Vite 2 supports a very nice multi-page setup.
The given "multi-page" feature shares one main vite dev server for multiple nested pages, which may now use different src,
components,
etc.
Similar, a "multi-vite" setup should be able to serve multiple independent vite (or webpack, nuxt, [vite|vue]press) projects in parallel. A developer should even be able to add a new vite+vue3 setup to another existing legacy webpack+vue2 setup within the same webhost. Such a setup requires running a proxy (i.e.: Nginx) in front of all projects and setting the base according to your proxies location paths.
- tmuxinator to start the preconfigured tmux sessions.
- local Docker daemon up and running for demonstration purpose.
- ports 80 and 443 free to bind by Nginx.
Yes, we still use Makefiles.
make install
: installs node_modules and configures docker-compose for Nginxmake start
: start the whole site in dev modemake stop
: stop the whole site
After start, you should be able to point your browser to both
And edit components independently with superfast hmr.
For Demo, we set up a directory tree like this one:
├── Makefile
├── package.json
├── repos
│ ├── admin // 1st repository serving an admin ui (as PWA for example)
│ └── profile // 2nd repository serving a profile page (as SPA for example)
├── services
│ └── nginx // nginx config consumed by docker container
└── yarn.lock
Configure all vite instances to serve a baseUrl from their directory.
// ./repos/admin/vite.config.js
import vue from '@vitejs/plugin-vue'
export default {
base: '/admin/',
plugins: [vue()],
server: {
port: 8082 // <-- port dedicated to admin
}
}
// ./repos/profile/vite.config.js
import vue from '@vitejs/plugin-vue'
export default {
base: '/profile/',
plugins: [vue()],
server: {
port: 8081 // <-- port dedicated to profile
}
}
And connect those instances under the same domain with an Nginx proxy config like so:
location /admin/ {
proxy_pass http://host.docker.internal:8082;
}
location /profile/ {
proxy_pass http://host.docker.internal:8081;
}
With the base
option in place, there is no particular dev setup needed. Each run of:
$ yarn build
will build your project with a base path as prefix right into ./dist
. So, for example, the default index.html of admin repository will render like:
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<link rel="icon" href="/admin/favicon.ico" />
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Vite App</title>
<script type="module" crossorigin src="/admin/assets/index.068891ee.js"></script>
<link rel="modulepreload" href="/admin/assets/vendor.cee31988.js">
<link rel="stylesheet" href="/admin/assets/index.b77793a3.css">
</head>
<body>
<div id="app"></div>
</body>
</html>
To finally serve such a PWA on a production environment, Nginx will get configured with a location like:
location /admin/ {
alias /path/to/your/admin/dist;
try_files $uri $uri/ /admin/index.html;
}
- TBD
- since vite@2.0.0-beta.50: Droped extra locations for
/@vite/
and/node_modules/
in favor ofbase
option
- initially bootstrapped