[fields] formatDate's 'short' compact form keeps a Western field order after localization #4543

Observation-class, recorded so it is not rediscovered as a surprise. No user is blocked; nothing here is a regression.

What it is

#4272 made formatDate's 'short' branch honor the threaded locale, but deliberately localized only the month token — the compact shape around it is assembled by hand:

const month = date.toLocaleDateString(options?.locale, { month: 'short' });
const day = date.getDate();
const year = String(date.getFullYear()).slice(-2);
return `${month} ${day}, '${year}`;

So a zh session renders 8月 15, '26: a localized month inside a month-day-year layout with a Western comma and apostrophe-year. zh's own compact convention would be year-first.

This was the ruled scope for #4272 and is correct as far as it goes — the alternative, handing the whole thing to Intl.DateTimeFormat, changes the English output too (Aug 15, '26 becomes Aug 15, 26, losing the apostrophe), which that card pinned as byte-identical must-not-change. So the mixed form is a deliberate consequence, not an oversight.

Why it might still be worth a decision later

The compact form exists for narrow mobile cards, where the apostrophe-year is a density trick rather than a locale convention. Whether a zh / ja reader should get their own compact convention (26/8/15) instead of a localized token in a Western frame is a product call, and it trades against the byte-identical English pin. Recording it as an observation rather than proposing a change.

Only consumers today are ObjectGrid's two mobile-card date cells. Context: #4272, #4468 / PR #4512.

Generated by Claude Code