5.1 KiB
vite-node
Vite as Node runtime.
The engine powers Vitest and Nuxt 3 Dev SSR.
Features
- On-demand evaluation
- Vite's pipeline, plugins, resolve, aliasing
- Out-of-box ESM & TypeScript support
- Respect
vite.config.ts
- Hot module replacement (HMR)
- Separate server/client architecture
- Top-level
await
- Shims for
__dirname
and__filename
in ESM - Access to native node modules like
fs
,path
, etc.
CLI Usage
Run JS/TS file on Node.js using Vite's resolvers and transformers.
npx vite-node index.ts
Options:
npx vite-node -h
Options via CLI
All ViteNodeServer
options are supported by the CLI. They may be defined through the dot syntax, as shown below:
npx vite-node --options.deps.inline="module-name" --options.deps.external="/module-regexp/" index.ts
Note that for options supporting RegExps, strings passed to the CLI must start and end with a /
;
Hashbang
If you prefer to write scripts that don't need to be passed into Vite Node, you can declare it in the hashbang.
Simply add #!/usr/bin/env vite-node --script
at the top of your file:
file.ts
#!/usr/bin/env vite-node --script
console.log('argv:', process.argv.slice(2))
And make the file executable:
chmod +x ./file.ts
Now, you can run the file without passing it into Vite Node:
$ ./file.ts hello
argv: [ 'hello' ]
Note that when using the --script
option, Vite Node forwards every argument and option to the script to execute, even the one supported by Vite Node itself.
Programmatic Usage
In Vite Node, the server and runner (client) are separated, so you can integrate them in different contexts (workers, cross-process, or remote) if needed. The demo below shows a simple example of having both (server and runner) running in the same context
import { createServer } from 'vite'
import { ViteNodeServer } from 'vite-node/server'
import { ViteNodeRunner } from 'vite-node/client'
import { installSourcemapsSupport } from 'vite-node/source-map'
// create vite server
const server = await createServer({
optimizeDeps: {
// It's recommended to disable deps optimization
disabled: true,
},
})
// this is need to initialize the plugins
await server.pluginContainer.buildStart({})
// create vite-node server
const node = new ViteNodeServer(server)
// fixes stacktraces in Errors
installSourcemapsSupport({
getSourceMap: source => node.getSourceMap(source),
})
// create vite-node runner
const runner = new ViteNodeRunner({
root: server.config.root,
base: server.config.base,
// when having the server and runner in a different context,
// you will need to handle the communication between them
// and pass to this function
fetchModule(id) {
return node.fetchModule(id)
},
resolveId(id, importer) {
return node.resolveId(id, importer)
},
})
// execute the file
await runner.executeFile('./example.ts')
// close the vite server
await server.close()
Debugging
Debug Transformation
Sometimes you might want to inspect the transformed code to investigate issues. You can set environment variable VITE_NODE_DEBUG_DUMP=true
to let vite-node write the transformed result of each module under .vite-node/dump
.
If you want to debug by modifying the dumped code, you can change the value of VITE_NODE_DEBUG_DUMP
to load
and search for the dumpped files and use them for executing.
VITE_NODE_DEBUG_DUMP=load vite-node example.ts
Or programmatically:
import { ViteNodeServer } from 'vite-node/server'
const server = new ViteNodeServer(viteServer, {
debug: {
dumpModules: true,
loadDumppedModules: true,
}
})
Debug Execution
If the process get stuck, it might because there is a unresolvable circular dependencies, you can set VITE_NODE_DEBUG_RUNNER=true
to vite-node warn about it.
VITE_NODE_DEBUG_RUNNER=true vite-node example.ts
Or programmatically:
import { ViteNodeRunner } from 'vite-node/client'
const runner = new ViteNodeRunner({
debug: true
})
Credits
Based on @pi0's brilliant idea of having a Vite server as the on-demand transforming service for Nuxt's Vite SSR.
Thanks @brillout for kindly sharing this package name.
Sponsors
License
MIT License © 2021 Anthony Fu