[labs/context] SSR compatibility

[justinfagnani]

Should this be an RFC?

• This is not a substantial change

Which package is this a feature request for?

Context (@lit-labs/context)

Description

Before graduating @lit-labs/context and @lit-labs/ssr we should implement context support in SSR. This may be done via general event support in SSR as in #2309, or if that's not feasible, in some other way.

Alternatives and Workarounds

None

[hrmcdonald]

It sounds like #3365 will help address this, but until it does, could @lit-labs/context be updated to not break SSR by not attempting to attach event listeners during SSR via the use of isServer checks?

If the priority of this is high enough that event support will be coming to SSR sooner than later this probably isn't necessary though.

[dunqan]

We really support @lit-labs/context idea, and it would be really nice to have it working seamlessly with SSR, but our main worries are:

1. It uses an Event based protocol, super nice solution on CSR, doesn't look so nice on SSR, but it looks like Lit is trying to push SSR support for Events because of this.

If lit SSR is a liniear process, idea of dispatching event to the parent (when the parent is partially rendered already) looks like not complete one (but I agree, will work for limited use cases, like context implementation).

2. What about partial hydration, when child component was hydrated before parent was able to setup event listeners?

Did you consider any mechanism of serializing context, so child ements are able to pick it up from the DOM on hydration, without having to hydrate parent first?

[bennypowers]

shouldn't the provide controller attach the listener in hostConnected anyways, rather than in the constructor? Our context implementation does not crash SSR, so long as DOM *Observers are shimmed.

https://github.com/RedHat-UX/red-hat-design-system/blob/cbd5fec0099cd6c8bfedc8b3c1bfd4a94969d4a3/lib/context/color/provider.ts#L104-L110

Why do the listeners need to be attached in the constructor?

[justinfagnani]

@sorvell I'm not sure this should be closed yet. Event support is opt-in and not documented.

[kyubisation]

Sorry, that's on me, as I referenced this issue in my PR. Feel free to re-open this.
I can try to add appropriate documentation to lit.dev in January.

[kyubisation]

lit/lit.dev#1390