mirror of
https://github.com/hanzoai/i18n.git
synced 2026-10-30 14:24:32 +00:00
zeekay
0e90bb83c9
The extractor reads the children of <T> as keys. JSX text is normalized by the compiler before T sees it, so the key is SWC's rule applied to the source (tools/extract/src/jsx.ts): entities decoded, lines cut at every terminator, spaces and tabs trimmed where they touch a break unless an entity wrote them, the rest joined by one space. Each case is compiled with esbuild and held equal to the children T receives; the same cases and the entity edges were measured against Next 16.3's SWC. useI18n().format (and format(locale) from the root) is Intl at the locale: number, percent, money, date, relative and list, with Intl's options passed through, so en-US output is toLocaleString's byte for byte. Right to left every result is one isolate and a list isolates each item, so Latin names read right to left; isolate leaves a whole isolate alone, so a figure handed to t carries one pair of marks. The browser test reads Arabic and ar-EG digits, one isolate per figure, and the list's order, which fails without the item isolates. Claude-Session: https://claude.ai/code/session_01BLLMyBFSvAfCt7ZGqYuiEz