feat: implement createLoadTranslations factory function

[lunarW1TCH]

I added the getFromMessages method on i18n object, a ready solution for server-side rendering instead of manually setting the cache or triggering the update and then waiting.

[ai]

It is 10% size increase.

1. Maybe we should do it as function, not a method (so we will not put it to client-side JS bundle)?
2. Also, I would like a more clear name, right now it is not clear for me

[lunarW1TCH]

I wanted to make it a separate method initially but opted for method bound to the i18n object to use the existing logic.
Maybe for a separate function I could do something like:

const get = async () => {}
const i18n = createI18n(locale, { get })

export const getTranslations = createGetTranslations(i18n, get) // factory function 

Get translations would still accept messages store, fetch translations, set them in cache and then return the translations. Is this a better approach? Do you think createGetTranslations is a fitting name?

[ai]

Maybe we can call is loadTransaltions/loadAllTranslations or something with load verb, because we already have messages.get() (with get verb) and need to explain the difference.

[lunarW1TCH]

loadTranslations should be fine, I will work on this new implementation tomorrow then

[lunarW1TCH]

I rewrote the logic into a separate factory function

[ai]

Do we need to pass locale? I think we can use it from i18n.locale

[lunarW1TCH]

you are completely right here, i didn't notice that i18n exposes locale

[ai]

What is use cases for that? SSR?

Why do we use it in React components?

[lunarW1TCH]

It's for rendering async server components which aren't updated on the client side. The translation must be available before server sends the rendered html to client.

async function Component(): Promise<React.ReactNode> {
 // ...
}

It's different from pre-rendered client components which are later hydrated with loaded translation.

[lunarW1TCH]

https://react.dev/reference/rsc/server-components

[ai]

Nice explanation, let’s put it to the docs.

[ai]

If we use cache do we need to check it (and return cached version is available)?

[lunarW1TCH]

I'll add checks for it

[lunarW1TCH]

I have simplified the logic, calling messages.get() begins the process of loading translations for the specific component, then i am using the built in translationsLoading function to block the return until they load. No need for passing anything else than i18n object.

i marked first get call with void to make it clearer that the return value for that line is not used and only the side-effect matters.

[ai]

Looks good to me.

But I am a little busy on this week. I will try to merge and release it a little later.

[ai]

After thinking about it, I hope to find way to simplify API: c94fd4d

[ai]

loadTransaltions() was released in 1.3.0. Thanks for the idea.