Fix https://github.com/vector-im/hydrogen-web/issues/686
Fix https://github.com/vector-im/hydrogen-web/issues/682
Instead of deleting the whole `target/` directory, leave it alone so the symlink
driving the `npm link`/`yarn link` stays in tact.
Leave Vite builds in their build directories (`/lib-build`/`/asset-build`)
so you can `vite build --watch` to build on local changes and still have a
consisent place to reference in the `package.json` `exports`. Previously,
everything relied on `build.sh` which does a bunch of moving and renaming
and made it hard to rebuild on changes.
Add back support for CommonJS (adding the `package.json` `exports`).
The last piece is making sure the `?url` imports (`import workerPath from 'hydrogen-view-sdk/main.js?url';`)
work still. It looks like this may have just been solved via
https://github.com/vitejs/vite/issues/6725 -> https://github.com/vitejs/vite/pull/7073
(literally 2 days ago) and we just need to wait for the next Vite release 🎉
- [`linkedom`](https://github.com/WebReflection/linkedom) is being used https://github.com/matrix-org/matrix-public-archive to server-side render (SSR) Hydrogen (`hydrogen-view-sdk`)
- This is being fixed by using a explicit HTML wrapper boilerplate with `DOMParser` to get a matching result in the browser and `linkedom`.
Currently `parseHTML` is only used for HTML content bodies in events. Events with replies have content bodies that look like `<mx-reply>Hello</mx-reply> What's up` so they're parsed as HTML to strip out the `<mx-reply>` part.
Before | After
--- | ---
 | 
Before:
```js
// Browser (Chrome, Firefox)
new DOMParser().parseFromString(`<div>foo</div>`, "text/html").body.outerHTML;
// '<body><div>foo</div></body>'
// `linkedom` ❌
new DOMParser().parseFromString(`<div>foo</div>`, "text/html").body.outerHTML;
// '<body></body>'
```
After (consistent matching output):
```js
// Browser (Chrome, Firefox)
new DOMParser().parseFromString(`<!DOCTYPE html><html><body><div>foo</div></body></html>`, "text/html").body.outerHTML;
// '<body><div>foo</div></body>'
// `linkedom`
new DOMParser().parseFromString(`<!DOCTYPE html><html><body><div>foo</div></body></html>`, "text/html").body.outerHTML;
// '<body><div>foo</div></body>'
```
`linkedom` goal is to be close to the current DOM standard, but [not too close](https://github.com/WebReflection/linkedom#faq). Focused on the streamlined cases for server-side rendering (SSR).
Here is some context around getting `DOMParser` to interpret things better. The conclusion was to only support the explicit standard cases with a `<html><body></body></html>` specified instead of adding the magic HTML document creation and massaging that the browser does.
- https://github.com/WebReflection/linkedom/issues/106
- https://github.com/WebReflection/linkedom/pull/108
---
Part of https://github.com/vector-im/hydrogen-web/pull/653 to support server-side rendering Hydrogen for the [`matrix-public-archive`](https://github.com/matrix-org/matrix-public-archive) project.
