feat: add setMessagesCompiler method

[timofei-iatsenko]

Description

1. Allow users to use uncompiled catalogs on production (json)
2. have a proper working fallback to default message defined in the code (with the proper macro config)

Related user request: #2012

To achieve these goals, i18n.setMessagesCompiler() method was provided, which allow to manually register message compiler on production.

Also, the format of compiled message was slightly changed.

Previously, if the message doesn't have any ICU features inside it was printed as a string, array used for the rest of cases. That make distinguishing between compiled and uncompiled message particularly hard and sometimes lead to double compilation which was tried to be fixed here #1913

Now uncompiled message is always a string, and compiled is always an array.

Some examples:

"Hey '{name}'!" -> ["Hey {name}!"] // escaped placeholder
"Hey {name}!" -> ["Hey", "name", "!"] // icu placeholder
"Hey!" -> ["Hey!"] 

Types of changes

• Bugfix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Examples update

Fixes # (issue)

Checklist

• I have read the CONTRIBUTING and CODE_OF_CONDUCT docs
• I have added tests that prove my fix is effective or that my feature works
• I have added the necessary documentation (if appropriate)

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
js-lingui	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Oct 4, 2024 10:02am

[github-actions]

size-limit report 📦

Path	Size
./packages/core/dist/index.mjs	2.88 KB (0%)
./packages/detect-locale/dist/index.mjs	723 B (0%)
./packages/react/dist/index.mjs	1.65 KB (0%)
./packages/remote-loader/dist/index.mjs	7.26 KB (0%)

[timofei-iatsenko]

Had to exclude test from @lingui/remote-loader, because it's not published since version 3 and has outdated dependencies. That causes integration tests to fail, because snapshots are created using source code and new compiler format, but integration tests use packages from node_modules and for @lingui/remote-loader it takes stale versions.

[timofei-iatsenko]

Something happened to Verdaccio. The WF uses:

npm i -g {verdaccio,verdaccio-auth-memory,verdaccio-memory}

Which is installing latest possible version, they published version 6 couple days ago and that might be the reason. I tried to fix v5 version, but it didn't help. The error:

info attempt registry request try #1 at 10:04:12 AM
http request PUT http://0.0.0.0:4873/-/user/org.couchdb.user:test
info retry will retry, error on last attempt: Error: connect ECONNREFUSED 0.0.0.0:4873

However, the verdaccio itself is up and running:

[vonovak]

Hi, apologies for late reply. I have the same feedback as Andrii - docs :)

[timofei-iatsenko]

@andrii-bodnar @vonovak i added docs for that method. The better overall tutorial will be in next iterations after this epic finished

[vonovak]

Suggested change

	import { compileMessage } from "@lingui/message-utils/compileMessage";
	// `compileMessage` is Lingui's default implementation
	import { compileMessage } from "@lingui/message-utils/compileMessage";

[timofei-iatsenko]

It's coming from @lingui/ scope, isn't it obvious enough? More over, this method wasn't built to make it possible to override a default implementation, so it's intended to use always with a default implementation.

[vonovak]

this method wasn't built to make it possible to override a default implementation, so it's intended to use always with a default implementation.

That's a bit confusing to me because then I don't understand the use of this so much. But it's probably because other changes need to be done before this can be properly documented, as you said before.

isn't it obvious enough?

Well, just a few moments ago it felt obvious to me that I can use also something else other than the default implementation, so the word "obvious" can be a bit tricky.

If no other implementation should be used, the docs should say that imho.

Anyways, I don't want to block this PR from merging - I've raised the points I wanted to. 🙂👍
I'll defere to @andrii-bodnar