PO Gettext extract does not generate correct plural placeholders

[garikkh]

Description

When running lingui extract with locales that have different plural rules, we would expect the PO to have the correct number of msgstr for the number of plurals, but instead it just uses two (source locale, not even source code. Even if your source code has zero, one, many it will only put msgstr[0], msgstr[1]

Verifications

• I've checked the docs and this isn't covered there.
• I've searched existing issues on GitHub.

Reproduction Steps

Add any plural

        <div>
          <Plural
            value={count}
            one={"There's one book"}
            other={"There are # books"}
          />
        </div>

        <div>
          {plural(count, {
            one: "one book",
            other: "many books",
          })}
        </div>

And use ru in locales - good example, as it has 4 plural forms, and rule is:

"Plural-Forms: nplurals=4; plural=(n%10==1 && n%100!=11 ? 0 : n%10>=2 && n%10<=4 && (n%100<12 || n%100>14) ? 1 : n%10==0 || (n%10>=5 && n%10<=9) || (n%100>=11 && n%100<=14)? 2 : 3);\n"

Expected Behavior

Expect to see:

#. js-lingui:icu=%7Bcount%2C+plural%2C+one+%7BThere%27s+one+book%7D+other+%7BThere+are+%23+books%7D%7D&pluralize_on=count
#: src/App.tsx:52
msgid "There's one book"
msgid_plural "There are # books"
msgstr[0] ""
msgstr[1] ""
msgstr[2] ""
msgstr[3] ""

Instead only get

#. js-lingui:icu=%7Bcount%2C+plural%2C+one+%7BThere%27s+one+book%7D+other+%7BThere+are+%23+books%7D%7D&pluralize_on=count
#: src/App.tsx:52
msgid "There's one book"
msgid_plural "There are # books"
msgstr[0] ""
msgstr[1] ""

Macro Support

Babel with babel-macro-plugin

Lingui Version

5.5.1

Babel Version

No response

Framework

No response

[timofei-iatsenko]

I believe it's not a formatter responsibility at all, it's a responsibility of software managing your catalogs. Crowdin / PoEditor / etc.

But that could be easily changed here:

js-lingui/packages/format-po-gettext/src/po-gettext.ts

Lines 108 to 123 in 1cb5ecd

	// If there is a translated value, parse that instead of the original message to prevent overriding localized
	// content with the original message. If there is no translated value, don't touch msgstr, since marking item as
	// plural (above) already causes `pofile` to automatically generate `msgstr[0]` and `msgstr[1]`.
	if (message.translation?.length > 0) {
	const ast = parseIcu(message.translation)[0] as Select
	if (ast.cases == null) {
	console.warn(
	`Found translation without plural cases for key "${id}". ` +
	`This likely means that a translated .po file misses multiple msgstr[] entries for the key. ` +
	`Translation found: "${message.translation}"`
	)
	item.msgstr = [message.translation]
	} else {
	item.msgstr = ast.cases.map(stringifyICUCase)
	}
	}

Just populate an item.msgstr with empty elements equal to ast.cases.length

[garikkh]

I believe it's not a formatter responsibility at all, it's a responsibility of software managing your catalogs. Crowdin / PoEditor / etc.

I agree - and the way we use lingui currently that's how it works - we extract english (source) locale, and upload that to TMS, which generates the other ones. I only noticed this when building #2344