdio' | 'random' | 'clocks' | 'http')[],
}): {
component: Uint8Array,
imports: string[]
}
http
provides support for the host APIs used by the fetch
method and is disabled by default,
while this API is still being developed. Contributions very welcome to improve fetch
support.
Converts a JS source into a component binary.
Imports provides the list of used guest imports only, while the StarlingMonkey engine may pull in additional imports. Direct component analysis should be used to correctly infer the real imports list.
git submodule update --init --recursive
to update the submodules.- Stable Rust with the
wasm32-unknown-unknown
andwasm32-wasi
targets installed. wasi-sdk-20.0
installed at/opt/wasi-sdk/
Building and testing the project can be performed via NPM scripts (see package.json
):
npm install
npm run build
npm run test
Before being able to use componetize-js
(ex. via npm link
, from jco
), you'll need to run:
npm run build:weval
This will produce a few files, most importantly lib/starlingmonkey_embedding_weval.wasm
.
To clean up a local installation (i.e. remove the installation of StarlingMonkey):
npm run clean
This project is licensed under the Apache 2.0 license with the LLVM exception. See LICENSE for more details.
Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in this project by you, as defined in the Apache-2.0 license, shall be licensed as above, without any additional terms or conditions.