"Day keyword + relative time" combined expressions don't work in Portuguese, Japanese, or Chinese

[Amato21]

Context

PR #39 fixed getParsedDateResult() picking the wrong candidate when chrono-node returns multiple disjoint matches for one string (e.g. English "today in 3 minutes" / French "aujourd'hui dans 3 minutes" used to give the right date but the current time instead of +3 minutes).

That fix generalizes automatically to most languages since the root cause (only ever checking results[0]) was language-agnostic. Verified across all 11 supported languages:

Works correctly (+3 min): English, French, Spanish, Italian, Russian, Ukrainian, German, Dutch.

Does not work: Portuguese, Japanese, Chinese.

Root cause (different from PR #39's bug)

For these 3 languages, chrono-node's own locale parser doesn't recognize the relative-time portion of the combined phrase as a match at all — verified with raw chrono-node calls:

chrono.pt.casual.parse('hoje em 3 minutos', ...)   // -> only "hoje", "em 3 minutos" is dropped entirely
chrono.ja.casual.parse('今日あと3分', ...)           // -> only "今日", "あと3分" is dropped entirely
chrono.zh.hant.casual.parse('今天在3分鐘', ...)      // -> only "今天", "在3分鐘" is dropped entirely

Since chrono-node itself only ever returns one candidate for these languages, PR #39's fix (comparing every candidate to pick the best one) has nothing to choose between — there's no second, better candidate to find.

Possible fix

Build a dedicated custom regex pattern for "day keyword + relative time" (similar to the existing regexWeekdayWithTime pattern for "next Monday at 3pm"), generated per-language via TranslationCollector, instead of relying on chrono-node's fallback for this specific combination. This would make the behavior consistent across all languages instead of depending on chrono-node's per-locale completeness.

Scope note

Standalone "in 3 minutes" (no day keyword) already works correctly in every language — this only affects the combined phrasing.

[Amato21]

Korean joins this list. It has no chrono-node locale at all (unlike Portuguese/Japanese/Chinese, which at least have their own chrono-node parser for some patterns) — verified: chrono-node ships de/en/es/fi/fr/it/ja/nl/pt/ru/sv/uk/zh, no ko. So Korean relies entirely on this plugin's own regex patterns, with no chrono-node fallback for anything they don't cover.

What works (verified empirically): 오늘/내일/어제, 지금, weekday names, 이번/다음/지난 + weekday (this/next/last, prefix position), and suffix-style relative expressions (3일 후, 2주 전) — Korean naturally suffixes the relative-time marker after the number+unit, the same mechanism already used for Chinese's 後/后 (later/agosuffix keys).

What doesn't (also verified): weekday + specific time (e.g. 다음 월요일 에 3시) silently drops the time instead of erroring, and date ranges (e.g. 월요일 부터 금요일 까지) don't match at all — likely because Korean expresses "at"/"from"/"to" as particles attached directly to the previous word (no space), while this plugin's regex patterns expect them as a space-separated standalone word. That's a structural mismatch with the current template system, not just a missing translation.

Given three (now four) languages hit variations of "chrono-node has no/partial locale support, so combined expressions need dedicated custom regex instead of relying on chrono's fallback," might be worth building the generalized TranslationCollector-based pattern mentioned in this issue's description once, rather than one-off fixes per language. Not scheduled — noting it here per the maintainer's request so it isn't lost.

---

Generated by Claude Code